IA CRYPTOGRAPHIC SERIES

THE INDEX OF COINCIDENCE

AND ITS APPLICATIONS
IN CRYPTANALYSIS
by
William F. Friedman
From Aegeon Pork Press
THE INDEX OF COINCIDENCE
AND ITS APPLICATIONS
IN CRYPTANALYSIS
by
William F. Friedman
MiDDLEBURY COLLEGE LIBRAS*
1987 AEGEAN PARK PRESS
ISBN: 0-89412-137-5
ISBN: 0-89412-138-3
(soft cover)
(library bound)
AEGEAN PARK PRESS
P.O. Box 2837
Laguna Hills, California 92654
(714) 586-8811
Manufactured in the United States of America
TABLE OF CONTENTS
Introduction
Part I
The Vogel Quintuple Disk
Part II
The Schneider Cipher 65
A Special Problem
89
THE INDEX OF COINCIDENCE AND ITS APPLICATIONS IN CRYPTANALYSIS
INTRODUCTION
Frequency tables in the analysis and solution of ciphers have commonly been employed
to make assumptions of plain-text equivalents for the cipher letters constituting a message.
The significance of the various phases of the curves themselves, i.e., the crests and troughs
and their relative positions in such frequency tables, has been recognized to some extent, but
largely only in connection with the determination of two more or less preliminary points in
their analysis: (1) whether the frequency distribution approximates that of a substitution
cipher involving only one alphabet or more than one alphabet; (2) whether this approximation
corresponds to that of a standard alphabet, direct or reversed, or that of a mixed alphabet.
It will be shown in this paper that the frequency tables of certain types of ciphers have
definite characteristics of a mathematical or rather statistical nature, approaching more or less
closely those of ordinary statistical curves. These characteristics may be used in the solution
of such ciphers to the exclusion of any analysis of the frequencies of individual letters consti-
tuting the tables or curves, and without any assumptions whatever of plain-text values Jor the
cipher letters.
It is true that cipher systems admitting of such treatment are not very commonly encoun-
tered. But inasmuch as such systems are always of a complex nature, which the ordinary
methods of cryptanalysis would find rather baffling, a description of a purely mathematical
analysis that may be applied to other cases similar to the ones herein described may be con-
sidered valuable. In fact, it is possible that the principles to be set forth may find con-
siderably wider application in other phases of cryptanalysis than is apparent at this time.
Two examples of such a treatment will be given in detail: One dealing with a substitution
cipher wherein a series of messages employing as many as 125 random mixed secondary alpha-
bets can be solved without assuming a plain-text value for a single cipher letter; the other a
multiple alphabet, combined substitution-transposition cipher, solved from a single message
of fair length.
(1)
PART I
THE VOGEL QUINTUPLE DISK CIPHER '
This cipher system involves the use of J ive superimposed disks bearing dissimilar random
mixed alphabets. These disks are mounted upon a circular base plate, the periphery of which
is divided into 26 segments; one of these is marked "Plain", indicating the segment in l i ne
with which the successive letters of the plain text as found on the five revolving disks are to
be brought for encipherment. The remaining 25 segments of the base plate bear the numbers
from 1 to 25 in a mixed sequence, which we have called the "Numerical key." This key may,
however, consist of less than 25 numbers, in which case one or more of the segments of the base
plate will remain blank. The numbers constituting the key are written on the base plate in u
clockwise direction beginning immediately at the right of the plain segment (fig. 1).
MET HOD OF ENCIPHEHMENT
In the accompanying example illustrating the details of encipherment it will be seen that the
numerical key consists of 21 numbers, leaving blank, therefore, the four segments immediately
preceding the plain segment.
2
Assuming a series of messages, let us suppose the first three
begin as follows:
1. Prepare for bombardment at Harvey . . .
2. Enemy attack on Hunterstown . . .
3. Second Field Artillery Brigade . . .
Revolving the five cipher disks successively, and thus bringing the first 5 letters of message
1, PREPA, in line with the plain segment, reading from the outer disk inward in the order 1-2-3-
4-5, the cipher letters for this first set of 5 plain-text letters are then taken in the same order from
the segments of the disks directly in line with that segment of the base plate that bears the number 1.
In this case it is the eighteenth segment after the plain, in a clockwise direction, and, as shown
in figure 1, the equivalent cipher letters for this group are MEKJ R. The second set of live plain-
text letters of message 1, REFOR, are then in a similar manner set in line under the plain segment,
and their equivalent cipher letters are taken from the segment immediately following segment
1 of the numerical key, in a clockwise direction, viz, segment 13. The cipher letters in this case
are VZQWH. The third group of letters in message 1 finds its cipher equivalents at segment
1
W hile on duty in the Code and Cipher Section of the Intelligence Division of the General St uf f , G.H.Q .,
A.E.F., Lt. Col. F. Moorman, Chief of Section, turned over to the writer for st udy a cipher system t oget her
with a series of 26 test messages submitted by Mr. E. J. V ogel, Chief Clerk, who had taken considerable interest
in cryptography and had, as a result of his studies, devised t he system presented for exami nat i on. T he writer
worked upon the cipher during his leisure moments, but the problem involved considerable labor and solution
was not completed before being relieved from dut y at t hat st at i on. T he mai n pri nci pl es for s ol ut i on, however,
were established and only the detailed work remained to be completed. Af t er mi interval of more than a year,
while Director of the Cipher D epartment of t he Ri ver bank Laboratories, Geneva, 111., the wr i t e r t ur ned his
attention once more to this cipher and succeeded in completely solving the problem by carryi ng out those
principles to their logical conclusion.
It is recommended that the reader prepare a duplicate of t he set of disks in order t hat he may more readily
follow the various steps in the analysis.
(1!)
18; the fourth, at segment 8. The fifth group of plain-text letters, however, will take its
cipher equivalents from the first segment to the right of the plain segment, inasmuch as the
segments immediately following segment 8 are blank. Plain-text letters TATHA, therefore, will
be enciphered on segment, 9, becoming XONJE. This method is continued in like manner through-
out message 1. Jf message 1 contains more than 21 groups of 5 letters, the twenty-second group
will take its cipher equivalents on segment 1 again; the twenty-third.on segment 13, and so on.
F IGURE 1
Proceeding now to message 2, the first 5 plain-text letters, ENEMY, are set up under the plain
segment and the cipher letters are taken from segment 2, becoming LTVBM. The second group
of 5 letters is enciphered on segment 1 , the third group on segment 13, and so on, throughout
the message. The first 5 letters of message 3, SECON, are enciphered on segment 3, becoming
Y APAC. The first group of cipher letters in n message is always to be taken from the segment
bearing the number which coincides with the serial or accession number of the message in the
day's activity.
It will be seen, therefore, t hat no matter what repetitions occur in the plain-text beginnings
of messages, the cipher letters will give no evidences of such repetitions, for each message has a
different starting poi nt , det er mi ned, ns said before, by the serial number of the message in the
II
4
day's activity. This automatic prevention of initial repetitions in the cipher text is true, how-
ever, only of a number of messages equal to the length of the numerical key; for, with a number of
messages greater than the length of the key, the initial segments from which the cipher letters
are to be taken begin to repeat. In this case, message 22 would necessarily have the same initial
point as message 1; message 23, the same as message 2, and so on. Messages 43, 64, 85, . . . ,
would all begin with segment 1; messages 44, 65, 86, . . . with segment 2, and so on.
The secrecy of the system is dependent upon a frequent change of alphabets and a still more
frequent change of numerical key, since it must be assumed, as with all cipher systems or devices,
that sooner or later the general method of encipherment will become known to the enemy.
The only reliance, therefore, for the safety of the messages must be placed in keeping the specific
alphabets and the numerical key for a given series of messages from the enemy.
PRINCIPLES OF SOLUTION
The following messages ' are assumed to have been intercepted within one day and therefore
to be in the same alphabets and key:
MESSAGE No. 1
MLVXK QNXVD GIRIE IMNEE FEXVP HPVZR UKSEK MVQCI VXSFW
GVART YBZKJ WVUPV XZCBD BDOLS GHINZ LJCTE KSLPY VPBYD
WTRJK BDDFA ANJXE XGHED ERYVP YPWDJ DFTJV ZHTWB WXTMF
OZDOJ
MESSAGE No. 2
ULJCY GXAEU DTEIL UZBRW GJZSS QLUOX PTFOO NWSHD BPTJO
HQRYY YAXRZ KTEMP UAYMK ISRDZ VUVKW HXAYD YAGSM CURBZ
LBXOV EBBPI BMLCB UMAXF ZSLXV QFXUE MPZMK MQZZT KMURW
EJVB
MESSAGE No. 3
YLMKW CBGSF VGABP HOZFV QQNSQ NQLQL DIGXM XCWAI QFJOQ
TYDEL MBMJB SEPSO DHREM ELKIP KXNMW QYBIH BHFDC GLYWC
YGMMP EEXZH UBBSB SBONG URQKW YRAYU NYUCS LNEMV VNSXN
WGVME MXPDF WGTZE KRLGU ZJFZJ W
MESSAGE No. 4
UFHUJ LTMKJ PONFG RIUGG OZGWS UBNMW WGILB JNXTD BPREX
MMWHB OBFVO TFGSJ SLXEH RTZMI LLUUX FIFWC PGSBA KRCAS
XWKQV SLDKS NTESD QVQBN RZDMB JQLYH LTMXB BSVWL ILKYU
NMFEB HB
MESSAGE No. 5
VJZCQ KZJJK DCVQD KSSXY TSUXE GRROH PXKZF ZKFMS VGDWU
XLTBW EXHEF AWQWF ZESMX GWCEM JPNVB GRJGB IBWOH YMOAP
IZYNX GXMSB OZZGK FVURN NJFGQ DTPLV STOID DWVLR TXTBH
WNWIE YJXXW BKOJQ FOUHO T
J
These are the actual messages submitted by Mr. V ogel.
MESSAGE No. 6
XGORF GCHAX DUEIQ XAOWK BKBUH SKWWW XNEYZ
IWTVU NPCLU KQDIG YLMNC KYJJF SDSTU DCBUK
VGQSW OGHPI XCYIO SUAEU BQAPY RMDMW FWGSQ
LNGQT IPJRI HUMAD DZUTW BFMO
MESSAGE No. 7
UFUCL HJYDY ZTHHA NWJWJ IFMZI VZIJE QODUT
DKXXR TOSVL MNHNO RZRTX QIPFN HOUNL FGUVO
RTDCO LNTTM MXMXR TUIIG ZOJCU BTXJK KGUEJ
IKGSL APA
MESSAGE No. 8
NWWVY UMHVB RPQHF XOHQN IPATI CFMZT DIMQI
VTEGJ IAFEO BMUUT VPSKO HUYNA VPRXS SCUZB
GXHRP OIHWF LBTKF QESIO YXVNK AWDAA EQLVJ
KLXHR NCVNY XSQMC XVXJC TJVSC UJEGZ TFONC
MESSAGE No. 9
NYMPE YZYRZ AWLPP IMPBB VWAXZ QSVFG ITZMT
NNCMP SOJKI GDPQZ IIPMV ZOCBO UCKXR SEPEM
NDEUH YUEIP RFOHI QLQHG IJFRT UTNQC JEGAF
YKXXQ VRQUW
MESSAGE No. 10
TEPDA XXHHC FYMFK QRBJV YVJID JBNXF JBLXU
UFRXL WELQJ QJKFW RLSSF BQJWR BZKYN EAUWP
CQRVW ZVXXH NZHCW SVEYH NEANW G
MESSAGE No. 11
SEYBZ MGSOZ CMPSQ BASFH VFSCG CHKSB ZOPRZ
LGRXO XXCKK MVQGJ XYUOC FFFVJ OZCJE AQSJY
KEXYA VWONX GTCWT TGLOI IWORT JJVQE HYNK
MESSAGE No. 12
OMTXX WUXZE YEHOJ ALJCO EPLPJ RBCVX WARW
XEUJG ROHUP OFSGQ PONLW RAIBP KACIB GMYSB
WKZTE KZYIZ ZXFJO HZIVG ESGOX YYCEO B
MESSAGE No. 13
QNKIT FZHNC GJBSA JIQBX PFTAO RJLUD IKPKI
LJUFZ UBSMD VNMNZ UWVYQ DJPFY KETMN CLEQS
JBCPG RNNNH BSYZR STGBV E
JAKUK BEQMG
OQUVA WHSXE
HYMRQ LPUVN
MZPRX SWNFC
TPOWI VHNCQ
MSJBS ESVWJ
IRUJI NSBWR
JHCDE WUHJV
MYYTH GSEYA
HNPM
KZNFC DHSUA
OTZNM AMHWX
SQRWO DAJMT
KSOXR KNHJK
AYSKO UWJDX
CZEVH OCADC
WZCMO TOXVM
DYBDQ FTZDA
VKHEU XSPFG
TNUWU EVGHU
DQXNA GEYHC
6
ZJRKK CLYZK RINOK
YVWFL DXZSK ZPWNI
KKGEF FJWWR ZXROX
ZMKMI D
OSULC WZAFW, SQAUC
QQDCX CBIMT DGIAS
UUHTF FOQOK KNRPI
XDCAS BEHQV OPWEY
VHXVD RSRYI PVNWQ
RRQDM STTAL GKPPO
DIPSG ZUFVH OZIGB
UMMOL HVVEQ GWTJI
RAABX ZCVWR UUJEG
CJEDW AKHZA SDZIE
MESSAGE No. 14
NBTAK NOBBI WRZRX DQTAR AEKMY MOXLT
JUULI TPUJR TPQGH RJZCQ XCNHU NOKMI
PZUGG HMYNB DUHQZ BJDFA FJAYU RKHKB
MESSAGE No. 15
WPRBI WCFUO JKQHY
BAQRK RUHHF JYUJG
PLLXQ DRDEM JMCFB
UTISE NZFCU BXI
MESSAGE No. 16
QEDMJ LTKFN RMDGT
PZTCU BNCEE HAWMB
TQXND HMHEW VHTLT
MESSAGE No. 17
PQTNS AEFZH TOFQH
NMJHQ CHZUM TTSIU
HIHWG LBUFZ DVPUB
OIWVX KSMSX MBXOZ
DYNMA CWULT PKWEE
WLHVX FIFZR STVJV
DNMBM JDGVP KJMND
KRGNU ZXCWQ VEZTC
OOXNQ NPDDT QVSJR
GHELW HUMFH LZHNH
DCMXO NAYEY GQHID
MESSAGE No. 18
LALNH QUDUA ZBZUD VSJFE MEHXW EUWZT OKWNO OOSIL TASEG
OVQVX PMKKW BQRBI VGDGJ JDAHW RZDJA WQAXB FBRLA AHJEP
PUMEU HJQJP ZSPPQ VZWDL HECDA LPJJS ZOJYB MO
MESSAGE No. 19
YTVUL TWEVD MBHKV IHTPI GNXBQ XAUAQ OUFVO GSMKB BAKIG
YRNAF BKIJC ZJSRN WBQHM UYJPT GHCCB RLNVH OLDQA ZZDCV
UWMNZ OPRFG RDONY RCZAM ZYNVQ WFONZ CITES IRWER GKETA
YUSUK TFECD BMQVB VBWVV VWCZP TWCTJ FHFEH VNDCO MZVLK
YUJPZ BHLDY VVPMD KFHPB VCYU
MESSAGE No. 20
OBHOK UWRON AJDFH FRQMI ULOTG XXIEV HMAKV PVMAV OITKD
LDQIW UYVWI JEJRQ MCUZP KGUDN QSPBF TQVZP IZJTU RBUFZ
JUSIZ DCIGX QESJD LIZVM AOPME YNEXI HOXJK KUYHK AORUY
LVD
MESSAGE No. 21
OMXOW LTWEW KFJHN ZMSKK GXFDL, YLGPT YYYNQ CYYMA ZWRAD
HGWBI HHCGM FDGMN XIQDN NPLYQ NPJNZ OWFVV KMGVH KHCUC
STNVZ CGXHV LZZXX SVTKG POEAC OJYQU MEULH KHYDD ETPDN
QYZVU HIMGG RBHEW CUSO
MESSAGE No.
SFDFM
YJOUN
CXEUF
JJMIV
AQMQW
AEGQG
UKCTB
JFCEU
FFONY
HTZPB
BETEX
QFPMO
GVFZH
YRNEI
ZSLRK
TFSOQ
JTPSY
BWIVZ
QFXUV
XKESC
MESSAGE No.
LHKGP
XUWVE
LCSMI
DZAUO
KUKNY
YZAJE
YVSWW
W
WJROY
NFDEX
DQNAP
XZHPK
GDZGA
KFPAO
JJIHU
JBZET
FIQTR
MESSAGE No.
OMOOK
YJODO
WFKKZ
OKWHR
XIRIX
PXXZL
MPXQI
KDCBX
XFRWL
PIQMN
UEQTU
FHTEG
ALWNK
VPIWA
INNJI
MESSAGE No.
FPETJ
YXDKR
GWILM
FJYAZ
CDVNY
GWMQL
FGCVS
YQNVO
LMKQU
FQMEJ
NCZGB
WSMSS
CDALX
KOBMS
HUIRT
CCMFY
FIYGR
ZURAN
XUWAI
KVKRA
MESSAGE No.
XFYKU
VDCYO
ZAECE
TRJTG
VWRIH
ESOWP
DLZCS
HQWVZ
YQXGD
DJBCW
VEASF
XRLCH
NDBFZ
GYZLI
XQDYB
PZJCM
MPMHG
MJQEC
EMVZL
XZUAB
SECCE
IHSAM
TZCWL
IOKZO
KNDMF
TRURC
FDBPB
LSFXW
DURHG
GGJBF
GAMFQ
GLHMH
CIJTI
RHCMJ
AUHKA
EPLQU
GBVIJ
GBIWL
SEBMA
ANHJO
IJKIR
SGMFC
ULGXC
ZFTIQ
BUILZ
RAQBJ
MSTZZ
XJZKI
TKNDA
VEKYG
SRUEU
UZHRQ
UFRNA
TQRGT
WIZWK
OTJEP
ZUAPV
IWORY
DDQYB
JNSSI
22
IAONR
JNPXG
CLRIL
CJCFI
23
LUVBU
XYHAL
KQQIE
24
HZIKQ
NFWOH
ETVUP
25
HQMIP
ULZFE
DGAML
Y
26
LGPKS
YOOHV
RUWKE
ULXFU
APKQP
JODEE
IIZYY
VZXCL
QRYXB
OISIE
RCMUS
HCHXX
TWSZJ
SXXSZ
LHIGG
BN
ZRUDA
DGECA
HHYCA
XQNUY
GWXXY
QHTJY
FAONR
YDLOT
QBTFK
CWPSK
IAUNO
RIYIT
XSINC
KAXUF
XHRXE
HTAVV
YFYOE
VHMWL
RESKE
UXWQX
ZMZKW
OJUGA
HVRUC
FVDSQ
XXHRF
CDPYM
GHEBF
GQZNG
EKBMG
OP
QCAVJ
UZMJM
VYIGT
BSNWH
LPSJJ
BDWVG
MVKNT
CACVC
KIOZG
OTTTO
LSIJU
LPXGC
PAUJD
WDPDB
PVUGD
HOGXW
QEHQR
RJCSH
MOAFT
BTMYK
PFTYF
DFTFO
MGETD
MFIAC
SETNP
FLUVK
LTTXN
UTNIE
MZNGB
ZWXXQ
LNICD
DLRZV
RFTHL
SYBMC
RLETI
SMQVB
BYGMQ
QCVQ
r
I
It may be of advantage to begin the elucidation of {he principles of solution by translating
this cipher into terms of the sliding of primary alphabets against one another with the con-
sequent production of a multiplicity of secondary alphabets. For example, by using ordinary
sliding alphabets such as are commonly used in cryptanalysis, we may produce the same
results as are given by the set of concentric disks. Let us use the alphabets of the illus-
trative disks, mounted upon sliding strips in pairs, and let us slide each pair of alphabets 8
letters apart. Thus, if we consider the upper one of each pair of alphabets in figure 2 as the
plain-text alphabet and begin each alphabet arbitrarily with the letter A, we have the following:
[Plain text
[Cipher
F IGURE 2
A U F Q Z E R H Y G W J O I D M N C T X S B L P K V A U F Q Z E R H Y G W J O I D
A U F Q Z E R H Y G W J O I D M N C T X S B L P K V
fPlain text
Cipher
A O X G F Z M Y L P U K E T D J V S W I R B N H C Q A O X G F Z M Y L P U K E T D
A O X G F Z M Y L P U K E T D J V S W I R B N H C Q
[Plain text
Cipher
A W J B Q I H V K P U F O G T N E D S Z X C M L R Y A W J B Q I H V K P U F O G T
A f f J B Q I H V K P U F O G T N E D S Z X C M L R Y
oo
[Plain text
Cipher
A C V X Z f f M T F N U I O S E H J R D L G K Q B P Y A C V X Z W M T F N U I O S E
A C V X Z W M T F N U I O S E H J R D L G K Q B P Y
fPlain text
5
|Ciphe ler
A E T J U D M V Z N W H X O G Y K F R L Q B P C I S A E T J U D M V Z N W H X O G
A E T J U D M V Z N f f H X O G Y K F R L Q B P C I S
Note now that the first set of 5 plain-text letters, PREPA, yields the same set of 5 cipher
letters, MEKJR, that we found on page 2 by using the disks. The only thing which these five
pairs of independent sliding alphabets have in common in figure 2 is the fact that each pair has
been slid apart the same number of letters, viz, 8; if we consider the upper alphabet in each pair
as the stationary alphabet, then the lower one has been shifted 8 intervals to the right, or 18
intervals to the left, of the upper alphabet. This corresponds to the position of number 1 in
figure 1, for the latter number occupies the eighteenth segment to the right of the plain segment,
or the eighth to the left. The relative positions of the numbers in the numerical key, therefore,
correspond to the numbers of intervals the primary alphabets in the form of sliding strips would
have to be displaced in order to produce the same results as the disks.
Now the sliding against itself of a primary sequence containing 26 letters will give rise to
a series of 25 secondary cipher alphabets;' likewise, each primary concentric sequence will
give rise to a series of 25 secondary alphabets. If the numerical key consists of 25 numbers,
all these secondary alphabets will be employed; if it consists of less than 25 numbers, then a
correspondingly decreased number of secondaries will be employed.
Since each primary sequence can give rise to a set of 25 secondaries, the total number of
possible secondary alphabets in the whole system is 126; but if the numerical key consists of
less than 25 numbers, then the total number of secondaries will be less than 125 by exact multi-
ples of 5, since the absence of one or more numbers from the key affects all five primary concen-
tric sequences. For example, if the key consists of 21 numbers, then there will be involved
21X5, or 105 secondary alphabets. In a message of exactly 105 letters, then, each letter will
be enciphered by a different secondary alphabet. If the message contains more than 105 letters,
then all the letters after the 105th will be enciphered by the same secondary alphabets as at
the beginning of the message and in the same sequence.
In the explanation of the method of encipherment it was made clear that the substitution
proceeds in a regular manner, taking successive groups of 5 letters; the cipher equivalents are
taken from the successive segments, proceeding in a clockwise direction from any given initial
segment. It follows, therefore, that in a single long message wherein the complete encipherment
requires the passing through of this sequence of segments more than one time, there exist periodic
or cyclic phenomena of a type found in various ciphers, due to the presence of a definite or regu-
lar cycle. In this case, the length of this cycle in terms of groups of 5 letters corresponds exactly
with the length of the numerical key; its length in terms of individual letters is five times the
length of the key. For the sake of clarity, we shall refer to this cycle when stated in terms of
letters as the period. Thus, with a key of 21 numbers, the length of the cycle is 21 groups, and
the length of the period is 105 letters. If a message consists of 315 letters, for example, the
letters would pass through three complete cycles; the 1st, 106th, and 211th letters would be
enciphered in exactly similar positions, and therefore by exactly the same secondary alphabet.
The 2d, 107th, and 212th letters would likewise be enciphered by the same secondary alphabet,
but of course not the same as the preceding secondary alphabet. W ith a key of 23 numbers,
the length of the cycle is 23 groups, the length of the period, 115 letters; the 1st, 116th, and
231st letters would be enciphered by the same secondary alphabet; the 2d, 117th and 232d
letters by a different secondary, and so on. If we represent the length of the period by n,
then the 1st, (n + l)th, (2n-f-l)th, (3n-f l)th, . . . letters fall in the same secondary alphabet;
the 2d, (n-f-2)th, (2n+2)th, (3n+2)th, . . . letters fall in another secondary alphabet; and
so on. If a message be longer than the period, therefore, it will follow that the 1st, 2d,
3d, . . . nth secondary alphabets must contain repetitions of cipher letters, representing
1
The twenty-sixth secondary alphabet coincides with the normal alphabet, since each plain-text letter
would be represented by itself in that secondary alphabet.
10
repetitions of plain-text letters, for these secondary alphabets, are after all only single mixed
cipher alphabets, and the repetition of high-frequency letters in ordinary plain text is a necessary
characteristic of all alphabetical languages. Such repetitions will be evidenced by repetitions
in the cipher text at n, 2n, 3n intervals, and they may be used to determine the length of the
period. Exactly how this is done will presently be demonstrated.
But the determination of the length of the period is only a slight step forward in the analysis.
It is true that it will give us the length of the numerical key, but that is all. W hat we must know
next is the sequence of numbers, or rather, the relative positions of the numbers in this key.
W e may ascertain this by further scrutiny of the theoretical and actual results of the method
of encipherment. It is often the cae with various ciphers that the method of encipherment is
excellent in principle, and will yield practically indecipherable messages when the messages arc
very few in number, but the weaknesses in the method are quickly disclosed when it is used for
regular traffic such as that necessary in military cryptography, where many messages arc to be
sent each day in the same key. In the cipher under examination, the weakness is introduced by
the fact that the initial segment for each message of the day's activity is determined by the serial
number of the message.
1
Now there are as many initial segments for each numerical key as there
are numbers in that key. Once the starting point is determined, all the messages pass through the
same cycle; different messages merely begin at different points in the cycle. Now, since the
numbers applying to these starting points constitute the sequence of numbers in the key, the
successive initial segments constitute a series or sequence which, when properly reconstructed;
will give us the sequence of numbers in the key.
After the numerical key has been reconstructed, we are yet a long way from solution, for w^
are still confronted by the more complex problem of reconstructing, or solving, the cipher
alphabets.
W e have so far analyzed the solution of the problem into the following three steps or phases;
1. The determination of the length of the period.
2. The reconstruction of the numerical key.
3. The reconstruction of the cipher alphabets.
Let us proceed, therefore, to perform each step.
1. The determination of the length of the period.It was explained above how the cipher
system will result in the production of repetitions in the cipher text at definite intervals dependent
upon the length of the period. The first, (n.-f-l)th, (2n-f-l)th, . . .letters fall in the same second-
ary alphabet; the second, (n + 2)th, (2n + 2)th, . . . letters fall in another secondary alphabet;
and so on. If there are repetitions in the plain text at n intervals apart, there will be correspond-
ing repetitions in the cipher text. There would be involved here only a slightly modified case
of the ordinary process of factoring the intervals between repetitions in the cipher text, as applied
in the solution of typical periodic multiple-alphabet ciphers. Thus, in this case, if it happens that
the first, second, and third letters of a message, and also the (n+l )t h, (n + 2)th, and (n + 3)th,
are the letters THE, then there must be a repetition of the initial trigraph of the cipher t ext ,
representing THE, at a distance of n letters. But in a cipher involving BO many alphabets as this
one, the repetition of trigraphs and polygraphs would naturally be rather infrequent, except in u
very long message.
However, the paucity of trigraphs and polygraphs, and even of digraphs, need not prove
to be a great obstacle, for the repetitions of individual letters may be used wi t h groat accuracy
for the same purpose, viz, the determination of the length of the period. The method is based
1
However, were the initial segments determined in some other manner, the final results would be t hem) me, and
the cipher could be solved by a slight modification of method. Even if the i ni t i al segments were subj ect to no
law, the cipher could still be solved by the method hereinafter set forth, wi t h some modifications.
XFYKU NDBFZ
FDBPB SEBMA
UFRNA APKQP
HTAW OTTTO
RLETI DJBCW
IQKZO EPLQU
u a o i t
KNDMF GBVIJ TKNDA
SRUEU RUWKE RIYIT
KAXUF CACVC LNICD
RFTHL HQWVZ XZUAB
IHSAM RHCMJ RAQBJ
XJZKI JNSSI HCHXX
F IOD HE 3.Met sage 26 transcribed upon the attumption of a period of 100 Utters
a O J 5 4 0 4 4 O M < ) 0 7 0 7 5 8 0 8 5 g O S 1 0 0
LGPKS CWPSK BSNHH LTTXN VDCYO GYZLI TRURC GBIHL VEKYG YOOHV IAUNO LPSJJ UTNIE ZAECE XQDYB
BDffVG MZNGB TRJTG PZJCM LSFXW ANHJO UZHRQ ULXFU XSINC MVKNT ZWXXQ VWRIH MPMHG DURHG IJKIR
ESOWP MJQEC GGJBF SGUFC TQRGT JODEE XHRXE KIOZG DLRZV DLZCS EMVZL GAMFQ ULGXC WIZWK IIZYY
GLHMH ZFTIQ OTJEP VZXCL YFYOE LSIJU SYBMC YQXGD SECCE CIJTI BUILZ ZUAPV QRYXB VHMWL LPXGC
IWORY OISIE RESKE PAUJD SMQVB VEASF TZCWL AUHKA MSTZZ DDQYB RCMUS UXffQX WDPDB BYGMQ XRLCH
ZMZKW PVUGD QCVQ
Number of coincidences WWtfU WMl I fU [ H I I I I /
J
11
upon the construction of what we have called a "Table of coincidence", which will show us
mathematically the most probable length of the period. W e may as well use the text of our
series of messages to illustrate the process.
If we assume the numerical key to consist of 20 numbers, then the length of the period would
be 100 letters. Let us write the longest message of our seriesviz, message 26in exactly super-
imposed lines containing 100 letters each, and then make a count of the recurrences, or more
accurately, the coincidences,' of letters within the individual columns thus formed.
Note the repetition of the letter F in the second column. This fact is indicated by placing
a check mark in the tabulation of coincidences. W here a letter appears three times within the
same column (B, in column 8), three check marks are recorded, for there we have a coincidence
between the first and second, second and third, and first and third occurrences. W here a letter
appears four times in the same column, six check marks are recorded. The number of coinci-
dences for each case corresponds to the number of combinations of two things that can be made
from a total of n things.
2
We note that on the assumption of a period of 100 letters there is a total of 39 coincidences.
Now, if the period is really 100 letters in length, then the repetitions of letters within columns
are not mere coincidences brought about by chance superimposition of identical letters but are
actual recurrences in the restricted sense of being the resultants of the encipherment of similar
plain-text letters by the same secondary alphabet. But there is no way of determining from
this single tabulation whether the assumption of a period of 100 letters is correct or not, and
therefore we do not know whether the repetitions in this case are recurrences or coincidences.
This we can determine, however, by a comparison of tabulations of coincidences made upon
various assumptions of length of period. Theoretically, the correct assumption should yield a
higher total of coincidences than the incorrect assumptions, because the recurrence of high-
frequency plain-text letters (in English, E, T, 0, A, N, I, R, S, H, D) is to be expected, and the
number of such causally produced repetitions should certainly be greater than the number
of repetitions produced by mere chance in the superimposition.*
Let us proceed, therefore, to make a table of coincidence for the various probable lengths of
period, first transcribing message 26 into lines corresponding to hypotheses of 105, 110, 115, 120,
and 125 letters.
4
Before doing so, however, we find it necessary to introduce a few remarks
upon the desirability of using a slight correction factor for this table.
1
W e draw the distinction between recurrences and coincidences on the grounds that the former term should,
and will here be used to indicate repetitions of letters in the cipher text causally related to each other by being
enciphenncnts of identical plain-text letters by identical alphabets; whereas the latter term indicates repetitions
not causally related to each other in this manner but simply the result of chance. A repetition may therefore be
either a recurrence or a coincidence. The process of factoring in ordinary multiple-alphabet ciphers of the periodic
type has for its purpose the separation and classification of repetitions into the two kinds. Until proved otherwise,
all repetitions must be considered coincidences.
1
The formula is: C=n(n1 )/ 2. Thus, when n = 5, the number of coincidences is 10; when n = 6, the
number is 15.
3
Since this paper was written, a further study of the concept of coincidences has made it possible to predict,
wi t h a fair degree of accuracy, j ust how many coincidences should be expected for correct and incorrect assump-
tions. The mathematical and statistical analyses are given in detail in W . F. Friedman, Analysis of a Mechanico-
Rlectrical Cryptograph, Section V I; S. Kullback, Statistical Methods in Cryptanalysis, Section V II (Technical
Publications, S. I. S., 1934). It, results from these mathematical studies that the ratio of the number of actual
coincidences to the total number of possible coincidences is .038 for an incorrect case and .066 for a correct one.
T his knowledge eliminates the necessity for tabulations corresponding to every possible case and gives a reliable
means of det ermi ni ng the correct assumption as soon as it is made. (See Notes 1 and 3 on pages 13 and 14
respectively.)
' The message need not be written out more than once if long strips of cross-section paper are used, writing
a line on each strip. Each line should contain 125 letters, and the various strips can then be arranged to bring
the proper letters into supcrimposition according to each hypothesis in turn.
12
For really accurate comparison, the totals of coincidence obtained for the various hypotheses
should be corrected in order to make proper allowance for the differences in totals due solely
to t he var i at i on in t he number of letters in each column when transcribed according to each
hypot hesi s.
1
From a cryptographic point of view, a total of 100 coincidences in an arrangement
where there are G letters in each column represents a slightly greater degree of coincidence than
in an arrangement of the same message also yielding 100 coincidences, where there are 7 letters
in most of the columns; there is less opportunity for coincidences to be produced in the former
case. W e should, therefore, reduce all the totals of coincidence to some common basis. The
reasoning we have followed in the establishment of a correction factor to be applied is as follows:
Mt^s.i^t. 20 contains exactly 539 letters. W hen transcribed into lines of 100, 105, . . . , 125
letters, t he columns in each of these five set-ups have the following number of letters:
T ABLE I
39 col umns of 0 l oi t ers and 61 columns of 5 letters.
14 columns of G letters and 91 columns of 5 letters.
90 col umns of f) l oi t er s und 11 columns of 4 letters.
70 col umns i;f .
r
> K' t t or s und 36 columns of 4 letters.
59 crhmins of 5 leHors and 61 columns of 4 letters.
39 col umns of 5 letters and 86 columns of 4 letters-
Period
(letters)
100
105
110
115
120
126
Assuming that perfect coincidence can occur in each column (all letters identical), then in
a column of G letters we can have 6X 5/2=15 coincidences; in a column of 5 letters, 5X 4/2=10
coincidences; and in a column of 4 letters, 4X 3/2 = 6 coincidences.
If now we find the total number of chances for coincidences for each of the arrangements
given in table I, we have the following:
TABLE II
Period
(letters)
100
105
110
1 15
120
125
Conditions
39 columns of 15 chances each, 61 columns of 10 chances each..
14 columns of 15 chances each, 91 columns of 10 chances each.. _ .
99 columns of 10 chances each, 11 columns of 6 chances each
79 columns of 10 chances each, 36 columns of 6 chances each
59 columns of 10 chances euch, 61 columns of 6 chances each
39 columns of 10 chances each, 86 columns of 6 chances each
Total
chances
1, 195
1, 120
1 056
1, 006
956
906
Choosing for our basis of comparison the hypothesis of a period of 100 letters, the various
proportions of chances for coincidences for each of the remaining hypotheses will constitute
correction factors to be applied in each case. They are as follows:
T ABLE III
Period
(letters)
100
105
110
116
120
125
Chances for
coincidence
1, 195
1, 120
1,056
1,006
956
906
Correction
factor
1.00
1.07
1. 13
1. 19
1.25
1.32
1
See footnote 3 on the preceding page. This correction factor is unnecessary if the number of actual coin-
cidences is reduced to a percentage basis, in terms of the total possible number of coincidences.
13
W e are now ready to establish the tables of coincidence for the various hypotheses. Space
forbids the actual demonstration of the several arrangements of message 26 to correspond
to the various hypothetical key lengthsthat shown in figure 3 is typical of them all. W e
shall give only the final result in table IV .
TABUB IV
Period
(Tetters)
100
105
110
115
120
125
Coincidence* on each hypothesis
t H J M M I M I n u r t u n u m i
n u wwt H i wwwww
wt H i wm wn u m wwi i
WWlH t WM WWM Wt t Um W
WWWWWWWI
WWWWWWI I I
Total
39
45
47
60
36
33
Correction
factor
1.00
1.07
1. 13
1. 19
1.25
1.32
Corrected
total
39.0
48.2
53. 1
71.4
45.0
43.6
There seems to be no doubt but that the period of 115 letters is correct. The cycle, there-
fore, consists of 115-f-5=23 groups, and the numerical key contains 23 numbers. This means
that the two final segments bear no numbers, and are therefore blank segments.
2. The reconstruction of the numerical key.Having ascertained the length of the period,
and thus the length of the numerical key, the next step is to reconstruct the sequence of
numbers constituting the key. As stated before, this process is made possible in this case
by the method of encipherment which is such that all the messages of the day's activity go
through exactly the same cycle, but the successive messages begin at different initial points
in this cycle, and these points coincide with the relative positions of the numbers making up
the sequence of numbers in the key.
W e do not know the absolute position of any numbers in the numerical key, hut we may
proceed first to find their relative positions, regarding the key in the nature of a continuous
cycle or chain. Later we may find the absolute positions of the numbers in this cycle, i.e., we
shall have reconstructed the numerical key itself.
Now, as stated before, all messages proceed through the same cycle; it is only the initial
points for the messages wlu'ch are different. Hence, if we can determine the relative positions
in which messages 1, 2, and 3 should be superimposed in order to make all three messages
coincide as regards the portion of the cycle through which they pass simultaneously, we shall
thus have determined the relative positions of the numbers 1, 2, and 3 in the cycle. For
example, if we should find that the first group of message 2 belongs under the twelfth group
of message 1, and the first group of message 3 under the sixth group of message 2, we would
conclude that the relative positions of these numbers in the cycle are these:
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 3 4 5
1 2 . . . . 3
1
As a verification of Note 3, page 11 above, the percentages of the actual number of coincidences to the
total possible number has been calculated:
Period Percentage
100 0.033
106 . 040
110 044
115 .060
120.
126.
. 038
.036
14
How are these relative positions for messages 1, 2, and 3 to be determined? Clearly, we
may use the same basis for this determination as for that concerned with the length of the
periodviz, a table of coincidences. For, when messages 1 and 2 are correctly superimposed
we should get a higher degree of coincidence between the letters of the superimposed columns
than when they are not correctly superimposed. The reasons are the same as in the preceding
case: The successive groups of cipher letters in one message represent encipherments by the
same secondary alphabets as apply in the other message; hence, repetitions of plain-text letters
within columns will result in repetitions of cipher letters within those columns. W e can there-
fore determine the correct superimposition by experiment and the recording of coincidences.
Having found that the key contains 23 numbers, it is obvious that message 24 has the
same starting point in the cycle as message 1; we may proceed at once to combine them by
direct superimposition. W e do likewise with messages 2 and 25, and 3 and 26. The purpose
of this stej ) is merely to afford greater accuracy through the increased number of letters with
which we shall have to deal in finding the correct relative superimposition of these three sets
of messages.
Since message 26 contains the greatest amount of text, we may regard it' as our base and
try to find the relative position of messages 1 and 24 with respect to it. W e now place messages
1 and 24 beneath message 26, beginning the first group of the former beneath the second group
of the latter
1
. A tabulation of the number of coincidences in each column is then made. Mes-
sages 1 and 24 are again placed beneath message 26, beginning the first group of the former
beneath the third group of the latter and again the total number of coincidences is ascertained.
In other words, messages 1 and 24 are moved successively 1, 2, 3 . . . 22 intervals
2
to the right
of message 26, and a table of coincidences is constructed. The greatest total number of coin-
cidences, as shown in table V , is given when messages 1 and 24 are placed three intervals to the
right of message 26. This means, then, that the numbers 1 and 3 occupy these relative positions:
1 2 3
1
T ABLE V
3
Intervals 1 2 ~ ~ 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Coincidences 59 49 86 40 53 39 55 60 59 47 55 45 61 64 53 54 46 58 50 48 52 40
Showing various totals of coincidence when messages 26 and 1 and 24 are superimposed at different
intervals corresponding to the successive hypotheses of relative position.
Messages 2 and 25 are next taken for experiment and a similar table of coincidences is made
for the various superimpositions with message 26, omitting, of course, the three-interval trial
since the position corresponding to that test we found to be occupied by the number 1. As soon
as the relative position of eacli number is found, the subsequent trials corresponding to those
numbers may be omitted. The labor is, of course, somewhat tedious but may be done by
1
It was t hought unnecessary to use both messages 3 and 26 as a base, since the latter alone seemed suf f i ci ent l y
long to give reliable i nformat i on.
2
An interval in this case is equal to a group of 5 letters, because the enciphernicnt proceeds in segments of
5 letters each. W e need try only the first 22 intervals since the cycle consists of only 23 positions, und messages 1
and 24 cannot have t he same beginning point as messages 3 and 20.
3
W i t h regard to the mathematical not i on discussed in Note 3 on p. 11, t he fol l owi ng remark is per t i nent :
The total number of coincidences possible when messages 1 and 24 are placed three i nt erval s to the ri ght of
message 20 is 1253. The number of actual coincidences, 8(5, when divided by 1253 gives .068. Since the expected
result for a correct assumption is .060, it is at once evident that this assumption is correct and it is consequently
unnecessary to consider any further cases.
1
FIOCRB 'nf cor r ect t u per i m por i t i on of
MESSAGES 1 AND 24
1
MLVXK
YPITOJ
24 |
MOOK
IETVUP
2 !"""
i QFXUE
25
FPETJ
DGAML
IYLMKW
1 YRAYU
26
XFYKU
RUWKE
MJQEC
YFYOE
AUHKA
QNXVD
DFTJV
OKWHR
QHTJY
GXAEU
HPZHK
CDVNY
QBTFK
CBGSF
NYUCS
NDBFZ
RIYIT
GGJBF
LSIJU
MSTZZ
GIRIE
ZHTWB
MPXQI
OP
DTEIL
MQZZT
LMKQU
VYIGT
VGABP
LNEMV
KNDMF
BDffVG
SGMFC
SYBMC
DDQYB
IUNEE
WXTMF
PIQMN
UZBRW
KMURW
CDALX
FLUVK
HOZFV
VNSXN
GBVIJ
MZNGB
TQRGT
YQXGD
RCMUS
FEXVP
OZDOJ
ALWNK
GJZSS
EJVB
FIYGR
FJYAZ
QQNSQ
WGVME
TKNDA
TRJTG
JODEE
SECCE
UXWQX
HPVZR
HZIKQ
QLUOX
HQMIP
YQNVO
NQLQL
MXPDF
LGPKS
PZJCM
XHRXE
CIJTI
ffDPDB
UKSEK
XQNUY
PTFOO
FAONR
WSMSS
DIGXM
WGTZE
CBPSK
LSFXW
KIOZG
BUILZ
BYGMQ
MVQCI
GQZNG
NWSHD
QCAVJ
CCMFY
XCWAI
KRLGU
BSNWH
ANHJO
DLRZV
ZUAPV
XRLCH
VXSFW
DFTFO
BPTJO
MFIAC
KVKRA
QFJOQ
ZJFZJ
LTTXN
UZHRQ
DLZCS
QRYXB
IOKZO
GVART
YJODO
HQRYY
YXDKR
Y
TYDEL
W
VDCYO
ULXFU
EMVZL
VHMWL
EPLQU
MESSAGES 2 AND 25
PTFOO NWSHD BPTJO HQRYY YAXRZ KTEMP UAYMK ISRDZ VUVKW HXAYD YAGSM CURBZ LBXOV EBBPI BMLCB UMAXF ZSLXV
GBMQL FQMEJ KOBMS ZURAN ULZFE YDLOT UZMJM SE7NP GWILM FGCVS NCZGB HUIRT XUWAI
MEB8AOE6 3 AND 26
MBMJB SEPSO DHREM ELKIP KXNMW QYBIH
GYZLI TRURC GBIWL VEKYG YOOHV IAUNO
XSINC MVKNT ZWXXQ VWRIH MPUHG DURHG
GAMFQ ULGXC WIZWK IIZYY HTAW OTTTO
LPXGC RLETI DJBCK IHSAM RHCMJ RAQBJ
XJZKI JNSSI HCHXX ZMZKW PVUGD QCVQ
BHFDC GLYWC YGMMP EEXZH UBBSB SBONG URQKH
LPSJJ UTNIE ZAECE XQDYB FDBPB SEBMA SRUEU
IJKIR UFRNA APKQP KAXUF CACVC LNICD ESOWP
RFTHL HQt VZ XZUAB GLHMH ZFTIQ OTJEP VZXCL
IWORY OISIE RESKE PAUJD SMQVB VEASF TZCWL
1
1
15
clerks. This process is continued in similar manner for all the remaining messages. The data for
messages 2 and 25 show that they belong five intervals to the right of messages 1 and 24, and the
relative positions of the numbers 1,2, and 3 in the cycle are therefore these:
2 3
1
1 2 3 4 5
2
The data for this determination of position for all messages are given in table VI.
TABLE V I.Data for determination of the position of the numbers in cycle
MESSAGE 20 USED As A BASK
{Position.... 2 3 4 5 6 7 8 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23
IN umber of coincidences 59 49 86 40 53 39 65 60 59 47 65 45 61 64 53 54 46 68 50 48 52 40
{Position 2 3 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
[Number of coincidences 45 49 58 67 62 61 99
.{Position 2 3 5 6 7 8 10 11 12 13 14 15 16 17 18 19 20 21 22 23
IN umber of coincidences 34 35 38 36 16 33 27 23 19 24 28 24 24 21 19 25 20 42 35 32
{Position 2 3 5 6 7 8 10 11 12 13 14 16 16 17 18 19 20 22 23
6
1 Number of coincidences 44 31 33 31 33 29 26 22 16 23 49
{Position. 2 3 5 6 7 8 10 11 12 13 16 16 17 18 19 20 22 23
IN umber of coincidences _ 25 33 37 25 40 31 20 11 66-
-{Position 2 3 5 6 7 8 10 11 13 15 16 17 18 19 20 22 23
7
[Number of coincidences.,._ . 23 33 26 27 25 20 24 31 31 16 19 28 28 28 46
[Position 2 3 5 6 7 8 10 11 13 15 16 17 18 19 22 23
IN umber of coincidences 36 27 53
{Position .... 2 3 6 7 , 8 10 11 13 15 16 17 18 19 22 23
IN umber of coincidences 27 32 26 17 21 21 30 17 13 21 27 58
{Position _ 2 . 3 6 7 8 10 11 13 16 16 17 19 22 23
{Number of coincidences... 18 18 15 13 23 19 14 18 13 20 10 15 18 52
{Position 2 3 6 7 8 10 11 13 15 16 17 19 22
J N umber of coincidences 16 28 10 15 14 16 16 26 17 28 20 31 19
u
j{Position 3 13 16 19
IN umber of coincidences 21 30 22 16
12
jPosition 2 .3 6 7 8 10 11 15 16 17 19 22
(Number of coincidences., 18 27 56
^{Position....-- 2 3 7 8 10 11 15 16 17 19 22
[Number of coincidences 20 1915 18 18 19 16 38
{Position 2 3 7 8 10 ll 15 17 19 22
[Number of coincidences '29 17 32 30 18 20 23 33 24 42
{Position
r
2 3 7 8 1 01 1 1 61 7 1 9
[Number of coincidences 38 32 37 27 24 '69
1
Secondary test, using messages 1 and. 24 plus 2 and 25 as the base.
16
T ABLE V I.Data for determination of the position of the numbers in cycleContinued
MESSAGE 26 USED As A BASE
(Position 2 3 7 8 10 15 17 19
[Number of coincidences 24 23 47
[Position 2 3 8 10 15 17 19
I Number of coincidences 15 44 30
18
[Position 2 8 10 15 17 19
[ N umber of coincidences 48 31 26
[Position 8 10 15 17 19
[Number of coincidences 28 37 66
[Position 8 10 17 19
\Numberofcoincidences 65 14
21
[Position - 10 17 19
J Number of coincidences 26 33 47
[Position 10 17
[Number of coincidences 48 22
23
[Position 17
(Number of coincidences 50
When in any trial the total of coincidences for a certain position stands out prominently
from the preceding ones, subsequent trials for the message concerned are omitted.
The final result of carrying out this work for all the messages tried against message 26 is
that the following cycle is established:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
3 18 17 1 8 12 16 20 2 22 15 6 11 5 19 13 23 9 21 7 4 14 10
This reconstructed cycle represents, as stated before, the relative, not the absolute, posi-
tions of the numbers in the key, because there is as yet no indication as to what number occupies
any given segment on the base disk. Furthermore, we must remember that there is a break
of two intervals somewhere within this cycle, representing the two blank segments on the base
disk. This fact may cause some difficulty later on but we shall find a way of overcoming it.
In the meantime, we may content ourselves with the cycle as established and proceed to an analy-
sis of the results of its reconstruction.
The immediate result is to enable us to superimpose all the messages of the day's activity,
as shown in figure 5. W e may begin with message 1, or with any other message, but for the
sake of convenience in analysis, we may as well transcribe them in regular order.
The letters of each of these 115 columns belong to a corresponding number of secondary
alphabets, all different, but all single, mixed, substitution alphabets. Individual frequency tables
are made, therefore, and are shown in table V II. The tables are given in groups of five, labeled
A, B, C, D, and E, corresponding to the five primary alphabets of the system. The groups of
alphabets are given in their proper cyclic sequence, so that each set of five alphabets is accom-
panied by a number which identifies its position in the sequence of segments. Thus, we may refer
to any of these secondary alphabets by number and letter, aa for example, 5B, meaning the second
single alphabet under segment 5. The A alphabets all apply to the outermost primary alphabet,
or alphabet 1; the B alphabets to primary alphabet 2, and so on. W e are now ready to attempt an
analysis of the cipher text with the object of solving these secondary alphabets and reconstructing
the primaries.
1
1. MLVXK
YPWDJ
LBXOV
HOZFV
VNSXN
WGILB
GWCEM
DCBUK
MZPRX
8.
YXVNK
NNCMP
YVJID
OZCJE
HZIVG
VNMNZ
WRZRX
D
CWULT
OZIGB
HWEQ
DCMXO
ZBZUD
HO
WBQHM
MZVLK
DCIGX
ZWRAD
QEHQR
YZAJE
24. OMOOK
ETVUP
GWILH
GBVIJ
UZNG8
TQRGT
YQXGD
RCHUS
8 12
10 11
QNXVD GIRIE
DFTJV ZHTWB
EBBPI BHLCB
QQNSQ NQLQL
WGVME HXPDF
JNXTD BPREX
JPNVB GRJGB
OQUVA WHSXE
SWNFC DKXXR
NWWVY UMHVB
AWDAA EQLVJ
SOJKI GDPQZ
JBNXF JBLXU
AQSJY WZCMO
12. OHTXX
ESGOX YYCEO
UWVYQ DJPFY
DQTAR AEKMY
PKWEE UUHTF
16.
TQXND HMHEW
GWTJI PQTNS
NAYEY GQHID
VSJFE MEHXW
UYJPT CHCCB
YUJPZ BHLDY
QESJD LIZVM
HGWBI HHCGM
CXEUF UKCTB
NFDEX GDZGA
OKWHR MPXQI
QHTJY OP
FGCVS NCZGB
TKNDA LGPKS
TRJTG PZJCM
JODEE XHRXE
SECCE CIJTI
UXWQX WDPDB
l e 20
30 29
IHNEE FEXVP
WXTIIF OZDOJ
2.
UMAXF ZSLXV
DIGXM XCWAI
WGTZE KRLGU
MMWHB OBFVO
IBWOH YMOAP
VGQSW OGHPI
TOSVL HNHNO
RPQHF XOHQN
MYYTH GSEYA
IIPHV ZOCBO
KSOXR KNHJK
TOXVM KEXYA
WUXZE YEHOJ
B
KETHN CLEQS
MOXLT YVWFL
FOQOK KNRPI
VHXVD RSRYI
VHTLT
AEFZH TOFQH
EUWZT OKWNO
RLNVH OLDQA
WPMD KFHPB
20. OBHOK
AOPME YNEXI
FDGMN XIQDN
BETEX ZSLRK
JBZET XYHAL
PIQMN ALWNK
25.
HUIRT XUWAI
CWPSK BSNWH
LSFXW ANHJO
KIOZG DLRZV
BUILZ ZUAPV
BYGMQ XRLCH
2 22
30 U
HPVZR UKSEK
ULJCY GXAEU
QFXUE HPZMK
QFJOQ TYDEL
ZJFZJ W
TFGSJ SLXEH
IZYNX GXMSB
XCYIO SUAEU
R2RTX QIPFN
IPATI CFJIZT
KLXHR NCVNY
UCKXR SEPEM
UFRXL WELQJ
VWONX GTCWT
ALJCO EPLPJ
DQXNA GEYHC
DXZSK ZPWNI
15.
PLLXQ DRDEM
PVNWQ QEDMJ
OOXNQ NPDDT
OOSIL TASEG
ZZDCV UWMNZ
VCYU
UWRON AJDFH
HOXJK KUYHK
NPLYQ NPJNZ
22. SFDFM
QFXUV CLRIL
DGECA CDPYM
HZIKQ XQNUY
FPETJ CDVNY
DGAML QBTFK
LTTXN VDCYO
UZHRQ ULXFU
DLZCS EMVZL
QRYXB VHMWL
IOKZO EPLQU
i s e
40 45
MVQCI VXSFW
DTEIL UZBRW
HQZZT KMURW
MBMJB SEPSO
RTZMI LLUUX
OZZGK FVURN
6. XGORF
BQAPY RMDMW
HOUNL FGUVO
DIMQI IRUJI
XSQMC XVXJC
OTZNM AMHWX
QJKFW RLSSF
11.
TGLOI IWORT
RBCVX WARW
JBCPG RNNNH
JUULI TPUJR
OSULC WZAFW
JMCFB WLHVX
LTKFN RMDGT
QVSJR RAABX
OVQVX PMKKW
OPRFC RDONY
FRQMI ULOTG
AORUY LVD
OWFW KMGVH
AQMQW FFONY
LHIGG FVDSQ
BTMYK LCSMI
GQZNG DFTFO
LMKQU CDALX
VYIGT FLUVK
GYZLI TRURC
XSINC MVKNT
GAMFQ ULGXC
LPXGC RLETI
XJZKI JNSSI
F
11 6
M U
GVART YBZKJ
GJZSS QLUOX
EJVB
DHREM ELKIP
FIFWC PGSBA
5. VJZCQ
NJFGQ DTPLV
GCHAX DUEIQ
FWGSQ HYMRQ
TPOWI VHNCQ
NSBWR VTEGJ
TJVSC UJEGZ
NDEUH YUEIP
BQJWR BZKYN
SEYBZ MGSOZ
JJVQE HYNK
DYBDQ FTZDA
BSYZR STGBV
TPQGH RJZCQ
SQAUC WPRBI
FIFZR STVJV
DNMBM JDGVP
ZCVWR UUJEG
BQRBI VGDGJ
19.
RCZAM ZYNVQ
XXIEV HMAKV
KHCUC STNVZ
GVFZH JTPSY
RJCSH JJMIV
YVSWW DQNAP
YJODO XIRIX
FIYGR HQMIP
FJYAZ YQNVO
GBIWL VEKYG
ZWXXQ VWRIH
WIZWK IIZYY
DJBCW IHSAM
HCHXX ZMZKW
IODRB 5
1* 13
0 W
WVUPV XZCBD
PTFOO NWSHD
KXNMW QYBIH
KRCAS XWKQV
KZJJK DCVQD
STOID DWVLR
XAOWK BKBUH
LPUVN LNGQT
RTDCO LNTTM
IAFEO BMUUT
TFONC HNPB
RFOHI QLQHG
EAUWP AYSKO
CMPSQ BASFH
XEUJG ROHUP
13. QNKIT
E
XCNHU NOKMI
WCFUO JKQHY
XDCAS BEHQV
KJMND RRQDM
NMJHQ CHZUM
JDAHW RZDIA
YTVUL TWEVD
WFONZ CTTES
PVMAV OITKD
CGXHV LZZXX
IAONR TWSZJ
JFCEU QFPMO
23.
KFPAO FIQTR
KDCBX UEQTU
FAONR QCAVJ
WSMSS CCMFY
YOOHV IAUNO
MPMHG DURHG
HTAVV OTTTO
RHCMJ RAQBJ
PVUGD QCVQ
23
n n
BDOLS GHIHZ
BPTJO HQRYY
BHFDC GLYWC
SLDKS NTESD
KSSXY TSUXE
TXTBH WNWIE
SKWWW XNEYZ
IPJRI HUMAD
HXMXR TUIIG
VPSKO HUYNA
9. NYMPE
IJFRT UTNQC
UWJDX CQRVW
VFSCG CHKSB
OFSGQ PONLW
FZHNC GJBSA
KKGEF FJWWR
OIWVX KSMSX
OPWEY UTISE
STTAL GKPPO
TTSRU GHELW
WQAXB FBRLA
MBHKV IHTPI
IRWER GKETA
LDQIW UYVWI
21.
SVTKG POEAC
OJUGA HOGXW
TFSOQ XKESC
LHKGP KUKNY
KQQIE HHYCA
VPIWA NFWOH
MFIAC YXDKR
KVKRA Y
LPSJJ UTNIE
IJKIR UFRNA
RFTHL HQWVZ
IWORY OISIE
21 7
M 85
LJCTE KSLPY
YAXRZ KTEMP
YGMMP EEXZH
4 .
QVQBN RZDMB
GRROH PXKZF
YJXXW BKOJQ
JAKUK BEQMG
DZUTW BFMO
7. UFUCL
ZOJCU BTXJK
VPRXS SCUZB
YZYRZ AWLPP
JEGAF SQRWO
ZVXXH NZHCW
ZOPRZ CZEVH
RAIBP KACIB
JIQBX PFTAO
ZXROX PZUGG
MBXOZ QQDCX
NZFCU BXI
PZTCU BNCEE
HUMFH LZHNH
AHJEP PUMEU
GNXBQ XAUAQ
YUSUK TFECD
JEJRQ MCUZP
OMXOW LTWEW
OJYQU MEULH
YJOUN AEGQG
CJCFI BN
WJROY XZHPK
GHEBF PFTYF
GffXXY EKBMG
GWMQL FQMEJ
ZAECE XQDYB
APKQP KAXUF
XZUAB GLHMH
RESKE PAUJD
4
n
VPBYD
UAYMK
UBBSB
UFHUJ
JQLYH
ZKFMS
FOUHO
IWTVU
HJYDY
KGUEJ
JHCDE
IMPBB
DAJMT
SVEYH
OCADC
GMYSB
RJLUD
14.
HMYNB
CBIMT
HAWMB
CJEDW
HJQJP
OUFVO
BMQVB
KGUDN
KFJHN
KHYDD
HTZPB
JJIHU
DZAUO
MGETD
KOBMS
FDBPB
CACVC
ZFTIQ
SMQVB
II !
14
M
WTRJK
ISRDZ
SBONG
LTMKJ
LTMXB
VGDWU
T
NPCLU
ZTHHA
MSJBS
WUHJV
VWAXZ
YKXXQ
10.
NEANW
LGRXO
VKHEU
IKPKI
ZJRKK
DUHQZ
DGIAS
KRGNU
AKHZA
ZSPPQ
GSMKB
VBWW
QSPBF
ZMSKK
ETPDN
YRNEI
LUVBU
W
WFKKZ
ZURAN
SEBMA
LNICD
OTJEP
VEASF
10
t oo
BDDFA
VUVKW
3.
URQKW
PONFG
BSVWL
XLTBW
KQDIG
NWJWJ
ESVWJ
GXHRP
QSVFG
VRQUW
TEPDA
G
XXCKK
XSPFG
TNUWU
CLYZK
BJDFA
BAQRK
ZXCWQ
SDZIE
VZWDL
BAKIG
VWCZP
TQVZP
GXFDL
QYZVU
BWIVZ
ZRUDA
PXXZL
ULZFE
26.
SRUEU
ESOWP
VZXCL
TZCWL
3
101
ANJXE
HXAYD
YLMKW
YRAYU
RIUGG
ILKYU
EXHEF
YLMNC
IFMZI
IKGSL
OIHWF
ITZMT
XXHHC
MVQGJ
WKZTE
EVGHU
RINOK
FJAYU
RUHHF
VEZTC
HIHWG
18.
HECDA
YRNAF
TWCTJ
IZJTU
YLGPT
HIMGG
JNPXG
XXHRF
XFRWL
YDLOT
XFYKU
RUWKE
MJQEC
YFYOE
AUHKA
18 17
no us
XGHED ERYVP
YAGSM CURBZ
CBGSF VGABP
NYUCS LNEMV
OZGWS UBNMW
NMFEB KB
AWQWF ZESMX
KYJJF SDSTU
VZIJE SQODUT
APA
LBTKF QESIO
KZNFC DHSUA
FYMFK QRBJV
XYUOC FFFVJ
KZYIZ ZXFJO
LJUFZ UBSMD
NBTAK NOBBI
RKHKB ZMKMI
JYUJG DYNMA
DIPSG ZUFVH
17. UMMOL
LBUFZ DVPUB
LALNH QUDUA
LPJJS ZOJYB
BKIJC ZJSRN
FHFEH VNDCO
RBUFZ JUSIZ
YYYNQ CYYMA
RBHEW CUSO
SXXSZ HVRUC
MOAFT XUWVE
FHTEG INNJI
UZMJM SETNP
NDBFZ KNDMF
RIYIT BDWVG
GGJBF SGMFC
LSIJU SYBMC
MSTZZ DDQYB
17
I
The first thought that comes to one is that these individual mixed alphabets may be solved
upon the basis of frequency alone, as is commonly done with such frequency distributions. For
example, we might assume the most frequently occurring letter in each alphabet to be the equiva-
lent of plain-text letter E, the next, of T, and so on; then substitute the Assumed values in the text
and try to build up words. But each of these alphabets contains only an average of 36 letters, so
that hardly any assumption would carry a considerable degree of certainty. This is especially
the case in English text where the letter E does not always stand out prominently as the most
frequently used letter in small amounts of text. Were an analysis of this kind absolutely necessary
to solution, it is doubtful whether this particular set of messages could be solved except after a
long period of patient labor. But it will be shown now that such an analysis is in fact not essen-
tial, because we may be able to effect a direct reconstruction of the five primary alphabets, which
will not only lead to the solution of all these messages, but will also give us every one of the possible
125 secondary alphabets of the entire system.
TABLE VII
1
A
B
C
D
E
F
O
H
I
J
K
L
M
N
O
P
Q
K
8
T
TJ
V
W
X
Y
Z
A
/
////
/
III
III
1
W
1
III
1
1
1
II
III
m
H
B
Il l l
III!
1
1
1
1
III
II
1
II
1
1
II
Il l l
1
Mi l
O
I
1
III
1
m
i
i n
i
i
i
i
H I
i
i
Ml
1
II
II
D
/
II
1
M
1
II
II
Il l l
II
1
II
1
II
Il l l
1
II
1
E
III
Il l l
II
1
II
M
III
1
1
II
1
1
1
II
II
III
1
8
A
B
C
D
E
F
O
H
I
J
K
L
M
N
O
P
Q
B
8
T
IT
V
W
X
Y
Z
A
II
1
II
II
1
1
1
Il l l
III
II
1
Il l l
III
III
III
1
1
1
B
/
II
II
II
III
1
III
II
II
1
M
1
II
1
M
II
1
C
/
//
III
1
1
1
M
M
II
III
1
II
m
III
1
D
II
1
1
1
Il l l
II
1
M
1
1
1
1
///
/
/
II
1
m
i
i
E
111
1
/
Il l l
M
II
1
Il l l
II
II
1
1
1
III
Il l l
1
Cfl
39
18
TABLE VIIContinued
12
A
B
C
D
E
P
O
H
I
J
K
L
M
N
0
P
Q
K
8
T
XT
V
W
X
T
Z
A
/
///
//
//
/
I tU
I I
1
I I
I I I
1
I I
I I
I I I
I I I
1
1
1
B
/
/
///
//
/
mi
in
i
ii
I lll
H I
mi
i
i
i
i
ii
c
nu
nu
in
i
nu
nu
in
i
ii
ii
in
D
///
///
///
/
////
//
/
//
/
///
///
///
/
I H l' l
E
/
mm
i
in
ii
ii
i
i
in
n
i
ii
i
n
n
in
i
16
A
B
O
D
B
F
O
H
I
J
X
L
M
KT
O
P
Q
B
8
T
TT
V
W
X
Y
Z
A
//
///
/
/
/
//
/
///
/
///
/
///
/
//
//
/
////
///
B
//
/
///
//
//
////
/
///
W
1
I I
I lll
1
1
1
I I
c
/
//
///
//
//
//
////
////
/
nu
in
nu
i
i
D
///
//
///
Mill
I I
1
I I
1
I lll
I lll
E
/
//
////
///
/
////
//
/
//
///
/
/
I H l
1
I I
1
1
19
VIIContinued
20
A
B
O
D
X
F
O
H
I
J
K
L
M
N
O
P
Q
K
8
T
TT
V
W
X
Y
Z
A
I I
1
1
1
1
1
M
1
Ml
1
Ml
I I I !
I lll
B
I I
1
I I I
1
1
1
1
I lll
1
M
I I I
I I
M
1
I I
1
1
O
1
1
I I
I I I
I I
Mllll
1
I I I
I I
I I
1
1
I I I
1
1
1
Ml
I lll
I I
D
I lll
1
1
1
1
1
I I
I I I
I I I
I lll
I lll
1
I I
1
I I
I I I
1
B
I I I
1
1
I I I
Ml
1
lltl
I I
I I
M
I I
1
1
I I I
9
A
B
C
D
E
F
O
H
X
J
K
L
M
N
O
P
Q
B
8
T
V
V
W
X
Y
Z
A
/
mi
i
in
in
i
i
i
H
H
mi
n
i
M
n
i
n
B
I I I
M
I I
1
M
I lll
I lll
1
1
1
1
I I
1
I
o
I I
1
I I
1
1
1
1
I I I
I I
n
i
i
in
i
i
i
Ml
I lll
I I
B
mi
i
n
i
i
i
mi
n
i
ii
mi
in
i
M
I I I
E
I I
1
1
1
I I I
I I
I I I
I I
I I I
Mil
I I I
1
I I
in
i
20
TABLE VIIContinued
22
A
B
O
D
E
F
0
H
I
J
X
L
M
K
0
P
a
B
8
T
TCT
V
W
X
T
Z
A
/
////
/
///
////
/
/
///
///
////
//
///
//
//
/
/
B
/
//
/
//
////
I I
1
1
1
1
I I I
1
Ml
1
1
1
I I
1
I I
1
o
I I
I I
mi
i
in
I I I !
1
I I I
1
I I
1
I I I
1
I I
I I
1
D
/
wtiii
mi
n
i
n
M
i
n
/
i
n
n
n
E
/
/
/
I I
1
I I I
I lll
I lll
I lll
1
1
I I I
I lll
I I I
I I
16
A
B
O
D
E
F
a
H
I
3
K
L
M
IT
0
P
Q
R
8
T
U
V
W
X
Y
Z
A
//
//
//
/
///
/
///
////
///
Ml
I I
I I
1
1
I I I
B
/
///
/
/
/
//
//
//
//
////
/
///
M
1
I I I
1
I I
1
o
/
///
/
/
///
///
/
I tti
I lll
I I
1
I I I
1
Mil
D
/
m
mi
i
n
i
in
in
mi
i
n
in
i
in
i
i
E
//
W
I I I
mm
n
n
i
/
i
i
n
i
i
1!
it
n
21
TABLE VIIContinued
6
A
B
0
D
B
F
O
H
I
J
X
L
M
N
O
P
Q
B
8
T
TT
V
W
X
T
Z
A
/
/
/
I H 1
I I
1
I I
I I I
1
1
Mil
1
I I
I I I
I I
I I
I I
B
/
/
I I
1
I I
I I
mn
nun
i
i
ii
Ml
i
i
ii
c
I lll
1
I lll
1
I I
I I
I I
1
M
1
till
1
Mill
1
D
/
111
1
I I I
1
1
I I
I lll
Mil
I lll
I I
1
I lll
1
1
E
I I I
I I
1
I I
I lll
1
1
I I I
1
1
I I I
Mil
M
I I
11
A
B
O
D
E
F
0
H
I
J
K
L
M
N
O
P
Q
B
8
T
U
V
W
X
Y
Z
A
I I I
I lll
1
M
M
1
1
1
I I I
I I
I I
I I I
1
1
I I
I I
B
/
I lll
1
1
I I
I lll
Mini
i
n
H I
n
in
n
i
i
c
I I
I I I
I I
1
I lll
1
I I
I I
1
1
n
i
n
i
mi
i
mi
in
D
I I I
I lll
1
I I
I I
I I I
1
1
I lll
I I I
Mill
I I
I I I
E
I lll
1
tin
n
i
n
in
i
mi
Ml
i
i
i
i
n
n
22
TABLE VIIContinued
6
A
B
O
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
B
8
T
TT
V
W
X
T
Z
A
/
///
/
/
////
//
///
/
/
/
/
///
//
Ml
1
1
11
I I
B
/
/
/
///
//
//
////
//
//
/
///
Mil
I I I
1
I I I
1
c
1
I I
mi
ii
i
in
in
M
I I
I I I
I I I
1
1
Ml
D
I I
I I I
I I I
1
1
I I I
Mil
1
I I I !
1
I I
1
1
I I I !
I ll
E
//
//
/
/
///
/
/
/
M
Ml
M
1
I I
I I
I I I
19
A
B
C
D
E
F
O
H
I
J
K
L
M
N
O
P
Q
B
8
T
T J
V
W
X
Y
Z
A
//
//
/
/
//
//
nut
i
i
i
in
in
i
i
mi
mi
n
B
M
I I
I I I
1
M
1
1
1
I I
1
I I
1
1
M
I I I
I I
c
//
M
1
I I I
I I
/I /I
I I
Mill
I I
M
1
1
D
////
/
/
//
/
Mil
1
I I
I I
M
1
1
I I
I I
I I
I I
E
/
///
//
/
/
//
/
/
M
1
I I
I I
I I I
I I
M
I I
i
1
23
TABLB VIIContinued
18
A
B
O
D
B
F
O
H
I
J
K
L
M
N
O
f
Q
B
8
T
TT
V
W
X
Y
Z
A
/
nn
n
in
i
i
i
i
in
n
n
mi
nn
i
n
i
n
B
I I I
I lll
I I
1
1
I I
I I
1
1
I lll
I I
1
I I
1
m
n
in
c
/
I I
1
1
1
1
I I
I I I
1
I I
mi
i
nn
mi
n
in
n
D
I I
1
' 1
I I
I lll
I I I
I I
1
I I I
1
m
nn
nn
n
i
i
E
/
m
n
in
i
in
in
nn
i
n
i
in
i
n
i
n
23
A
B
O
P
E
F
O
H
X
J
K
L
M
K
O
P
Q
B
S
T
tr
V
w
X
y
z
A
I I I
1
I lll
I I I
I I I
I lll
1
I lll
I I I
1
I I I
1
B
/
I I
m
I I
1
I I I
11
1
mi
I I
1
I
I I
I I I
1
I I
1
c
/
/
I I
1
I I
I I
I I
I I I
1
I I
I I
mn
m
i
nn
D
,/
/
/
I I
I I I
I I I
1
I lll
I I
I lll
1
1
1
I lll
1
I I
H i
E
///
/
lit
1
1
I I
1
1
1
I I
I I
1
I I I
I I
I I
1
1
1
I I
I I
I I I
24
TABLE VIIContinued
9
A
B
O
D
E
F
a
H
i
j
K
L
M
N
O
P
a
B
8
T
XT
V
W
X
Y
Z
A
I I
I I
Ml
W
1
I I
I I I
1
I I
I I
m
i
n
it
B
/
I I
m
i
n
in
i
n
in
in
n
I lll
I lll
1
I I
c
/
/
nut
i
in
n
H I
H I
i
mi
i
i
i
i
mi
H I
r>
//
/
////
/
///
////
/
I I I
1
mi
i
n
in
n
H
E
Ml
1
I lll
I I
Ml
1
1
I I
1
I I
I lll
1
I I
I I I
31
A
B
O
D
E
P
a
H
i
j
K
I.
M
N
O
P
Q
B
8
T
IT
V
W
X
Y
Z
A
//
/
/
nu
i
mi
i
i
i
n
i
i
n
i
i
i
Ml
M
B
////
/
I I I
1
I I
1
Ml
1
1
I I
I I
1
I I
I I
li
1
M
c
I I
I I
1
1
1
I I I
I I
I I I
1
I
I I
I lll
I I
I
I I
mil
it
D
I I
M
I lll
1
I I
1
1
M
I I I
I lll
I I
I I I
I lll
E
/
I I I
I I
I I I
1
I I
1
I I
I lll
I I
1
I lll
ill
I I
I I
I lll
25
TABLE VIIContinued
7
A
B
0
r>
E
y
0
H
X
J
K
X.
M
N
0
P
Q
&
8
T
TT
V
W
X
T
Z
A
I I
nun
i
ii
i
t
I lll
I I
I I
1
mi
i
i
n
i
i
in
B
I lll
I I
I lll
M
I I
1
I I
I lll
1
I I I
I
1
I I
Ml
c
/
I I
I I I
I I I
1
I lll
1
1
I I
I I I
1
1
1
I I
mil
i
D
I I
I lll
I lll
1
1
I I I
1
m
i
i
in
i
i
i
i
n
mi
E
m
n
i
in
mi
m
i
n
i
n
H I
n
i
n
i
i
4
A
B
C
D
E
F
0
H
I
J
K
X.
M
N
O
P
a
B
8
T
TT
V
W
X
Y
Z
A
/
I I I
I I
I I
1
m
n
H I
m
i
ii
i
n
in
i
n
B
I lll
I I
1
1
I I I
I I I
I I
W
/
N U
I I
1
1
1
1
1
1
1
c
I I
I lll
I I
I I I
I I
1
I I
I I
I I
t
I I I
I I
I I I
1
m
i
D
/
mi
i
in
i
i
mi
i
n
n
i
H I
m
H I
E
mini
n
mi
i
n
n
i
n
H I
i
i
n
ii
n
i
i
26
TABLE V IIContinued
14
A
B
C
D
E
P
G
H
I
J
K
X.
M
N
O
P
a
B
8
T
TT
V
W
X
T
Z
A
/
//
/
/
//
/
I H I
I
I I
1
1
I I
1
M
I lll
I I
W
B
//
///
/
///
/
////
/
/
/
//
M
Ml
I lll
1
a
I I I
1
1
1
1
W
I I
I I
1
I I I
1
1
I lll
nu
i
i
i
i
D
//
///
/
/
///
/
/
//
mi
i
i
H I
i
i
i
i
i
mi
i
E
I I I
I I I
1
I I
1
I I
1
I I I
I I
1
I I
I I
I H I
I I
1
I lll
10
A
B
C
D
E
P
O
H
I
J
X
L
M
N
O
P
O
B
8
T
tr
V
W
X
Y
Z
A
mi
i
H
in
i
i
n
H
n
tin
n
m
n
n
B
//
//
/
/
///
/
/
//
////
nu
i
in
m
i
in
O
I lll
I I I
1
1
1
1
1
1
1
I I
I I I
1
I I I
m
i
ii
i
in
D
/
/
////
/
mil
H I
in
n
i
n
Mil
I lll
E
////
I I
M
I I
I I I
Ml
I lll
1
I I I
I lll
1
27
TABLE VIIContinued
3
A
B
C
D
E
F
O
H
Z
J
X
L
M
N
O
p
Q
&
8
T
TT
V
W
X
Y
Z
A
//
II
1
Il l l
nu
i
n
i
Il l l
1
1
1
Il l l
Mi l
3
1
II
Il l l
M
II
II
Il l l
II
II
1
III
II
1
Il l l
1
c
III
II
III
Mi l
II
1
1
Il l l
II
1
II
1
1
1
II
III
D
/
/
II
III
III
Il l l
J
1
III
1
1
1
Il l l
III
II
Il l l
1
E
II
1111
1
Il l l
M
Il l l
1
II
1
II
III
Ml
1
18
A
B
C
D
E
F
O
H
I
J
X
L
M
N
O
P
Q
B
B
T
V
V
W
X
Y
Z
A
II
1
1
1
III
1
1
III
Ml
II
Il l l
1
Il l l
1
1
1
II
II
B
II
Ml
1
1
II
II
III
II
1
1
II
II
1
1
Ml
M
o
II
1
II
III
III
III
III
1
II
1
1
1
Il l l
Ml
1
III
D
1
1
I
f H J
Mi l
II
Ml
III
II
1
Il l l
II
1
E
II
III
1
1
M
III
II
II
II
1
III
II
1
1
Mi l
17
A
B
0
D
E
V
0
H
1
J
X
L
M
A
/
III
Il l l
1
1
II
1
1
1
1
B
///
///
///
/
//
/
/
//
C
/
///
////
/
III
1
1
II
D
III
I
1
II
III
Mi l l !
E
////
///
III
1
1
1
I
1
III
1
1
17
N
0
P
Q
B
8
T
TJ
V
W
X
Y
Z
A
/
Il l l
Il l l
III
II
1
Ml
B
////
III
II
Ml
II
1
III
c
///
/
/
//
Mi l l
1
II
II
D
/
II
1
1
M
M
II
E
/
///
///
/
/
II
I
1
II
28
The method which we are about to demonstrate is based upon the fact that the segments
from which the cipher groups are taken follow one another from a given initial point in a
regular succession, uninterrupted in this case except for a break of three segments representing
the two blank segments of the key plus one blank which is always present representing the
plain segment. To explain the principle of this method in detail, attention is directed to the
fact that, as a result of the system of encipherment, the series of successive cipher equivalents
for any given plain-text letter in any one of the five primary alphabets coincides with the
sequence of letters in that alphabet. The series will coincide with the complete alphabet except
for the omission of 1, 2, 3, or more letters depending upon the number of blank segments.
For example, turn to figure 1 and note that, in alphabet 1, the sequence of letters beginning
with A i s as follows: A U F Q Z E R H Y G W J . . . .
N ow it is patent that if we place letter A of the first primary alphabet in the plain segment,
its series of successive cipher equivalents coincides with the sequence of letters succeeding A in the
same alphabet, viz, U F Q Z E R H Y G W J O I D M N C . . . .
If we place another letter of the same primary alphabetfor example Zin the plain
segment, its series of successive cipher equivalents constitutes exactly the same sequence,
except with a different initial point, viz, E R H Y G W J O I D M N C . . . . In other
words the successive cipher equivalents for these 2 plain-text letters come from one and the same
cycle or sequence. Now, the same is true with respect to every other letter of alphabet 1, and
also of the other primary alphabets. Of course, the sequence is different for each primary
alphabet.
Since this cycle or sequence of letters is the same for all the letters of each primary alphabet,
only the series of successive cipher equivalents for one letter of each primary alphabet is necessary
in order to effect a complete reconstruction of that alphabet. In other words, if we can select
with accuracy the cipher equivalent for one and only one plain-text letter in each of the successive
115 secondary alphabets, we can then arrange these equivalents into five sequences of letters
which will coincide with the live primary alphabets, thus resulting in their reconstruction.
The reconstructed sequences will be complete except for the omission of one or more letters
representing the blank segments. If the numerical key consists of 23 numbers, three letters
will be missing from each sequence. These letters will be known, of course, but their relative
positions in the omitted section will have to be found later.
Obviously, the letter which will lend itself best to such a procedure is E, for it is the most
frequently occurring letter in English text. If, therefore, by a careful study of the individual
frequency tables applying to the columns of the superimposed messages, we can select the cipher
equivalent of only the letter E with certainty in the successive secondary alphabets, we shall
at once have the sequences of letters in the five primary alphabets and the solution of the
problem will be at hand. For example, if in a hypothetical sequence of these alphabets we
select the letters K, N, Q , and V , respectively, as the four successive cipher equivalents of E,
then this will mean that in primary alphabet 1 there is a sequence . . . K N Q V . . .,
providing a break in the numerical key does not exist between the members of the sequence
of key numbers applying to the segments concerned. Continuing this process, ultimately the
five primary alphabets can be completely reconstructed. But we must remember always that
this process is dependent upon the correct assumptions for the cipher equivalent of E in each
of the 115 secondary alphabets, or columns of cipher text.
Let us attempt such a reconstruction. Turning to the series of secondary alphabets given
in table V II, we try to find in each alphabet the letter which undoubtedly represents plain-text
letter E. At the very start we encounter difficulties. In alphabet 1A, the letters M and Y are
of equal frequency. There is no way of telling which letter represents E, so that we shall have
29
to consider both M and Y as possibilities. In alphabet 8A again we have difficulties, for both
J and Q have the same frequency. It begins to look like a very doubtful procedure. As we
go further along, the difficulties in selecting the representative of E increase rather than decrease
and the cryptanalyst becomes lost in a multiplicity of possibilities. Evidently this method,
as the preceding one, while theoretically correct, is practically out of the question because
of the limited size of each frequency table. In fact, it is doubtful whether we can select the
representative of E with certainty in any one of the A alphabets,
1
and certainly, if we cannot
do this with the letter that theoretically occurs the most frequently, we cannot do it with
any other letter.
It was at this point, when apparently a blank wall confronted the writer, and there seemed
little hope of solution, that he evolved the method which finally resulted in solution, and which
embodied such new principles that he was led to describe them in this paper. This method
had recourse to some simple mathematics, easy of comprehension and apph'cation when the
underlying principles have been grasped.
First, let us make what we have termed a "consolidated frequency" table for all of the
secondary alphabets applying to the first, or A, primary alphabet. This is done by collecting
the data contained in the individual frequency tables shown in table VII into one large table,
taking only the data applying to the letters of primary alphabet 1. This larger table is shown
below (table VIII).
1
It was found later that the cipher equivalent of E has the greatest frequency in only 3 out of the 23
alphabets. In one alphabet E did not occur at all, and in six cases it occurred only two times. It will be of
interest to the reader to study these tables for the information they contain with regard to the extreme degrees
of variation from the normal that small frequency tables can exhibit.
TABLB VIII.Consolidated frequency table for alphabet 1
Cipher
letter
A
B
C
D
K
p
< j
H
i
j
K
L
M
N
O
P
Q
B
S
T
tr
V
w
X
Y
z
Segment
1
1
4
1
3
3
1
5
1
3
1
1
1
2
3
5
2
S
2
1
2
2
1
1
1
4
3
2
1
4
3
3
3
1
1
1
12
1
3
2
2
1
5
2
1
2
3
1
2
2
3
3
1
1
1
18
2
4
1
1
1
2
1
3
1
3
1
3
1
2
2
1
4
3
20
2
1
1
1
1
1
5
1
6
1
1
6
4
4
2
1
6
1
3
3
1
1
1
2
2
4
2
1
5
2
1
2
22
1
4
1
3
4
1
1
3
3
4
2
3
2
2
1
1
i:
2
2
2
1
3
1
3
4
3
6
2
2
1
1
3
6
1
1
1
5
2
1
2
3
1
1
7
1
2
3
2
2
2
11
3
4
1
5
5
1
1
1
3
2
2
3
1
1
2
2
5
1
3
1
1
4
2
3
1
1
1
1
3
2
6
1
1
2
2
IB
2
2
1
1
2
2
6
1
1
1
3
3
1
i
4
4
2
13
1
4
2
3
1
1
1
1
3
2
2
6
4
1
2
1
2
23
3
1
5
4
3
3
4
1
4
3
1
3
1
l
2
2
6
5
1
2
3
1
2
2

1
2
2
21
2
1
1
6
1
4
1
1
1
2
1
1
2
1
1
1
6
5
7
2
7
1
2
1
1
4
2
2
1
6
1
1
2
1
1
3
4
1
3
2
2
1
5
2
3
5
1
2
1
2
3
1
2
14
1
2
1
1
2
1
5
1
2
1
1
2
1
5
4
2
5
10
6
1
2
3
1
1
2
2
2
4
2
5
3
2
1
2
2
2
4
5
1
2
1
4
1
1
1
4
7
18
2
1
1
1
3
1
1
3
6
2
4
1
4
1
1
1
2
2
17
1
3
4
1
1
2
1
1
1
1
1
4
4
3
2
1
6
Total fre-
quency
22
37
27
99
20
30
40
35
29
27
40
34
31
28
35
22
30
38
32
31
37
39
28
38
36
34
841
Num-
ber of
segments
occupied
14
13
16
17
13
17
14
15
12
14
15
14
16
15
14
11
12
16
14
17
15
16
14
17
12
12
Average
frequency
per
segment
1.58
2,85
1.69
2.30
1.64
1.76
2.86
2.34
2. 42
1.93
2.67
2. 43
1.94
1.86
2.50
2.00
2.50
2.37
2.28
1.82
2.47
2.45
2.00
2.24
3.00
2.83
841
Average frequency per cipher letter=-2g-=32.4 occurrences.
31
This consolidated frequency table is of a rather peculiar nature. Each column gives the
frequency of the cipher letters in a particular segment and there are 23 such columns, corre-
sponding to the 23 segments of the numerical- key. The numbers of the columns are determined
by, and coincide with,. the sequence of numbers in the cycle as given on page 16, viz, 1,8, 12, 16,
etc. Each row gives the frequency of a particular cipher letter in the successive segments and
since the columns succeed one another in the cyclic sequence, it follows that the frequencies
in the successive segments on a line with any given cipher letter form a definite sequence oj
frequencies. There being 26 cipher letters, there are 26 such rows or sequences of frequencies.
The total frequency for each cipher letter is given in the column labeled as such and the
average frequency for all cipher letters is then found to be 841-r-26=32.4 occurrences. The
number of different segments in which the cipher letter applying to any given line occurs is
indicated in the next column; and the average frequency per segment for each cipher letter
is given in the last column.
Before we can proceed it will be advisable to establish certain principles which will enable
us to follow the subsequent reasoning more easily. W e shall make use of alphabet 1 shown
in table V III, calling attention to the fact that the same principles apply to the other four
primary alphabets. In order to make the illustration comparable in all its details with the
real situation in the test problem, let us make the numerical key 23 numbers in length by
adding numbers 22 and 23 at the end of the key shown in figure 1 .
Let us see what successive plain-text letters the cipher letters A, B, and C represent in the
sequence of segments.
0. Plain text
A V K P L B S X T C N M D I O J W G Y H R E Z Q
B S X T C N M D I O J W G Y H R E Z Q F U A V K
C N M D I O J W G Y H R E Z Q F U A V K P L B S
It "will be noted that the successive plain-text letters which cipher letters A, B, and C repre-
sent constitute almost exactly the same sequence in the three lines. This follows from the
nature of the cipher system itself, and the cause of it has already been pointed out. In the B
line there is a section not present in the A line, consisting of the letters FUA; in ihe A line, the
section not present in the C line consists of the letters XT C; and in the C line, the section not
present in the B line consists of the letters PLB. This is due to the interruption in the numerical
key; the section omitted will consist of 3 sequent letters in each case, but these letters will be
d ifferent for every cipher letter.
Let us now accompany the sequence of the plain-text letters opposite each of the letters
A i B, and C, with a sequence of frequencies corresponding to their normal theoretical frequen-
cies * for English text.
1
These theoretical frequencies are given by Hitt on the basis of 200 letters of plain text. See Hitt, Parker.
Manual for the Solution of Military Ciphers, 1918, p. 0.
.
3Si
t 17 4 14 10 7 21 11 Id
V K P L B S X T C
32
FIGURE 7
A
12
I
6 JO 1U
N M D
a 6 is 2 i w is a 23 a
O J W G Y H K E Z Q
8 X T C N M D J W O Y H B, E Q F T J A V K
NMD O J W Q Y H B E Z Q F T 7 A V K P I . B
Now, since the sequences of plain-text letters represented by these sequences of frequencies
are the same, it follows that we can so arrange the latter as to make the successive individual
frequencies coincide; and if we make due allowance for the break in the sequences caused by
the omitted sections of 3 letters, the three sequences should coincide exactly. Thus:
33
FlCUKK 8
A
V K P L B S X T C N M D I O J W G Y H R E Z Q t J V K P L B S X T C N
6 X T O N M D I O J W O Y H R E Z Q F t T A V K [ W ] S X T C N
5* 5z^ -3z"^ = 3:5> : ^ ^ ^ St St Sz "^ S z rt 5r 5r--3f
M D I O J W O Y H B E Z Q F t T A V K P L B S [ M
^ = =2:3 ^ ^ ^ S^ Sr Sz ^ ^ 3z ^ =: "~" 5^ &
In order to make the sequences coincide, we displaced the B sequence five intervals to
the right of the A sequence, and the C sequence four intervals to the right of the B sequence.
Let us reverse the order of these letters, A, B, and C, and space them in accordance with the
number of intervals which each sequence of frequencies has been shifted relative to the others.
Thus:
4 3 3 1 8 6 4 3 3 1
C ... B .... A
Refer now to the illustrative cipher alphabet in figure 1 and note that this corresponds to the
order of these letters A, B, and C in this primary alphabet. W e have determined the order of
these letters in our alphabet merely by correctly superimposing or shifting the three sequences of
frequencies relative to one another so as to make the individual frequencies coincide.
Now had we not known what letters these individual frequencies in each sequence of fre-
quencies represented but had merely been given the sequence of frequencies themselves, it would
still have been just as easy to find the correct relative positions of the three sequences from a
comparison of the positions of high and low points in each sequence of frequencies. In other
words, we do not need to know what letters the individual frequencies in each sequence offrequencies
represent; it is still possible to determine the relative positions (in the primary alphabet) of the
letters applying to each sequence in the cipher alphabet by a study of the positions of the high and
low points in each sequence of frequencies. No analysis whatever of the individual frequencies
is necessary, the entire frequency table being treated as an ordinary statistical curve. This, in
its final analysis, is the meaning of the proposition stated in the opening paragraph of this
34
paper.
1
It thus follows that the five alphabets of our problem may be reconstructed, without
a knowledge of what letter any individual frequency in the sequences of frequencies (as shown in
table V III) represents, by an analysis of these frequency tables considered as true statistical
curves.
Let us return now to the test messages. Table V III represents a set of 26 sequences of fre-
quencies similar in origin to those for A, B, and C in the illustrative alphabet above. W e could
superimpose these sequences in the same manner and as easily as we did in the messages them-
selves were it not for two circumstances: First, we know that there is an interruption of three
blanks in the cycle wliich we have reconstructed but do not know where these blanks must be
inserted. Consequently, some allowance must be made for the blank segments in each sequence
of frequencies. Secondly, the individual frequencies in each sequence of frequencies in our prob-
lem do not exactly correspond to the theoretical frequencies of the plain-text letters to which
they apply but only correspond approximately to the theoretical. In some cases this approxima-
tion is far from close because of the paucity of text, and this will make the determination of
the correct relative positions of two sequences a much more difficult process than was the
case with the illustrative sequences above.
We are, therefore, confronted with the problem of superimposing the sequences of frequencies
correctly without a knowledge of these two factors, and this we shall accomplish by a slight modi-
fication of method and a recourse to some simple mathematics.
First, as to the modification of method due to our ignorance of the exact location of the
break of three intervals in the numerical key: this consists in superimposing sequences, not to
find the relative positions of any pair of sequences but to find such sequences as are one and only
one interval apart; i.e., sequences wliich represent a relative displacement of only one interval.
The reason for this step is now to be explained.
Let us consider the sequence of theoretical frequencies corresponding to the cipher letter A
and the letter which immediately follows it in the illustrative alphabet, viz, U, arranging the two
sequences as though we had only reconstructed the cycle and had not as yet determined the
numerical key. Let us begin both sequences with segment 9, the first segment in the key.
1
The ordinary frequency table applying to a plain text or a cipher alphabet does not correspond to the
ordinary frequency distribution of statistical work. In the latter, the position of the points along one of the axes
of the graph and their extension along the other axis are either causally related, or the curve treats of data which,
being subject to the operation of the laws of probability, form the normal, or Q uetelet, curve of error. In the
former, the positions and extensions of the coordinates are not related in any way unless one considers the arbi-
trary order of the letters of the alphabet as constituting a cause. The positions of the coordinates in a crypto-
graphic curve were determined many centuries ago when the English language was first evolved.
But the sequences of frequencies in table V III are not similar in origin to the ordinary plain-text or cipher
alphabet frequency tables of cryptographic work. They are, in fact, closely related to certain frequency distribu-
tions of statistical data because the position and extensions of the coordinates are absolutely determined by a cause
other than the arbitrary order of the letters of the English alphabet. These two characteristics of the curves of
a aeries of secozidary alphabets may be varied at will by changing the sequence of letters in the primary alphabet.
Any set of frequency distributions applying to a series of secondary alphabets derived from a variable primary
alphabet may be treated in the same mathematical manner as these will be treated in the subsequent pages.
35
FldDMl. 0
A
9 17 4 14 10 7 21 11 1 (I 20 ID 12 3 8 1.1 2 1 13 18 8 22 23
V K P L B S XT C N M D I O J WG Y H R E Z Q 1
8 17 4
]V K P
14 10 7 21 11 1
L B 8 . . .
K
rt
rt
17 4 14 10 7 21 11 18 90 18 12 3 6 14 2 1 IS 18 8 22 23
A V XP L B S XT C N M D I O J WO Y H B E Z [
8 17 4 14
]A VKP
7
B
21 11 ID
tf
CO
It is evident that we may superimpose these sequences correctly by shifting the A sequence
one space to the right of the U sequence or the U sequence one space to the left of the A sequence.
Thus:
F IGURE 10
A
8 17 4 14 10 7 21 11 1 20 19 12 1 5 15 2 1 13 18 8 22 23 17 4 14 10 7 21 11 1
V K P L B S XT C N M D I O J WG Y H R B Z Q [ ]V K P L B S . . .
TT
17 4 14 10 7 21 11 1 10 1 12 B 8 IS 2 1 13 18 8 32 23
A V K P L B S XT C N M D I O J WG Y H R E Z [
B 17 4 14 10 7 21 11 16
]A V K P L B S . .
rt
a
Note now that complete and perfect coincidence between successive pairs of superimposed
segments is obtained except in two cases, viz, those involving the letters A and Q in the U and A
lines, respectively. This is due to the fact that the break in the sequence comes between seg-
ments 23 and 9. The frequency of letter A of the U sequence should be matched with the fre-
quency of A in the A sequence, but the latter does not occur because of the break in the numerical
key. The same is the case with the letter Q of the A sequence.
36
Now suppose that we did not know where the break in the numerical key falls, and let us
superimpose the sequences again. Thus:
F IGURE 11
A
8 17 4 14 10 7 21 11 1 20 19 13 3 6 15 2 1 13 18 22 23 X 9 17 4 14 10 7 21 11 16
V K P L B S X T CNMDI O J W G Y H B E Z Q V K P L B S . . .
B 17 4 14 10 7 21 11 10 30 19 12 8 6 18 2 1 13 18 22 23 X 8 17
A V K P L B 8 X T CNMD I O J W O T H B E Z A V
4 14 10 7 21 11 10
K P L B 8 . .
It is seen that perfect coincidence is still maintained throughout except in the case of the one
pair of segments containing the letters A and Q of the U and A lines, respectively. By omitting
the three blank segments representing the place where the break occurs, we have brought the
letters A and Q into an incorrect superimposition. But the amount of error due to the super-
imposition of one pair of incorrect segments as against the correct superimposition of 22 other
pairs is of so little consequence that it may be neglected altogether. In other words, when we
superimpose sequences which are only one interval apart, relative to each other, we may neglect
the discrepancy that would be due to our ignorance of where the break in the numerical key
comes.
Now suppose that we did not know that the letter U immediately follows A in this illustra-
tive primary alphabet, and had only a table containing the frequencies applying to the cipher
letters (similar to those shown in table V III). It is evident that by placing the A sequence one
interval to the right of all other sequences successively, and choosing that sequence which most
closely coincides with the A sequence in the positions of the high and low points, we shall thus
have determined what letter immediately follows A in the primary alphabet; this follows because
the letter applying to that sequence of frequencies occupies a definite position in the cipher
alphabet, viz, it follows A. In this case, the U sequence would be chosen and we would conclude
that the sequence in the primary alphabet is . . . A U . . . . Taking the U sequence, the same
operation is performed with the other sequences as before, and we thus find the letter that
follows U in the primary alphabet. Theoretically, therefore, we should be able to reconstruct
the complete primary alphabet by this method, and thus overcome the difficulty due to our
failure to know exactly where the interruption in the numerical key falls.
In this process of matching sequences of frequencies to decide wliich one most closely
coincides with a given sequence, we cannot depend upon a mere ocular examination and com-
parison. W e must reduce the operation to a mathematical method. This, we shall proceed
to consider.
Let us return to table V III and select that line for experiment which gives the best indica-
tions of representing the closest approximation to a theoretical frequency table containing as
37
few elements as are contained in the average line in the table. In a theoretical frequency table
of small size such as the one shown in figure 12 only the high-frequency letters are represented;
the low-frequency letters are absent. The average frequency per letter that does occur is
evidently the total frequency, viz, 34, divided by the number of different letters that go to
make up this total, viz, 12. This quotient is 34-8-12=2.83.
FIGURE 12
A B C D B F O H I J K L M N O P Q R S T t T V W Z T Z
Total frequency=34
Number of different lettera** 12
Average frequency =*=2.83
Note now that the Y line of frequencies in table VTII averages 3.00 occurrences per segment;
that is, the average frequency per plain-text letter which Y represents is 3.00. This is even a
little better than the average theoretical frequency per letter as determined above. Let us con-
sider the Y sequence of frequencies, therefore, as representing the closest approximation to a
theoretical frequency table of a similar total of occurrences.
Following the method discussed above, let us see if we can find the letter in alphabet 1
which follows Y.
Taking our Y sequence of frequencies, let us apply it to all the other sequences, placing the Y
sequence one interval to the right of the other sequences. Thus, with A the Y sequences are
placed in these relative positions:
FIGURE 13
A
Y
1 8
2
5
l
12
1
1
8
16
2
1
12
20
2
10
2
1
4
20
22
1
2
15
2
22
6
1
15
11
6
6
2
11
i t
2
5
13
1
2
18
23
13
8
23
21
2
2
9
7
2
6
21
4
7
14
1
4
10
2
14
S
2
10
18
2
7
3
17
2
18
1
17 1
We shall now proceed to find an abstract number such as will indicate the degree to which
these two sequences of frequencies agree, or fit, when placed with reference to each other as in
figure 13. It is evident that when we strike the letter which really follows Y in alphabet 1, the
corresponding sequence of frequences concerned should give the best fit with the Y sequence
and thus produce the greatest degree of coincidence.
Let us now compare the two superimposed sequences above, segment by segment. In the
upper one of the first pair of superimposed segments there are 2 occurrences of A; in the lower one,
5 occurrences of Y. The first pair of segments agree, therefore, in 2 occurrences; i.e., there are 2
coincidences. In the next pair of segments, an occurrence of 1 in the Y sequence is matched by
an occurrence of 1 in the A sequence; i.e., there is 1 coincidence. Let us go through the rest of
the segments in the same manner. The results are given in figure 14.
I,
38
FIGURE 14
A
T
i
Coincidences
8
2
;
5
i
2
12
1
1
8
1
10
2
1
12
1
20
2
16
0
i
1
4
20
1
22
1
2
0
15
2
22
0
, f i .
1
16
0
11

6
6
2
11
0
IB
2
6
0
13
1
2
19
1
23
13
9
23
21
2
2
e
2
7
2
0
21
2
4
7
14
l'
4
0
10
2
14
0
a
2
10
0
18
2
7
3
2
17 >
/
2
18
0
1
17

"
1
W e have a totnl of 12 coincidences. But we must also take into consideration the number
of noncoincidences between the two sequences; for these, as can easily be demonstrated, are of
equal importance with the coincidences. In the two hypothetical sequences given below, both
with the sumo total frequency, the number of coincidences is very high, viz, 28, yet the two
sequences do not agree closely at all, for there are 45 noncoincidences.
F IGURE 15
Occurrences (101)
Coincidences (28)
Noncoincidences (45)
5
5
i n
,5
0
0
2
>
2
5
5
5
6
fi
6
0
4
4
S
4
0
n
5
5
i n
5
0
1
1
1
2
2
4
? ,
0
G
6
1?
(\
0
n
2
2
4
2
0
6
r>
6
4
4
4
4
4
4
5
5
6
2
?
2
4
4
8
4
0
5
5
6
5
5
5
Let us find the totnl of noncoincidences, therefore, between the Y and the A sequences.
In the first pair of segments there are 3 noncoincidences; in the second pair, none; in the third
pair 1, etc. Let us now add these to our table, and include also the number of occurrences in
the segments, for we shall have need of this information very soon.
Fl GUKE 16
A
Coincidences
Noncoincide
Occurrences
Y
i
i
nccs
s
2
5
i
2
3
7
12
1
1
8
1
0
2
16
2
1
12
1
1
3
20
2
16
0
2
2
2
1
4
20
1
3
5
22
1
2
0
1
1
15
2
22
0
2
2
G
1
If,
0
1
1
11
fi
0
0
0
5
2
11
0
2
2
in
2
6
0
2
2
13
1
2
19
1
1
3
23
13
0
0
0
9
23
0
0
0
21
2
2
9
2
n
4
7
2
6
21
2
4
8
4
7
0
0
0
14
1
4
0
1
1
10
2
14
0
2
2
3
2
10
0
2
2
18
2
7
3
2
5
9
17
2
18
0
2
2
i
17
= 12
= 34
= 58
W e find that the total of coincidences is 12, that of noncoincidences, 34. The difference
is 12 34 =22. Were the sequences in closer agreement, this difference would bo a
positive quantity; but as a rule, we shall find it to be a negative quantity in our work because
of the fact that the frequencies are relatively low throughout. In this case, then, the number
of noncoincidences is 22 greater than the number of coincidences. This difference between
39
the totals of coincidences and noncoincidences will be used as the basis for the determination
of the degree to which two sequences coincide, and inasmuch as we shall have a great many
such differences to compute, a short cut to their determination will be of use. If we subtract
the total of occurrences from three times the total of coincidences, we can find this difference
directly without having to count up the number of noncoincidences. Thus, in this case,
(3X12) 58 =22. In all subsequent determinations we shall use this method.
Now it is obvious that the number of coincidences as well as the number of noncoincidences
is not only a function of the distribution of the occurrences in each sequence of frequencies but
also of the total number of occurrences. It is patent that in one pair of sequences with a
greater total number of occurrences than in another pair, the totals of coincidences and non-
coincidences might be greater in the former than in the latter from the mere fact that there
are more opportunities for coincidence and noncoincidence in the former case. We should
therefore take into consideration the total number of occurrences in the two superimposed
sequences, and the most logical correction would be to divide the difference between the totals
of coincidences and noncoincidences by the total number of occurrences of all the segments.
For example, it is only reasonable to place more reliance upon a case in which out of 30 occur-
rences the difference between the totals of coincidences and noncoincidences is +10, than
upon a case in which out of 60 occurrences the difference between these same totals is also +10.
In the former case, the quotient obtained b}' dividing the difference, +10, by the total occur-
rences, 30, is +.33; in the latter case, the quotient obtained by dividing +10 by 60 is only
+.17, only half as much.
To this quotient, wliich indicates in a general way the "goodness of fit" of the two super-
imposed sequences, and which is obtained by dividing the difference between the totals of
coincidences and noncoincidences by the total occurrences, we have applied the name "Index
of coincidence ".' It is evident that the greater the index of coincidence, the better is the
agreement between the superimposed sequences, and thus, the closer is the fit. W here the
tw,o sequences are relatively low in frequency, the total of noncoincidences will, as a rule, be
greater than the total of coincidences, so that the difference will usually be a negative quantity
and the index will also be negative. As these negative indices approach 0, they become closer
to positive indices, so that when we are dealing with negative indices, the lowest absolute index
will indicate the greatest coincidence. Thus, an index of .03 will indicate a much better
fit than an index of .35.
Returning now to the case in hand, we found (he difference between the totals of coincidences
and noncoincidences to be 22. Since a total of 58 occurrences enters into the formation of
these two tables, then the index of coincidence for the assumption that A follows Y in alphabet 1
is -22 -T- 58 = -.38.
Let us perform the same calculations for the rest of the letters in table V1IT (p. 30), omitting,
of course, Y. The data arc given in table IX.
See S. KuJ lback, loc. cit., Sectious V I and V II for other tvnd more reliable for matching alphabets.
40
TABLE IX
Letter
A
B
C
D
E
F
G
H
I
Total
occur-
rences
58
73
63
75
56
66
76
71
65
Coinci-
dences
12
15
13
16
12
11
8
13
10
Differ-
ences
-22
-28
-24
-27
-20
-33
-62
-32
-35
Indices
of coin-
cidence
-.38
-.37
-.38
-.36
-.36
-.60
-.69
-.45
-.54
Letter
J
K
L
M
N
0
P
Q
R
Total
occur-
rences
63
76
70
67
64
71
58
66
74
Coinci-
dences
15
13
16
9
16
10
17
17
15
Differ-
ences
-18
-37
-22
-40
-16
-41
- 7
-15
-29
Indices
of coin-
cidence
-.29
-.40
-.28
-.60
-.24
-.58
-.12
-.23
-.34
Letter
S
T
U
V
W
X
z
Total
occur-
rences
68
67
74
75
64
74
70
Coinci-
dence*
14
11
18
12
7
16
11
Differ-
ences
-26
-34
-20
-39
-43
-26
-37
Indices
of coin-
cidence
-.38
-.61
-.27
-.62
-.67
-.35
-.53
As stated above, the best fit is obtained when the index of coincidence is the greatest positive
quantity. In none of the cases above is the index of coincidence positive, but the value for
P, viz, .12, approaches the nearest to a positive quantity, and therefore represents the greatest
degree of coincidence. The next greatest index is given by the letter Q ; but inasmuch as the
index for P is almost twice as great as that for Q, we may conclude that it is the letter P, and
not the letter Q, which immediately succeeds Y in the alphabet 1.
We may now proceed to find the letter that follows P. The same operations are performed
with the letter P as with the letter Y
1
, this time using the frequency of P as the base and
trying it one interval removed from the frequencies of all letters except Y and P, for, as the posi-
tion of each letter is determined, it can be automatically omitted from the succeeding calculations.
The data are as follows:
TABLK X
Letter
A
B
C
D
E
F
G
H
Total
occur-
rences
44
59
49
61
42
52
62
57
Coinci-
dences
11
10
13
U
9
10
19
13
Differ-
ences
-11
-29
-10
-28
-15
-22
-35
-18
Indices
of coin-
cidence
-.25
-.49
-.20
-.46
-.36
-.42
-.57
-.32
Letter
I
J
K
L
M
N
0
Q
Total
occur-
rences
51
49
62
66
53
50
67
52
Coinci-
dences
8
11
12
8
11
9
9
7
Differ-
ences
-27
-16
-26
32
-20
-23
-30
-31
Indices
of coin-
cidence
-.53
-.36
.42
-.57
-.38
-.46
-.53
-.60
Letter
R
S
T
U
V
W
X
z
Total
occur-
rences
60
54
53
60
61
50
60
56
Coinci-
dences
14
8
11
9
7
8
11
8
Differ-
ences
-18
-30
-20
-33
-40
-26
-27
-32
Indices
of coin-
cidence
-.30
-.56
.38
-.56
-.65
-.52
-.45
-.66
It is seen that the index for letter C, viz, .20, represents the greatest degree of coincidence.
But the index for A is .25 and that for R is .30. In other words, the index for C is only
.05 greater than that for A, and .10 greater than that for R. The question then arises: Is a
difference of .05 or .10 in favor of C over A and R, respectively, a significant difference? In
1
The method which has been given here for det ermi ni ng Die index of coincidence is not very effective when
the alphabets arc very much different in size. The mathematical study mentioned in Note 3 on page 11 has
provided a method which is effective in such cases, and with its help a great deal of work could have been saved
in the following calculations. For, as soon as the proper position of the P sequence of frequencies with respect
to the Y sequence had been determined, the two could have been combined so as to yield a base of twice as many
letters. Similarly, the knowledge that P is followed by C would permit an additional adj unction of frequencies
to the base. Since the accuracy obtained in matching increases with the number of letters involved in the test,
the further results could have been obtained with much less difficultv.
41
other words, might hot the letter A or R follow P, instead of C? The answer may be found
by modifying the method in one particular. W e have been superimposing sequences with a
relative displacement of but one interval. If now we superimpose sequences with a relative
displacement of two intervals, the error, due to the failure to take into account the break in the
numerical key, will be greater than it is with a relative displacement of one interval, for now
there will be two incorrect pairs of superimposed letters, but still the results will be significant.
Let us, therefore, test out all the letters which are less than twice the index of C, with the Y
sequence removed two intervals. Thus with A, the sequences are in this position:
FIGURE 17
A
Y
2
5
1
1
2
1
2 1
4
1 2 1
2
1
2 2
2 2
2 6
1
2
2
2
7 2
2
The data for the letters which may possibly follow P, when tested with the Y sequence
at two intervals removed, are as follows:
TABLE XI
Letter
A
C
E
Total
occur-
rences
58
63
56
Coinci-
dences
9
20
13
Differ-
ences
-31
- 3
17
Indices
of coin-
cidence
-.55
-.05
-.30
Letter
H
J
M
Total
occur-
rences
71
63
67
Coinci-
dences
16
10
14
Differ-
ences
-23
-33
25
Indices
of coin-
cidence
-.32
-.52
-.37
Letter
R
T
Total
occur-
rences
74
67
Colod-
denow
11
13
Differ-
ences
-41
-28
Indices
of coin-
cidence
-.65
-.42
These calculations show conclusively that C follows P in alphabet 1. In the tables showing
the calculations for the reconstruction of alphabet 1, whenever the first calculation, using
one-interval data, fails to show a letter whose index of coincidence is at least twice as great
as its nearest rival, a secondary calculation will be made using two-interval data. In two
cases, viz, for the letters following K and Z , it will be noted that three-interval data were
employed to determine the correct letter subsequent to an inconclusive secondary calculation.
The tables containing the rest of the data for the reconstruction of alphabet 1 are given below:
TABLE XII.Data for reconstruction of primary alphabet N o. 1
Letter
Total
occur-
renow
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
Y
A
B
C
D
E
F
G
H
I
J
K
' L
M
N
0
P
Q
R
S
T
U
V
W
X
z
68
73
63
76
56
66
76
71
65
63
76
70
67
64
71
58
66
74
68
67
74
75
64
74
70
12
16
13
16
12
11
8
13
10
16
13
16
9
16
10
17
17
15
14
11
18
12
7
16
11
-22
-27
-24
-27
-20
-38
-52
-32
-35
-18
-37
-22
-40
-16
-41
- 7
-15
-29
-26
-34
-20
-39
-43
-26
-37
-.38
-.37
-.38
-.36
-.36
-.50
-.69
-.45
-.54
-.29
-.40
-.28
-.60
-.24
-.58
-.12
-.23
-.34
-.38
-.51
-.27
-.52
-.67
-.35
-.53
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
F
A
B
C
D
E
F
G
H
I
J
K
L
M
N
0
< J
R
S
T
U
V
W
X
z
44
59
49
61
42
52
62
57
51
49
62
56
53
50
57
52
60
54
53
60
61
50
60
56
11
10
13
11
9
10
19
13
8
11
12
8
11
9
9
7
14
8
11
9
7
8
11
8
-11
-29
-10
-28
-15
-22
-.35.
-18
-27
-16
-26
-32
-20
-23
-30
-31
-18
-30
-20
-33
-40
-26
-27
-32
-.25
-.49
-.20
-.46
-.36
-.42
-.57
r-.32
-.53
-.36
-.42
-.57
-.38
-.46
-.53
-.60
-.30
-.56
-.38
-.55
-.65
-.52
-.45
-.56
Y at two intervals
A
C
E
H
J
M
R
T
58
63
56
71
63
67
74
67
9
20
13
16
10
14
11
13
-31
- 3
-17
-23
-33
-25
-41
-28
-.55
-.05
-.30
-.32
-.52
-.37
-.55
-.42
Letter
Totsl
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
C
A
B
D
E
F
G
H
I
J
K
L
M
N
0
Q
R
S
T
U
V
W
X
z
49
64
68
47
67
67
62
, 56
54
67
61
58
55
62
57
65
59
58
65
66
56
65
61
14
14
18
9
13
15
14
11
13
11
18
17
12
20
11
16
10
15
9
16
14
13
13
- 7
-22
-12
-20
-18
-22
-20
-23
-15
-34
- 7
- 7
-19
+ 2
-24
-17
-29
-13
-38
-18
-13
-26
-22
-.14
-.34
-.18
-.43
-.32
-.33
-.32
-.41
-.28
-.51
-.11
-.12
-.35
+ .03
-.42
-.26
-.49
-.22
-.58
-.27
-.24
-.40
-.36
43
TABLE XII.Data for rtcomiruction of primary alphabet N o. /Continued
Letter
Total
occur-
rences
Colnd-
dcnoM
Differ-
ences
Indlot*
of coin-
cidence
O
A
B
D
;
f c
'P
G
'H
1
J
K
L
'11
ti
Q
"R
's
T
U
V
W
X
z
58
72
74
65
B5
76
70
64
62
75
69
66
63
65
73
67
66
73
74
64
73
69
13
11
17
11
15
13
13
13
9
16
16
12
17
14
14
15
17
25
17
14
13
9
-19
-89
-23
-22
-20
-36
-81
-25
-35
-27
-21
-30
-12
-23
-31
-22
-16
+ 2
-23
-22
-34
-42
-.33
-.84
-.31
.40
-.31
-.48
-.44
-.89
-.7
-.36
-.30
.46
-.19
-.35
-.43
-.33
-.23
+ .03
-.31
-.34
-.47
-.61
Letter
ToUl
occur-
rences
Oolncl-'
denoes
blffer-
eocee
Indices
of coin-
cidence
U
A
B
D
E
F
G
H
1
J
X
L
11
tt
Q
R
S
T
V
W
X
Z
69
74
76
57
67
78
72
66
64
77
71
68
65
67
12
15
15
13
13
26
15
13
15
13
14
19
14
9
75 ! 16
69
68
76
65
75
71
12
16
16
21
14
17
-23
-29
-31
-18
-28
- 3
-27
-27
-19
-38
-29
-11
-23
-40
-27
-33
-23
-28
- 2
-33
-20
-.39
-.39
-.41
-.32
-.42
-.03
-.38
.41
-.30
-.49
-.41
-.16
-.35
-.60
-.36
-.48
-.34
-.37
-.03
44
-.28
O at two intervale
G
W
75
63
26
12
+ 3
-27
+ .04
-.43
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
0
A
B
D
E
F
H
I
J
K
L
M
N
4
R
S
T
V
W
X
Z
62
77
79
60
70
76
69
67
80
74
71
68
70
78
72
71
79
68
78
74
16
16
15
12
13
15
14
19
14
15
16
10
12
16
13
13
20
11
17
8
-14
-29
-84
-24
-81
-30
-27
-10
-88
-29
-28
-88
-34
-80
-38
-32
-19
-35
-27
-50
-.23
-.38
-.43
-.40
-.44
-.40
-.89
-.16
-.48
-.89
-.32
-.66
-.49
-.39
-.46
-.45
-.24
-.52
-.35
-.68
XT at two intervals
A
J
V
59
64
76
17
14
18
- 8
-22
-22
-.14
-.34
-.29
44
TABLE XII.Data /or reconstruction of primary alphabet N o. 1Continued
I/etter
Total
occur-
rences
Coinci-
dences
Dlfler-
noes
Indices
of coin-
cidence
A
B
D
E
F
H
I
a
K
L
11
N
Q
R
5
T
V
W
X
z
59
61
42
62
67
61
49
62
66
63
60
62
60
64
63
61
60
60
66
13
16
8
14
11
9
9
17
13
12
Q
8
13
12
11
13
8
16
12
-20
-16
-18
-10
-24
-24
-22
-11
-17
-17
-23
-28
-21
-18
-20
-22
-26
-16
-20
-.34
-.26
-.43
-.19
-.42
-.47
-.46
-.18
-.80
-.32
-.46
-.64
-.36
-.33
-.38
-.36
-.62
-.25
-.36
Q at two intervals
D
F
K
X
79
70
80
78
14
17
29
21
-37
-19
+ 7
-16
-.47
-.27
+ .09
-.19
Letter
Total
occur-
rences
Coinci-
dences
D iffer-
ences
Indices
of coin-
cidence
K
B
D
E
F
H
I
J
L
II
N
Q
R
S
T
V
W
X
Z
77
79
60
70
76
69
67
74
71
68
70
78
72
71
79
68
78
74
11
23
8
13
20
12
11
13
9
16
20
16
13
14
17
11
16
23
-44
-10
-36
-31
-16
-33
-34
-36
-63
-23
-10
-38
-33
-29
-28
-36
-30
- 6
-.67
-.13
-.60
-.44
-.20
-.48
-.51
-.47
-.75
-.84
-.14
-.49
-.46
-.41
-.35
-.51
-.39
-.07
A at two intervals
D
H
Q
Z
61
74
69
73
16
11
8
12
-16
-41
-45
-37
-.26
-.55
-.65
-.60
Q at three intervals
D
Z
79
74
28
18
+ 6
-20
+ .06
-.27
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
D
B
E
F
H
I
J
L
M
N
Q
R
S
T
V
W
X
Z
76
59
69
74
68
66
73
70
67
69
77
71
70
78
67
77
73
16
17
13
16
16
16
16
19
16
16
17
22
22
24
20
16
14
-31
- 8
-30
-26
-23
-21
-26
-13
-19
-21
-26
- 6
- 4
- 6
- 7
-29
-31
-.41
-.14
.44
-.38
-.34
-.32
-.34
-.19
-.28
-.30
-.34
-.07
-.06
-.08
-.10
-.38
-.43
1C at two intervale
E
S
T
V
W
60
72
71
79
68
13
20
18
28
16
-21
-12
-17
+ 5
-20
-.35
-.17
-.24
+ .06
-.30
TABLE XII.Data for reconstruction of primary alphabtt N o. 1Continued
Utter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
V
B
E
F
H
I
J
L
U
N
Q
R
S
T
W
X
z
78
59
60
74
68
66
73
70
67
69
77
71
70
67
77
73
14
16
13
16
11
12
10
18
17
12
16
14
20
16
26
13
-84
-11
80
-29
-36
-30
-48
-31
-18
-33
-29
29
-10
-12
+ 1
-34
-.46
-.19
-.44
-.39
-.82
-.46
-.69
-.44
-.24
-.48
-.38
-.41
.14
-.19
+ .01
-.47
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
Z
B
E
F
H
I
J
L
M
N
Q
R
S
T
W
Z
76
68
68
73
67
65
72
69
66
68
76
70
69
66
72
16
11
20
19
17
16
20
16
16
16
28
16
16
11
13
-30
-25
- 8
-16
16
-20
-12
-21
-21
-20
+ 8
-25
-21
-33
-33
-.40
-.43
-.12
-.22
-.24
-.31
-.17
-.30
-.32
-.29
+ .11
-.36
-.30
-.50
-.46
V at two intervals
F
L
R
69
73
77
15
19
26
-24
-16
+ 1
-.35
-.22
+ .01
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
B
B
E
F
H
I
J
L
U
N
Q
S
T
W
Z
76
68
68
73
67
66
72
69
66
68
70
69
66
72
16
10
21
11
16
10
18
14
18
14
23
19
10
14
-27
-28
- 3
-40
-22
-26
-18
-27
-12
-26
1
-12
-36
-30
-.36
-.48
.04
-.65
-.33
-.39
-.25
-.39
-.18
-.38
-.01
-.17
-.65
-.42
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
S
B
E
F
H
I
J
L
H
N
Q
T
W
Z
69
62
62
67
61
69
66
63
60
62
63
60
66
13
12
12
20
14
14
13
17
12
9
12
16
13
-30
-16
-26
7
-19
-17
-27
-12
-24
-36
-27
-12
-27
-.44
-.31
-.42
-.10
-.31
-.29
-.41
-.19
-.40
-.67
-.43
-.20
.41
S at two intervals
H
U
W
73
69
66
23
16
13
4
-21
-27
-.06
-.30
-.41
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
H
B
E
F
I
J
L
M
N
Q
T
W
Z
72
55
65
64
62
69
66
63
65
66
63
69
9
14
14
10
15
18
15
17
12
15
20
16
-45
-13
-23
-34
-17
-15
-21
-12
-29
-21
O
-21
-.63
-.24
-.35
-.63
-.27
-.22
-.32
-.18
-.46
-.32
-.05
-.30
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
W
B
E
F
I
J
L
W
N
Q
T
Z
65
48
68
57
55
62
59
66
68
59
62
20
10
13
7
13
11
12
14
14
18
10
~ ~ o
-18
19
-36
-16
-29
-23
-14
-16
- 5
-32
-.08
-.38
-.33
-.63
-.29
-.47
-.39
-.25
-.28
-.09
-.52
H at two intervals
B 72
RA
DO
26
1 T
17
+ 3
1 C
16
+ .04
OQ
-2o
46
TABLE XII.Data for reconstruction of primary alphabet N o. 1Continued
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
B
E
F
I
J
L
M
N
Q
T
Z
67
67
66
64
71
68
65
67
69
71
8
15
23
12
11
15
8
5
14
11
-33
-22
+ 3
-28
-38
-23
-41
-62
-27
-38
-.68
-.33
+ .05
-.44
.64
-.34
-.63
-.78
-.39
-.64
Q
E
F
J
L
M
T
Z
50
60
57
64
61
62
64
11
9
9
13
20
14
6
-17
-33
-30
-25
1
-20
-46
-.34
-.55
-.53
-.39
-.02
-.32
-.72
Z
E
J
L
T
54
64
68
66
16
9
9
16
- 6
-37
41
8
-.11
-.58
60
-.12
F at two intervals
E
T
60
61
14
15
- 8
-16
-.16
-.26
M at three intervals
E
T
51
62
16
18
-3
-8
-.06
-.13
Letter
Total
occur-
rences
Coinci-
dences
D iffer-
ences
Indices
of coin-
cidence
I
E
F
J
L
M
N
Q
T
Z
49
59
66
63
60
57
69
61
63
10
11
6
13
9
20
10
15
12
-19
-26
-38
-24
-33
+ 3
-29
-16
-27
-.39
.44
-.68
-.38
-.55
+ .05
-.49
.26
-.43
M
E
F
J
L
T
Z
51
61
58
65
63
65
10
20
14
13
18
11
-21
1
-16
-26
- 9
-32
-.41
.02
-.27
-.40
-.14
-.49
E
J
L
T
47
61
59
13
12
g
Q
' ' Q
-25
32
-.17
-.41
54
Letter
Total
occur-
rences
Colnol-
denoae
Differ-
ences
Indices
of coin-
cidence
N
E
F
J
L
M
Q
T
Z
48
68
65
62
59
58
60
62
11
10
15
13
16
14
13
14
-16
-28
-10
-23
-11
16
21
20
-.31
-.48
-.18
-.37
-.19
-.28
-.36
-.32
I at two interval*
J
M
Q
56
60
59
13
10
17
-17
-30
- 8
-.30
-.50
-.14
F
E
J
L
T
Z
50
60
64
62
64
8
13
12
13
18
-26
21
-28
-23
+ 8
-.52
-.35 1
-.44
-.37
+ .18
J
L
T
61
58
18
12
- 7
-22
-.11
-.38
L
T 65 22 + 1
+ .02
47
The now completed alphabet 1,- which is written tentatively on a disk and mounted upon
the base, is as follows:
1 2 3 4 S < 7 8 9 10 11 12 13 14 IS 18 17 18 19 20 21 23 23 24 28 28
Alphabet 1 _ _ _ _ Y P C O U G A K D V X R S H W B I N Q M F Z E J L T
This reconstruction is, so far, purely the result of hypothesis. Moreover, we do not as yet
know where the break in the numerical key comes, and this we shall proceed now to ascertain.
There are several methods. We might, for example, assume that the first segment to the right
of the plain segment bears the number 1 of the numerical key. Then, turning to the secondary
alphabet applying to that segment in table V III (the first column), we note what the plain-text
equivalents for the letters A, B, C, . . . Z, would be on this hypothesis by actual trial with this
alphabet mounted on the base disk, comparing the frequencies given with their normal ex-
pectancy. Thus, the letters D, M, and Y with frequencies of 4, 5, and 5, respectively would
represent the letters K, Q , and T, respectively; this would be far from a good agreement with
expectancy. Hence, we would conclude that the first segment of our base disk does not bear
the number 1. Let us assume that the second number of our cycle applies to the first segment
of our base disk, and proceed again to note the plain-text letters corresponding to this assump-
tion. The cipher letters with their frequencies and equivalents would be as follows:
F IGURE 18
Cipher..
Plain text.--
Frequency._
A B O D E F G H I J K
- G W P K Z M U S B E A
L M N O P Q R S T U V W X Y Z
J Q I C Y N X R L O D H V T F
The agreement of these frequencies with their normal expectancy is fair, although not
striking. There are too many occurrences of low-frequency letters, G, P, K, and Q , and although
the frequencies for E, 0, I , N, and R are excellent, there are not enough occurrences cf the high
frequency letters, T, A, and S." We must give this hypothesis further scrutiny. If the first
segment bears the number 8, then the second segment would bear the number 12. Again let
us match the frequencies given with then- corresponding plain-text letters on this hypothesis.
Cipher
Plain text..
Frequency
A B C D E F G H I
. U H Y A F Q O R W
19
J K L M N O P Q R S T U V W X Y Z
Z G E N B P T I V X J C K S D L M
This assortment of letters is also fair, but still the evidence is not conclusive. W e might
continue along these lines, attempting to establish definitely that the numerical key begins
with the number 8, or does not. But let us try another method.
Referring now to the Y sequence of frequencies in table V III still retaining the assumption
that the numerical key begins with the number 8, the Y sequence should be broken and re-
arranged in order to conform to the aesumed break in numerical key in this manner:
FIGURE 20
Y
8
1
12
1
16 20
4
2 22 16 8 11
2
5
2
18
2
IS 23 8
2
21
6
7 4 14
2
10 3
7
18
2
17 1
5
[Brea k]
48
Below these frequencies let us place the plain-text letters which they represent upon the
assumption that the break does occur between the numbers 1 and 8 in the key.
FlOCRE 21
Y
Plain text.
8
1
T
12
1
L
18
J
20
4
E
2
z
22
F
16
u
8
Q
11
2
N
6
2
I
10
2
B
13
W
23
H
9
2
S
21
6
R
7
X
4
V
14
2
D
10
K
3
7
A
18
2
G
17
u
1
5
0
[Break]
This too, looks like a fairly good assortment of high-frequency letters, with the single
exception of T. It may be that we have really struck the correct place for the break after all.
Let us corroborate it by still another method.
Assuming the break to be as shown in figure 20 let us make a list of the cipher letters which
should represent E in the successive segments. Thus:
F IGURE 22
8 12 18 JO 2 32
C
IS
0
8
U
11
G
8
A
10
K
18
D
28
V

X
31
R
7
S
4
H
14
W
10
B
3
I
18
N
17
Q
i
M
[Brea k]
Let us turn now to the respective frequencies of these cipher letters as given in table V II,
taking the frequencies of the letters J, L, T, . . . in the A columns, successively beginning
with segment 8.
F IQ URE 23
Equivalents of plain-text E_ _
Frequency
8
J
4
12
L
2
18
T
2
20
Y
1
4
2
P
2
22
C
4
IS
0
6
6
U
3
H
G
5
s
A
i
K
6
is
D
3
23
V
3
9
X
2
21
R
2
7
S
2
4
H
5
u
W
4
,0
B
6
3
I
5
18
N
4
17
Q
4
i
M
6
[Break)
On the whole, this is as good as can be expected for the small number of occurrences in
each table. Let us do the same for the letters T, 0, and A. If the results are as good as those
for E we may conclude that we have really found the absolute positions of the numbers in
the numerical key:
F IGURE 24
Equivalents of plain-text T _ _
Frequency
Equivalents of plain-text 0..
F requency..
Equivalents of plain-text A-.
Frequency
8
Y
1
U
3
K
0
12
P
2
G
5
D
2
16
C
1
A
2
V
4
20
0
6
K
5
X
6
2
U
5
D
5
R
2
22
G
4
V
2
S
4
u
A
2
X
3
H
1
6
K
2
R
7
W
2
11
D
4
S
2
B
3
6
V
6
H
4
I
2
IB
X
4
W
4
N
1
13
R
4
B
4
Q
6
23
S
4
I
5
M
3
V
H
5
N
3
F
2
21
W
1
Q
1
Z
5
7
B
7
U
2
E
2
4
I
2
F
2
J
3
14
N
2
Z
5
L
6
10
Q
2
E
2
T
4
8
M
2
J
1
Y
7
18
F
3
L
0
P
0
17
Z
G
T
0
C
3
l
E
1
Y
5
0
3
[Break]

49
These distributions and frequencies are certainly excellent and we may regard our numerical
key as established. The initial segment after the plain is number 8. (See Addendum, p. 56.)
We could proceed to reconstruct alphabets 2, 3, 4, and 5 by exactly the same principles as
were used in the reconstruction of alphabet 1. To do so would be purely of theoretical interest
because it would represent a case where the reconstruction of five primary alphabets, from
the frequencies of 115 unknown secondary alphabets, is accomplished without a preliminary
tentative decipherment of even so much as a single word. In short, it would represent a case
where the cryptanalyst, without attempting the decipherment of any part of the text, comes
at once, after such a reconstruction as the above, to be in the same position as the correspond-
ents, and can decipher any message as rapidly as the legitimate recipients.
Where a staff of clerics and experts is available, this method would indeed be followed,
for the personnel could be divided into five groups, each group being assigned an alphabet to
reconstruct. Each group could be subdivided into two sections, one working forward from a
given letter, the other working backward from the same letter. After not more than 3 to 6
hours all five alphabets will have been reconstructed in their entirety. Or perhaps it would
be more practicable to reconstruct only three of the primary alphabets, say the first, third, and
fifth, filling in the other two from the resulting decipherment.
In the present instance, however, only alphabets 1 and 3 were reconstructed, the recon-
struction of alphabet 3 being successfully accomplished by the application of the same principles
as were used for alphabet 1. Then a partial decipherment, in which the repetitions of digraphs
and trigraphs within adjacent columns played an important part, led to the reconstruction of
the other three primary alphabets.
The data for the reconstruction of alphabet 3 are given in table XIV. Since the location
of the break in the numerical key was determined after the reconstruction of alphabet 1, the
columns in table XIV , upon which the reconstruction of alphabet 3 is based, are given in the
correct numerical key order so that any two sequences of frequencies could be superimposed
at any intervals. However, one-interval and two-interval data were used almost exclusively
except in one or two doubtful instances where greater intervals were employed.
T ABLE XIII.Consolidated frequency table for alphabet S
Cipher
letter
1
A
B
C
D
E
F
a
H
I
j
K
L
M
N
O
P
Q
3.
S
T
tr
V
w
X
Y
Z
Segment
8
1
2
3
1
1
1
5
5
2
3
1
2
5
3
1
12
5
6
3
1
5
1
5
3
1
2
2
3
16
1
2
3
2
2
2
4
4
1
5
3
4
1
1
20
1
1
2
3
2
9
4
1
1
1
3
5
2
2
2
1
2
1
1
1
1
3
2
2
1
1
3
1
1
1
6
4
2
22
2
2
6
1
3
4
1
3
1
2
1
3
1
2
2
1
15
1
3
1
1
3
3
1
5
4
2
1
3
1
7
6
4
1
4
1
2
2
2
1
S
1
4
1
8
1
11
2
3
2
1
4
1
2
O
1
1
2
1
2
1
4
1
4
3
s
1
1
4
2
1
3
4
5
2
3
3
1
1
6
19
2
5
1
3
2
4
2
8
2
5
1
1
13
1
2
1
1
1
1
2
3
1
2
> 6
1
4
4
2
3
2
23
1
1
2
1
2
2
2
3
1
2
2
7
5
1
4
g
1
1
6
1
3
2
3
2
1
4
1
1
1
1
3
3
21
2
2
1
1
1
3
2
3
1
1
2
4
2
1
2
7
2
7
1
2
3
3
1
4
1
1
2
3
1
1
1
2
7
1
3
4
2
4
2
3
2
1
2
2
2
1
3
2
3
1
5
1
14
3
1
1
1
1
5
2
2
1
3
1
1
4
5
1
1
1
1
10
4
3
1
1
1
1
1
1
1
2
3
1
3
6
1
2
1
3
3
3
2
3
7
2
1
1
4
2
1
2
I
1
1
2
3
18
2
1
2
3
3
3
3
1
2
1
1
1
4
6
1
3
17
1
3
4
1
3
1
1
2
3
1
1
2
8
1
2
2
1
1
1
3
1
5
1
3
1
1
1
1
3
1
1
6
1
2
2
Total
fre-
quency
30
19
37
32
29
25
22
45
28
33
26
21
44
29
26
32
32
38
40
34
45
28
29
39
30
34
814
Number
of seg-
ments
occupied
17
11
15
14
13
13
14
14
13
16
14
9
16
15
12
16
14
15
16
15
15
11
13
16
12
12
Average
frequency
per seg-
ment
1.76
1.74
2.47
2.29
2.23
1.92
1.58
3.22
2. 15
2.07
1.86
2.34
2.75
1.94
2. 17
2.00
2.29
2.53
2.50
2.27
3.00
2.55
2.24
2.44
2.50
2.84
814
Average frequency per crpher letter=-r-r = 31.3 occurrences.
51
TABLE XIV.Bate f or r econ t lr u et i on of pr i m ar y alphabet n o. S
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
H(l)
A
B
C
D
E
F
0
I
a
K
L
II
N
0
P
Q
R
S
T
U
V
W
X
Y
Z
76
65
82
76
76
71
68
72
76
74
68
88
78
73
70
79
84
85
79
91
74
72
81
76
82
14
12
13
12
13
16
15
15
20
14
7
16
15
10
16
18
17
15
19
18
12
14
18
23
13
-34
-29
-43
-40
-36
-23
-23
-16
-16
-32
-47
-43
-28
-43
-31
-25
-33
-40
-22
-37
-38
-30
-27
- 7
-43
-.45
-.46
-.52
-.58
-.48
-.33
-.34
-.22
-.21
-.48
-.69
-.49
-.38
-.59
-.39
-.32
-.39
-.47
-.28
-.41
-.51
-.42
-.33
-.09
-.53
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
Y (2)
A
B
C
D
E
F
G
I
J
K
L
M
N
0
P
Q
R
S
T
U
V
W
X
Z
59
48
65
59
58
54
51
55
59
87
51
71
56
56
62
62
67
68
62
74
67
55
57
65
12
7
13
12
10
8
8
10
13
12
9
23
13
6
16
10
22
15
7
10
7
8
12
11
-23
-27
-26
-23
-28
-30
-27
-25
-20
-21
-24
- 2
-17
-41
-17
-32
- 1
-23
-41
-44
-36
-31
-21
-82
-.39
-.56
-.40
-.39
-.48
-.56
-.53
-.46
-.34
-.37
-.47
-.08
-.30
-.73
-.28
-.62
-.02
-.34
-.66
-.60
-.63
-.66
-.37
-.49
H at two intervals
M
R
88
81
29
20
1
-21
-.01
-.26
Letter
Total
occur-
rences
Coinci-
dences
Differ-,
ences
Indices
of coin-
cidence
M(3)
A
B
C
b
E
F
G
I
J
K
L
N
0
P
Q
R
S
T
U
V
W
X
Z
67
66
73
68
66
62
59
63
67
65
59
64
64
70
70
76
76
70
82
65
62
65
73
14
12
17
16
13
13
10
14
14
15
6
9
13
13
21
12
16
21
31
10
8
13
12
-25
-20
-22
-20
-27
-23
-29
-21
-25
-20
-41
-37
-25
-31
- 7
-39
-28
- 7
+ 11
-35
-38
-26
-37
-.37
-.36
-.30
-.29
-.41
-.37
-.49
-.33
-.37
-.31
-.70
-.58
-.39
-.44
-.10
-.52
-.37
-.10
+ .13
-.54
-.61
-.40
.51
"i(i.
52
TABLE XIV . -Data for reconstruction of primary alphabet no. S Continued
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
XT (4)
A
B
C
D
E
F
G
I
J
K
L
N
0
P
Q
R
S
T
V
W
X
z
73
63
79
73
72
68
65
69
73
71
65
70
70
76
76
81
82
76
71
69
78
79
17
17
13
14
11
14
13
11
14
14
7
10
10
14
17
13
20
16
11
9
8
10
-22
-12
-40
-31
-39
-26
-26
-36
-31
-29
44
-40
-40
-34
-25
-42
-22
-28
-38
-42
-54
-49
-.30
-.19
-.51
-.43
-.54
-.38
-.40
-.52
-.43
-.41
-.68
-.67
-.57
-.45
-.33
-.52
-.27
-.37
-.54
-.61
-.69
-.62
M at two intervals
A
B
S
65
54
74
17
16
20
-14
- 6
14
-.22
-.11
-.19
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
B(6)
A
C
0
E
F
G
I
J
K
L
N
0
P
Q
R
S
T
V
W
X
Z
47
53
47
46
42
39
,
47
45
39
44
46
50
50
55
56
50
45
43
52
53
8
9
6
7
7
6
11
12
10
2
7
7
12
8
13
11
7
7
.7
7
10
-23
-26
-32
-25
-21
-21
-10
-11
-15
-33
-23
-25
-14
-26
-16
-23
-29
-24
-22
-31
-23
-.49
-.49
-.68
-.54
-.50
-.54
-.23
-.23
-.33
-.85
"-.52
-.54
-.28
-.52
-.29
-.41
-.58
-.53
-.51
-.60
-.43
U at two intervals
I
J
K
F
R
69
73
71
76
81
21
16
14
16
18
- 6
-25
-29
-28
-27
-.09
-.34
-.41
-.37
-.38
Letter
Total
occur-
rences
Coinci-
dences
Differ"-
ences
Indices
of coin-
cidence
1(6)
A
C
D
E
F
G
J
K
L
N
0
P
Q
R
S
T
V
W
X
Z
49
65
49
48
44
41
49
47
41
46
46
52
52
57
58
52
47
45
54
55
1.1
8
15
10
7
8
8
11
6
13
7
12
7
13
11
7
6
12
8
7
-16
-31
A
-18
-23
-17
-25
-14
-23
- 7
-25
-16
-31
-18
-25
-31
-29
- 9
-30
-34
-.33
-.56
-.08
-.38
-.52
-.42
-.51
-.30
-.56
-.16
-.54
-.31
-.60
-.32
-.43
^.60
-.62
-.20
-.56
-.62
B at two intervale
D
K
N
W
47
45
44
43
8
7
10
8
-23
-24
-14
-19
-.49
-.53
-.32
-.44
IT at three intervals
D
K
N
W
73
71
70
69
13
13
16
13
-34
-32
-22
-30
-.47
-.45
-.32
-.44
53
TABIJB XIV .Data for nconttruction of primary alphabet no. 3Continued
Letter
Total
ooour-
nnoM
Coinci-
dences
Differ-
enoes
Indices
o( coin-
cidence
N( 7 )
A
C
D
E
F
G
J
K
L
0
P
Q
R
S
T
V
W
X
z
69
66
60
68
64
61
69
67
61
66
62
62
67
68
62
57
55
64
65
13
22
8
9
13
7
16
11
12
12
16
12
17
11
12
10
7
14
14
-20
+ 1
-35
-31
-16
-30
-11
-24
-16
-20
-14
-26
-16
-35
-28
-27
-34
-22
-23
-.34
+ .02
-.69
-.63
-.28
-.59
-.19
-.42
-.30
-.36
-.23
-.42
-.24
-.52
-.42
-.47
-.62
-.34
-.35
W (10)
A
D
E
F
G
J
K
L
0
P
Q
R
S
V
X
z
58
58
57
53
50
58
56
50
55
61
61
66
67
56
63
64
9
6
13
8
6
15
12
14
6
13
9
22
8
6
17
11
-31
-40
-18
-29
-32
-13
-20
- 8
-37
-22
-34
0
-43
-38
-12
-31
-.54
-.69
-.32
-.55
-.64
-.22
-.36
-.16
-.67
-:36
-.56
0
-.64
-.68
-.19
-.48
Letter
Total
occur-
rences
Coinci-
dences
Differ-
enow
Indices
of coin-
cidence
C (8)
A
D
E
F
G
J
K
L
0
P
Q
R
S
T
V
W
X
Z
63
63
62
68
55
68
61
55
60
66
66
71
72
66
61
59
68
69
17
11
10
10
17
14
17
10
9
17
19
12
15
21
7
9
12
15
-12
-30
-32
-28
- 4
-26
-10
-25
-33
-15
- 9
-36
-27
- 3
-40
-32
-32
-24
-.19
-.48
-.52
-.48
-.07
-.38
-.16
-.46
-.85
-.23
-.14
-.49
-.38
-.05
-.66
-.54
-.47
-.35
B (11)
A
D
E
F
G
J
K
L
0
P
Q
S
V
X
z
64
64
63
59
56
64
65
56
61
67
67
73
62
69
70
14
15
12
12
12
15
12
10
13
17
13
13
13
22
12
-24
-19
-27
-23
-20
-19
-29
-26
-22
-16
-28
-34
-23
- 3
-34
-.38
-.30
-.43
-.39
-.36
-.30
-.45
-.47
-.36
-.24
-.42
-.47
-.37
-.04
-.49
Letter
Total
ooour-
renoM
Coinci-
dences
Dlfler-
enoM
Indices
of coin-
cidence
T (9)
A
D
E
F
G
J
K
L
0
P
Q
R
S
V
W
X
z
66
65
64
60
57
65
63
67
\
62
68
68
73
74
63
61
70
71
11
11
15
14
10
16
13
14
10
16
15
21
17
5
24
15
10
-32
-32
-19
-18
-27
-17
-24
-15
-32
-20
-23
-10
-23
-48
+ 11
-25
-41
-.49
-.49
-.30
-.30
-.47
-.26
-.38
-.26
-.52
-.30
-.34
-.14
-.31
-.76
+ .18
-.36
-.58
X (12)
A
D
E
F
G
J
K
L
0
P
Q
S
V
Z
64
64
63
59
56
64
65
56
61
67
67
73
62
70
17
16
16
9
9
13
12
18
7
13
13
10
8
16
-13
-16
-15
-26
-29
-25
-29
- 2
-40
-28
-28
-43
-38
-'22
-.20
-.25
-.24
-.44
-.52
-.39
-.45
-.04
-.66
-.42
-.42
-.59
-.62
-.31
TABLE XIV .Data for reconstruction of primary alphabet no. SContinued
Utter
ToteJ
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
I. (13)
A
D
E
F
G
J
K
0
P
Q
S
V
Z
50
60
49
45
42
SO
48
47
53
53
59
48
56
13
6
7
8
6
10
8
5
12
11
8
4
8
-11
-32
-28
-21
-24
-20
-24
-32
-17
-20
-85
-36
-82
-.21
-.64
-.57
-.47
-.64
-.40
-.50
-.68
-.32
-.38
-.59
-.75
-.57
Z (16)
E
F
G
J
K
0
P
Q
S
V
56
52
49
67
55
54
60
60
66
55
6
11
15
12
10
17
13
10
14
5
-28
-19
- 4
-21
-25
- 3
-21
-30
-24
-40
-.50
-.37
-.08
-.37
-.46
-.06
-.35
-.50
-.36
-.73
D at two intervals
G
0
53
58
17
11
2
-25
-.04
-.43
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
A (14)
D
E
F
G
J
K
0
P
Q
S
V
Z
57
56
52
49
57
55
64
60
60
66
55
68
15
9
14
11
14
12
10
15
13
14
16
15
-12
-29
-10
-16
-15
-19
-24
-15
-21
-24
- 7
-18
-.21
-.52
-.19
-.33
-.25
-.85
-.46
-.25
-.86
-.37
-.13
-.29
L at two intervals
D
F
J
K
P
V
50
45
47
46
48
48
13
6
7
7
10
9
-11
-27
-26
-25
-18
-21
-.22
-.60
-.55
-.54
-.38
-.44
0 (17)
E
F
J
K
0
P
Q
S
V
50
46
51
49
48
54
54
60
49
10
15
14
8
9
11
9
10
7
-20
- 1
- 9
-25
-21
-21
-27
-30
-28
-.40
-.02
-.18
-.51
-.44
-.39
-.50
-.50
-.57
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
D (16)
E
F
G
J
K
0
P
Q
S
V
Z
56
52
49
67
55
54
60
60
66
65
63
10
12
9
16
13
8
11
16
10
14
24
-26
-16
-22
- 9
-16
-30
-27
-16
-86
-12
+
-.47
-.31
-.45
-.16
-.29
-.66
-.45
-.25
-.65
-.22
+.14
F(18)
E
J
K
0
P
Q
S
V
52
53
51
50
56
56
62
51
14
9
15
6
12
8
14
11
-10
-26
- 6
-32
-20
-32
-20
-18
-.19
-.49
-.12
-.64
-.36
-.67
-.32
-.35
Z at throe intervals
E
K
P
S
50
57
62
68
12
17
10
23
-14
- 6
-32
+ 1
oc
.AO
-.10
-.52
+ .05
55
TABLE XIV.Data for reconstruction of primary alphabet no. SContinued
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
S(19)
E
J
K
0
P
Q
V
67
68
66
65
71
71
66
14
13
14
14
12
11
16
-25
-29
-24
-23
-35
-38
-21
-.37
-.43
-.37
-.35
-.49
-.54
-.82
0 (22)
E
J
P
Q
64
55
58
58
11
7
9
12
-21
-34
-31
-22
-.39
-.62
-.53
-.38
F at five intervals
E
Q .<
46
49
13
4
_ L. tj
-37
-.15
-.75
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
V (20)
E
J
K
0
P
Q
51
52
60
49
55
55
7
12
15
5
10
10
-30
-16
- 6
-34
-25
-25
-.59
-.31
-.10
-.69
-.46
-.46
15(23)
J
P
Q
67
60
60
18
13
9
- 3
-21
-33
-.05
-.35
-.55
Letter
Total
occur-
rences
Coinci-
dences
Differ-
ences
Indices
of coin-
cidence
X(21)
E
J
0
P
Q
55
56
63
59
59
10
14
16
14
13
-25
-14
- 6
-17
-20
-.46
-.25
-.09
-.29
-.34
J (24)
P
Q
66
65
24
18
+ 17
-16
+.31
-.29
P (25)
Q
(26)
63 22 + 3 + .05
The completely reconstructed alphabet 3 is as follows:
Alphabet 3 R X L A D Z Q F S V K O E J P Q H Y M U B I N C T W
W ith alphabets 1 and 3 at hand, the partial decipherment of the initial groups of a few
messages soon leads to the complete reconstruction of the other three alphabets. For example
note what these two alphabets give for the beginning of message 11 :
Segments 11 5
Alphabets 1 2 3 4 5 1 2 3 4 5
Cipher S E Y B Z M G S O Z
Plaintext O S V T
12 34 2 12 8 4 6 1
The word OBSERV ATION comes to mind almost at once.
19
1 2 3 4 5
C M P S Q
N
This means that we have
determined the following values:
Alphabet 8
Segment 11, B
P
=E
C
Segment 5, A
P
==G
C
Alphabet 4
Segment 11, E
P
=B
0
Segment 5, I
p
=0.
Alphabet 5
Segment 11, R
P
=Z
C
Segment 5, O
p
=Z
e
56
New values thus determined are inserted in their correct positions with the result that
after a short time alphabets 2, 4, and 5 are completely reconstructed. The five alphabets
are found to be as follows:
Al pha be t 1 Y P C O U G A K D V X R S H W B I N Q M F Z E J L T
Al pha be t 2 E Q M S L P Y V C J T A K H U Z F B R X I G N W D O
Al pha be t 3 R X L A D Z Q F S V K O E J P Q H Y M U B I N C T W
Al pha be t 4 A T E V X Z Y N F G S B I N Q R P C M D . K W O J U L
Alpha bet 5 A E F B N K L T I X M V O R Y W H Q J U G C Z P D S
Proof of their correctness may be at once established by deciphering the first few groups
of message 1. They are as follows:
MLVXK QNXVD GIRIE IMNEE FEXV P HPVZR UKSER MV Q CI etc.
EV ENI NGREP ORTSS HOULD INCLU DEDAI LYLOS SESSE etc.
"Evening reports should include daily losses . . . "
Once the primary alphabets have been reconstructed, the solution of subsequent messages,
written by means of them but employing a wholly different numerical key, is very easy; for
the cryptanalyst has only to set the first 5 letters of any message in one segment and look for
the plain-text beginning of the message in some other segment. Once the starting point has
been found, the number corresponding to the serial number of the message can be inserted in
its proper position, and all other numbers of the key determined in a similar manner.
This cipher is of interest also in view of its striking similarity to the Bakeries Disk Cipher,
or the "Star Cipher", described in a previous paper
!
by the author.
In the latter cipher, instead of only 5 primary alphabets, there are 20 to 25, and instead
of having a numerical key to determine where the cipher equivalents are to be taken, no key
of that nature is used, or needed, since the correct line, or generatrix, is easily found because
it is the only one that gives intelligible text throughout its length. The segments in the V ogel
Cipher correspond to the generatrices in the Star Cipher. In the former, the cipher letters are
taken from a definite cycle of generatrices, and not all generatrices are used; in the latter, no
such cycle obtains, for it is unnecessary, and all 25 generatrices may be used at random.
The fatal defect in this cipher, from the point of view of its practical indecipherability, is
that the segments, or generatrices, are used in a definite sequence or cycle, giving rise to an
internal period within messages and an external period between successive messages. It was
through them that the recovery of the numerical key and the primary alphabets was possible.
AD D END UM
The method presented above for determining the index of coincidence will in most cases lead
to the correct fitting of two frequency distributions. Since writing the foregoing description
of the method for determining the index of coincidence, the author has applied another method
which, although it involves additional calculations, will lead to even more accurate results.
Since cases wlu'ch may necessitate a greater refinement in method than that elucidated above will
arise, it may be worth while to present the more detailed method.
This method more closely approximates the actual methods used by statisticians and biome-
tricians in their analysis of data, in that it involves the squares of deviations or differences
between corresponding observed values.
1
Several Machine Ciphers and Methods for their Solution. Riverbank Publication No. 20. (1918.)
57
Let us consider the Y sequence of frequencies of table VIII. We desire to find that distribu-
tion which, when shifted one interval to the left of the Y distribution, will most closely approxi-
mate the Y distribution; i.e., comes from the same "parent" distribution. Returning again to
figures 13 and 14, we have the following:
FIGURE 25
A
Y
2
5
1
1
2
1
2 1
4
1 2 1
2 2
1
2
2
2
2
6
1
2
2 2
7 2
Pearson ' gives, in his discussion concerning "Testing Goodness of Fit", the method for
determining the probability that random sampling would lead to as large or larger deviations
between observations than those observed.
We shall here summarize the paper referred to. Let the samples be given by the frequencies
in the same class as follows:
Total
First sample: / /, / . . . /., N
Second sample: /', /a, /i, . . . . / N '
where the totals N and N ' differ as little or as widely as we please. The value x
2
given
by x
2
xnny S / /- ~ is calculated. Tables
2
have been calculated which give, for various
l

1 J i 'J i
values of n' , the probability of obtaining a value of x
z
&s big or bigger than the one calculated.
In matching alphabets we must be careful to choose the proper value of n' in the tables. It
is known that not all 26 letters will appear in a monoalphabet until the total number of letters is
about 500. The number of different letters corresponding to various total letters per monoalphabet
can be determined, and a curve plotted that will enable one to obtain this value readily.
2
The
procedure in applying the method for matching is as follows:
1. The value of x
2
is calculated as above.
2. The number of different letters, n' , corresponding to a total of N +N ' letters is
obtained from the curve.
3. In the tables for x
2
the column n' gives the probabib'ty corresponding to the calcu-
lated value of x
2
.
This probability expresses the chance of getting another set to yield a value of x
2
at least
as large as the one obtained; in other words, the closeness with which the two alphabets match.
Let us proceed to find the value of P for the two distributions shown in figures 13 and 14.
W e first find the differences between the weighted frequencies in the respective superimposed
sequences and then square the differences. Thus:
1
Pearson, Karl. On the probability thai two independent distributions of frequency are really samples from
the same population. Biometrika, vol. 8 (1911), pp. 250ff.
> See, S. Kullback, loc. cit., Table IV and chart 1. The table and chart are reproduced on pages 61-64 of
this paper.
FIOUBB 26
Total
A
T
W eighted fre-
quencies
Differences
Squan
ence
of diff-
& .
2
5
72
110
38
1,444
1
1
36
22
14
196
2
1
72
22
50
2,500
2
72
72
5, 184
1
4
36
88
52
2,704
1
36
36
1,296
2
72
72
5,184
1
36
36
1,296
2
44
44
1,936
2
44
44
1,936
1
2
36
44
8
64
2
2
72
44
28
784
2
6
72
132
60
3,600
1
36
36
1,296
2
44
44
1,936
2
72
72
5,184
2
7
72
154
82
6,724
2
44
44
1,936
22
36
59
Next, we must divide each difference squared by the sum of the corresponding original
frequencies. Thus:
TABLB XV
Square of differ-
ence
1,444
196
2,500
6,184
2,704
1,296
5,184
1,296
0
1,936
1,936
Sum of original
frequencies
7
2
3
2
5
1
2
1
0
2
2
Q uotient
206.3
98. 0
833.8
2, 592. 0
540.8
1, 296. 0
2, 592. 0
1, 296. 0
0.0
968.0
968.0
Square of differ-
ence
64
0
0
784
3,600
0
1,296
1,936
5, 184
6,724
1,936
Sum of original
frequencies
3
0
0
4
8
0
1
2
2
9
2
Q uotient
21.3
0.0
0.0
196.0
450.0
0.0
1, 296. 0
968.0
2, 592. 0
747. 1
968.0
To find x* we must now add these quotients and divide the sum by N XN ' or 22X36.
The sum of the quotients, 18,628.8, when divided by 792 gives x
s
=23.5. Jn this particular case
the further refinement was not applied. Since there are 22 segments we look up the table for
x*=23.5 and n'=22 and find P=.32 which means that in 32 of 100 cases we would get by chance
a match as bad as or worse than that observed.
We would ordinarily then go through exactly the same steps for all the other sequences in
our table VIII against the Y sequence; the sequence which gives the greatest value for P would
be the correct one and would indicate which letter follows Y in alphabet 1. But it will be
unnecessary here to go through all the calculations since it is desired only to show that this
method will give more trustworthy results than the method of coincidences and noncoincidences
presented above, and we shall, therefore, show only a few cases.
Let us take only those five sequences in table V III which when tested against the Y se-
quence by the previous method gave the greatest indices of coincidence, viz, the P, Q, N, L, and
J sequences which gave indices of coincidence of .12, .23, .24, .28, and .29, respec-
tively. The data for these sequences upon the method of squares are as follows:
TABLE XV I
Total
P
Q
N
X.
J
Y
Y
Y
Y
Y
1
5
4
5
3
5
5
4
5
2
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
1
2
4
4
4
1
4
1
4
4
3
3
2
4
3
1
3
1
3
1
1
2
1
2
2
2
3
2
3
2
2
1
2
1
2
2
2
2
6
2
2
2
3
2
1
2
3
2
3
1
2
1
2
1
2
1
2
4
2
6
6
1
6
1
G
2
c
6
3
1
2
5
2
2
2
2
1
2
2
2
1
7
7
4
7
6
7
1
7
2
4
2
1
2
1
2
1
2
22
36
29
36
27
36
33
36
27
36
vl c^l 1 8 P=7R
X
J.-.OO o p oo
t /1 91 *7 P A * \
60
The values of P (index of probability or coincidence) show .that the probability that the
P sequence belongs one interval to the left of the Y sequence is a little more than 1.7 times as great
as that of the nearest rival to the P sequence, viz, the N sequence. The result is quite gratifying.
W e shall show another case. For this we shall choose a case in which a tertiary trial had to
be made on the previous methodthat is, the method in wliich a trial with sequences only one
interval and two intervals removed failed to give definite and clear-cut results necessitating
a third test with a sequence throe intervals removed. Note the difficulty in finding the letter
which follows K in table V III. Here even the second trial failed to show whether D, H, Q, or
Z follows K. Let us test these sequences against the K sequence using the method of squares.
The data are as follows:
T ABLE X V II
Tots)
D
H
Q
Z
X
K
K
K
2
1
4
2
2
1
1
1
1
3
3
3
4
3
5
5
8
5
4
5
2
5
1
1
1
3
1
1
1
2
1
1
1
2
1
1
i 4
2
1
2
2
2
2
3
1
4
1
1
1
2
1
1
3
6
1
6
6
6
6
4
5
4
4
4
1
2
1
2
1
2
5
2
1
2
4
5
4
4
2
4
2
5
5
1
5
5
5
1
1
2
1
2
1
1
4
1
1
1
1 4
3
2
3
4
3
6
3
35
39
32
39
29
89
32
39
vt 1O A . f > i \<\
1/|__O1 1 f >
t
A * )
vf 91 7 P 4*>
vl=1 7 7 P fll
The difference between D and Z is not conclusive enough to decide definitely yet, but H and Q
can be eliminated at once. Testing for interval 2 will show that D is the proper letter.
It is to be noted that for our purposes it is not even necessary to find the actual value of P;
for given n' , constant in value, P increases as x
2
decreases, not in a regular manner it is true but
approximately so. Therefore, if n' is constant, throughout a series of calculations, the index of
coincidence in our case will vary in a general manner inversely with the value of x
2
> the lowest
value of x
2
indicating the greatest degree of probability or the greatest index of coincidence.
It is believed that this method of squares will be found exceedingly valuable in many other
cases where considerable refinement of method is necessary to produce clear-cut results in fitting
frequency distributions to one another.
61
TABLE IV.
1
T eat for Goodness of F it. Values of P.
X *
1
g
s
4
6
6
7
S
9
10
11
IS
IS
14
16
16
17
18
19
SO
SI
SS
3
S4
5
s e
S7
S3
2 9
SO
4 0
50
C O
70
'=3
60G331
367879
223130
135335
082036
049787
030197
018316
011109
OOG738
004087
002479
001503
000912
000553
000335
000203
000123
000075
000045
000028
000017
000010
000006
000004
000002
000001
000001
000001
oooooo
oooooo
oooooo
oooooo
oooooo
n'-4
801253
672407'
381625
261464
171797
111610
071897
046012
"029291
018566
011726
007383
004637
002905
001817
001134
000707
000440
000273
000170
000105
OOOOG5
000040
000025
000016
O00010
000006
O00004
000002
000001
OOOOOO
oooooo
oooooo
oooooo
n'5
909796
735759
657825
406006
287J9S
199148
135888
091578
061099
040428
026564
017351
011276
007295
004701
003019
001933
001234
000786
000499
000317
000200
000127
000080
000050
000032
000020
000012
000008
000005
oooooo
oooooo
oooooo
oooooo
'~6
062566
849U6
699986
649416
415880
306219
220640
156236
109064
075235
051380
O34787
023379
015609
010363
006844
004500
O02947
O01922
001250
000810
000524
000338
000217
000139
000090
000057
000037
000023
000016
oooooo
oooooo
oooooo
oooooo
n' = 7
985612
819699
608847
C76676
643813
423190
320847
238103
173578
124652
083376
O61969
043036
029636
020256
O13754
009283
006232
004164
002769
001835
001211
000796
000522
000341
O00223
000145
O00094
O00061
O00039
O00001
OOOOOO
OOOOOO
OOOOOO
n'=8
994829
059840
886002
779778
659963
639750
^428880
332594
252656
188673
138619
100568
O72109
061181
O36000
O25116
O17390
O11970
O08187
O05570
003770
O02541
O01705
O01139
000759
O00504
O00333
O00220
O00145
000095
O00001
oooooo
OOOOOO
OOOOOO
n'=9
998249
981018
934357
857123
757676
647232
636632
433470
342296
265026
201699
151204
111850
OS 1765
O59145
O42380
O30109
O21226
O14860
O10336
O07147
O04916
O03364
O02292
001554
O01050
O00707
O00474
O00317
O00211
O00003
OOOOOO
OOOOOO
OOOOOO
'=10
999438
991468
964296
911413
834308
739919
637119
634146
437274
360486
276709
213308
162607
122326
O90937
066881
O48716
O35174
O25193
O17913
O12650
O03880
O06197
O04301
O02971
002013
O01399
O00954
O00648
000439
O00008
OOOOOO
OOOOOO
OOOOOO
n'-ll
999828
096340
081484
947347
881178
815263
726444
628837
632104
440493
367618
286067
223672
172992
132061
O99632
074364
O54964
O40263
O29253
021093
015105
010747
O07600
005345
O03740
O02604
O01806
O01246
000867
O00017
OOOOOO
OOOOOO
OOOOOO
1
Copied from T ablei for StaliitMani ant Blomilriciatu, Edited by Karl Pearson, Part I, 2nd Ed., Cambridge
University.
62
TABLE IV (continued).
*
1
*
s
4
6
6
7
S
9
10
11
It
IS
14
IB
16
17
IS
19
SO
SI
S3
t s
* *
s s
S6
r r
S8
9
s o
4 0
50
60
70
n'=13
909950
993496
990726
969917
031167
873365
799073
713304
621892
630387
443S63
362642
293326
232993
-182498
141130
107876
031681
061094
O40341
033371
024374
017676
012733
009117
006490
004595
003238
002270
001685
000036
000001
000000
000000
n'=13
999986
999406
995644
983436
957979
916082
857613
785131
702931
616960
628919
445680
309041
300708
241436
191236
149597
116691
088629
067086
050380
037520
027726
020341
014822
010734
007727
005532
003940
002792
000072
000001
000000
000000
n'=.14
999997
999774
997034
991191
975193
946153
902151
843601
772943
693934
610817
527643
447812
373844
307354
249129
199304
157620
123104
095210
072929
055362
041677
031130
023084
017001
012441
009050
006546
004710
000138
000003
000000
000000
n'=15
999999
999917
999074
995466
985813
966491
934711
889327
831051
762183
666036
606303
626624
449711
378154
313374
256178
206781
164949
130141
101632
078614
060270
045822
034506
025887
019254
014228
010450
007632
000255
OOOOOG
oooooo
oooooo
n'=16
1-
999970
999698
997737
992127
979749
957650
923783
877617
819739
752594
679028
602298
625629
451418
382051
318864
262666
213734
171932
136830
107804
084140
OG5093
O49943
038023
028736
021569
016085
011921
000453
000012
oooooo
oooooo
n'=17
1-
999990
999830
998903
995754
988095
973260
948867
913414
866628
809486
743980
672768
698714
624638
452901
385597
323897
268663
220220
178510
143191
113735
089504
069824
054028
041483
031620
023936
018002
000778
000023
O00001
oooooo
n'=I8
V
999997
999931
999483
997771
993187
983549
966547
940261
903610
866564
800136
736186
667102
695482
623834
454366
388841
328532
274229
220291
184719
149251
119435
094710
074461
0580G8
O44938
034526
020345
001294
000042
000001
oooooo
'=19
I-
999999
999972
999763
998860
996197
990125
978637
959743
931906
894357
847237
791673
729091
661967
692547
523105
455053
391823
332819
279413
231985
190590
155028
124915
099758
078995
062055
048379
037446
002087
000075
000002
oooooo
'=20
1-
1-
999989
999894
999431
997929
994213
986671
973479
952946
923839
885624
838571
783691
722598
657277
589868
622438
450836
394578
336801
284256
237342
196152
160542
130189
104653
083428
005985
051798
003272
000131
000004
OOOOOO
63
TABLE IV (continued)
r
X'
1
g
3
4
6
6
7
8
9
JO
11
It
IS
14
15
16
17
18
19
s o
SI
M
t s
S4
t s
t 6
S7
S8
S9
SO
4 0
60
60
70
'=21
I-
1-
899996
999954
999722
998898
996686
991868
982907
868171
646223
916076
877384
830496
776408
716624
652974
687408
621626
457930
397132
340611
288795
242392
201431
166812
136264
109399
087769
069854
004996
000221
000007
000000
'-22
1-
1-
999998
999980
999868
999427
998142
996143
989214
978912
962787
939617
908624
869599
822952
769650
711106
649004
685140
621261
458944
399510
343979
293058
247164
206449
170853
140161
114002
091988
007437
000365
000013
000000
'=2S
!
!
999999
999992
999939
999708
998980
997160
993331
986304
974749
967379
933161
901479
862238
815886
763362
705988
645328
583040
620738
469889
401730
347229
297076
261682
211226
175681
144861
118464
010812
000686
000022
000001
'=24
I-
!
1*
099997
999972
999865
999452
998371
995957
991277
983189
970470
951990
926871
894634
655268
809251
757489
701224
641912
681087
520258
460771
403808
360286
300866
255967
216781
180310
149402
016369
O00921
000038
000001
n'=25
!
I-
1-
999999
999987
999929
999711
999086
997696
994547
989012
979908
966121
846650
920769
888076
848662
803008
761990
696776
638726
679267
519798.
461597
406760
353166
304463
260040
220131
184762
021387
001416
000064
000002
'=26
1-
!
1-
!
999994
999966
999861
999494
998596
996653
992946
986567
976601
901732
941383
914828
881793
842390
797120
746825
692609
635744
677664
519373
462373
407598
355884
307853
203916
224289
O29164
002131
000104
000004
n'=27
!
1-
!
!
999998
999984
999924
999726
999194
997981
995549
991173
983974
973000
957334
936203
909083
875773
836430
791656
741964
688697
632947
676965
618976
463105
409333
358458
311082
207611
039012
003144
000168
000007
'=28
1-
!
1-
1-
999999
999993
999962
999853
999646
998803
997239
994294
989247
981264
969438
952947
931122
903519
870001
830756
786288
737377
685013
630316
674462
618600
463794
410973
360899
314164
051237
004551
000264
000011
n'=29
I-
I-
1-
!
1-
909997
999981
999924
999748
999302
898316
996372
992900
987189
978436
965819
948589
926149
898136
864464
825349
781291
733041
681635
627836
673045
618247
464447
412628
363218
066128
006467
000407
000019
n'=80
!
!
1-
!
1-
999999
999991
999960
999863
999699
898988
997728
895384
891377
985016
976536
962181
944272
921288
892927
869149
820189
776543
728932
678248
625491
571705
617913
465066
414004
063937
009032
000618
000030
64
MONOALPHABETIC DISTRIBUTION PROBABILITY CHART 1
f O id Jo"so
NUMBER OF LETTERS PtR ALPHABET (OR ALPHABETS)
PART II
THE SCHNEIDER CIPHER
1
DETAILS OF ENCIPHEHMENT
This system aims to make the entire operation of encipherment and decipherment depend*
ent upon the knowledge of a single key word agreed upon in advance by the correspondents.
Let us suppose the word t o b e T R E B I Z O N D .
An arbitrarily mixed alphabet is derived from a generating rectangle based upon this
key word, with a slight departure from the usual method. Instead of inscribing the letters in
the generating rectangle in the key-word sequence, as is usual, the initial letter of the second
line may be any letter except one in the key word itself. The remaining letters then follow in
the key-word sequence. Thus suppose we choose K as this initial letter for the second line.
Our rectangle is constructed in this manner:
FIQ UBB 27
T R E B I Z 0 N D
K L M P Q S U V W
X Y A C F G H J
By taking the columns in succession and writing them in two lines of 13 letters each we
have the following:
Alphabet 1
FIQ URB 28
T K X R L Y E M A B P C I
Q F Z S G O U H N V J D W
These two lines constitute alphabet 1, in which T=Q , K=F , X=Z , etc. The values
are all reciprocal.
All the other alphabets are derived from alphabet 1 by the simple expedient of carrying
the upper half of alphabet 1 to the third line and revolving the sequence one letter to the right.
Thus:
F IQ URB 29
Alphabet 1
(1) T K X R L Y E M A B P C I
( 2) Q F Z S G O U H N V J D W 1
A
i
phabet 2
(3) I T K X R L Y E M A B P C J
The juxtaposition of lines 2 and 3 results in the formation of alphabet 2, in which Q =I,
F=T, Z =K, etc., also completely reciprocal.
1
Schneider, L. Description d' un sys&me cryplographique d. I ' usage de Varmte. Paris, 1912. This cipher
was submitted for examination by Maj. Otto Holstein, M.I.-Res. who furnished the writer, who was then
at the Riverbank Laboratories, with a translation of Commandant Schneider's paper. Attempts to locate
in American libraries and bookstores the work cited to see if the inventor had offered a sample message for
solution proved of no avail; fortunately, a copy of the paper was sent in quite accidentally by an obscure book
dealer shortly after the first draft of this manuscript was prepared. No sample message is given. Upon my
request an assistant prepared the message which will presently be given, and the solution of which waa attained
by the principles to be elucidated.
(65)
66
Alphabet 3 is constructed by moving line 2 to line 4, revolving the sequence two letters
to the right. Thus:
FIGURE 30
Alphabet 1
Alphabet 3
(1) T K X R L Y E M A B P C I
(2) Q F Z S G O U H N V J D W
(3) I T K X R L Y E M A B P C
(4) D W Q F Z S G O U H N V J
Alphabet 2
The juxtaposition of lines 3 and 4 results in the formation of alphabet 3, in which I=D,
T=W , K=Q , etc.
Continuation of this process can result in the production of a total of 13 different secondary
reciprocal alphabets. As a rule only a limited number of these secondary alphabets are
employed, usually not to exceed the first 6 or 7.
One of the main features of the method of encipherment is that groups of unequal lengths
are enciphered in cyclic fashion by means of several alphabets. That is, the text is broken up
into groups containing 1, 2, 3, . . . letters enciphered by different alphabets, the groups
being repeated in sets, as explained below.
Let us, as one of the correspondents, determine that the cycle is to consist of six groups.
Taking the first 6 letters of line 1, that is the upper half of alphabet 1, their numerical values on
the basis of their relative positions in the normal or straight alphabet, are as follows:
T K X R L Y
4 1 5 3 2 6
This sequence of numbers, 4-1-5-3-2-6, constitutes a cycle designated hereafter as the
"Interruption key." It partakes of the nature of an "interrupter" in that it dictates the
number of letters in each of the groups of irregular length treated in enciphennent. Thus, the
first group contains 4 letters; the second, 1 letter; the third, 5 letters, etc. The seventh group
would begin a repetition of the cycle, and it would contain 4 letters; the eighth group, 1 letter,
etc, This cycle is repeated many times within a message.
Encipherment within each group is regular in the succession of the alphabets employed.
Thus, if three alphabets are determined upon by the correspondents, the alphabets would be
employed in the following sequence in the interruption key given above:
F IGURE 31
Length of group
Alphabets
4
2 3 1 1 2 3 1 2
3
1 2 3 1 2
6
1 2 3 1 2 3
Note that in groups containing more letters than the number of alphabets decided upon,
the sequence of alphabets is repeated. Thus, in the 5-letter group, we have the sequence of
alphabets 1-2-3-1-2; in the 6-letter group, 1-2-3-1-2-3. Enciphennent then proceeds by
alphabets according to the distribution of numbers within the groups. For example, group 1
containing the sequence of numbers 1-2-3-1, the first letter is enciphered by means of alphabet
1; the second, by alphabet 2, and the third, by alphabet 3. The fourth letter, however, is again
enciphered by alphabet 1.
After encipherment, the order of the cipher letters within the groups is reversed throughout
the message. Thus, suppose that the encipherment using the three alphabets and key above
were as follows:
li
67
FIGURE 32
4
i 2 a i
E N EM
UM 0 H
1
i
Y
0
5
1 2 3
I SI
W X D
1 2
N T
A F
3
i 2 3
REN
SHE
2
1 2
C H
D E
1 2
I N
W M
6
3 1 2 3
G A L 0
Y N 0 E
4
1 2 3 1
NG WE
A R T U
1
i
S
R
5
1 2 3 1 2
TERNS
Q H Z A X
3
1 2 3
LOP
G L V
The letters within the cipher groups are then reversed yielding the following:
FIGURE 33
HOM U 0 F A D X W BHS ED E O N Y M W UT RA R X A ZHQ VLG
It is now necessary that the information necessary to decipher the message be conveyed
to the recipient, who, of course, is already in possession of the key word. This information is
transmitted in the form of an "Indicator group ", consisting of 3 letters, whose location within the
message has been previously determined in a manner to be explained presently. The first letter
of the indicator group gives the initial letter of the second line of the generating rectangle; the
second, the number of alphabets employed; and the third, the length of the interruption key,
from which its sequence can be derived from the upper half of alphabet 1, as already explained.
In the case above, the initial letter is K. The number of alphabets being three, the third
letter in the upper half of alphabet 1, viz, X, forms the second letter of the indicator group.
The length of the interruption key being six, the sixth letter of the upper half of alphabet 1,
viz, Y, forms the third letter of the indicator group. The knowledge of this letter enables the
recipient to construct the interruption key. The indicator group for the message above is,
therefore, KXY.
This indicator group is then inserted in the cipher text in a position determined by a pre-
viously agreed upon letter of the key word, usually either the first or the last letter. Thus, if the
correspondents agree upon the first letter of the key word, this indicator group would be inserted
after the twentieth letter of the cipher text, because T, the first letter of the key word, occupies
the twentieth position in the normal or straight alphabet. The cipher message above would
read, therefore, as follows:
MESSAGE
HOMUO FADXW BHSED EONYM KXYW U TRARX AZHQV etc.
By varying the elements of the indicator group, a great number of combinations can be
obtained from one key word. Schneider says:
W ith a key of 6 letters, if one sets out to modify the table of alphabets, one obtains only 26 6=20
different cryptograms from the same text.
If, in addition to the table of alphabets, one modifies the number of alphabets employed and the number of
letters to form the numerical key, assuming that one employs 3, 4, 5, or 6 alphabets (a total of 4 combinations)
and 5, 6, 7, or 8 letters to form the numerical key (also a total of 4 combinations), one would obtain 20 X 4 X 4=320
different cryptograms from one and the same text
W e shall first elucidate the principles by means of which a single message may be solved,
and then proceed to the solution of a series of messages in the same key.
68
1. Solution of a single message.Given the following message:
KHNVL
KZRGM
ESPEB
EMENF
PAECL
XNKGD
AQAHM
KGIQQ
KIEBC
OEWJO
WWCLR
KFSVB
UEVOP
MKHWU
TALOV
IQKGK
NJGKD
DDGWP
SBVYZ
RNIHN
IHTKS
DCCKK
KTWYG
CDXLF
JWWR
AOOWZ
ZRNLE
WUQQQ
EODRL
JKVHD
NNHKV
QKVVM
JPXGD
KWDMK
NMODQ
NPGDU
UGDJW
JWNPK
MDODO
MDEWA
VJVVW
IONGM
JNMPV
ISYDK
UETCL
QEEXK
FQBKE
ZVWUX
VLODO
NIKAO
RQFMD
ILIAB
NWBHE
EANKH
FYWHX
RMDEW
NVFTD
FQJWN
VZKWK
MHVWR
XYXOP
NEIHA
LZAYU
OBFMB
JKEOZ
ONAZA
ECESG
ULJVQ
MHFZT
ENMEN
AFYWO
HGBJO
KKOLI
MIQXK
HFUNN
MSIEE
EAAME
ENILH
BYWOQ
TCFWD
ZMWCL
SPFEP
ZZWIM
CPMPP
APRNP
XENGZ
LIPOW
POWKO
THYNS
GFCKK
TPDKU
KHFLW
AJUUA
YVKVI
JJODW
RHEOF
KOPIS
LWWNJ
WDAVL
VCUCL
FKVWP
NPPWK
VBHFY
VHEQA
ESEYE
LISYZ
XRKEO
EABRE
XDGIK
ZUAES
ZWFHV
HHODN
UCDQQ
MPXKR
HYFMN
XUNUM
NREPM
MEPMD
AEVIX
JUPWW
HWBRE
KMHFM
PWKFV
HEONE
ZDWKG
KZYGE
DMKFI
KKFDG
OJKFN
LKHVP
AYMGX
DWXYE
AFMHF
PZTQK
BC
SATHR
WAFTW
JJKIP
EIWKW
KKBDT
VPNPK
SYQMZ
WPMFS
MVPRA
NOJKF
VNKOJ
XNZON
ABEFD
IAMFT
PRINCIPLES OF SOLUTION
This cipher belongs to the class of combined substitution-transposition methods. The
substitution, however, does not follow the method of the ordinary periodic multiple-alphabet
system, nor is the transposition regular or geometrical in its nature. W e shall consider first the
results of the method of substitution.
While at first glance it may be thought that in this system, the encipherment of the groups
of different lengths eliminates the regularity or cyclic nature of the ordinary periodic multiple-
alphabet cipher, such is not strictly the case. It is true that the groups are irregular in length,
due to the action of the interrupter, but at the same time, since the interruption key repeats
itself many times throughout a message, there will be regular and cyclic repetitions of constant sets
of groups. In other words, there is a cycle in the system, composed of a constant number of
groups of irregular lengths. We shall first proceed to determine the length of this cycle.
T he length of this cycle is dependent upon the length of the interruption key, and is the sum oj
the constituent numbers of the key. It is obvious that the key can be 2, 3, 4 . . . up to 13 numbers
in length. It cannot be more than 13 because of the manner in wlu'ch the key is derived from the
upper half of alphabet 1. (See p. 66.)
Now the numbers in these interruption keys do not repeat themselves, and it follows,
therefore, as a simple mathematical fact, that the sums of the numbers composing every possible
keyin other words, the lengths of the possible cyclesare constant and determinate quantities.
Thus, if the key consists of 2 numbers, the cycle will be 1+2=3 letters in length; if it consists
of 3 numbers, the cycle will be 1 -4-2+3=6 letters in length. The following table gives the length
of all possible cycles of the entire system.
TABLE X V III
Length of key
2
3
4
5
6
7
Length of cycle
1 + 2 3
3 + 3= 6
6+4=10
10+5 = 15
15+6=21
21 + 7 = 28
Length of key
8
9
10
11
12
13 .-
Length of cycle
28+ 8 36
36+ 9=45
45 + 10 = 55
65 + 11 -66
66+12=78
78+13=91
69
In order to determine the correct cycle length we proceed to apply the coincidence method
described in the first part of part I of this paper. Tho message is written out in lines of 3, 6, 10,
15 . . . letters in length to correspond to tho various hypotheses regarding length of cycle, as
given in table XV III.
W e must first take into consideration, however, tho fact that somewhere within the body
of the message there is a group of 3 letters, composing the indicators of the message, and since
these 3 letters do not form a part of any cycle, they might throw all the cycles succeeding them
out of phase with the cycles preceding them. But, from the method given for fixing the location
of the indicator group, the latter must occur somewhere within the first 26 letters of the cipher
message. By discarding the first 26 letters of our message then, we may be sure that we have
eliminated this source of distortion. Starting with the twenty-seventh letter of the message,
viz, S, we proceed to set up the message according to various hypothetical cycle lengths. W e
may omit the lengths 3 and 6, at the start as highly improbable. In this case no table of coinci-
dences is really necessary because the arrangement of the message on the assumption of a cycle
of 28 letters gives so many repetitions within the same columns that further corroboration is
unnecessary. Note the repetitions of long polygraphs below:
FIGURE 34
12 a H is
S Y Z H
22 23 24
T H R
8 9 10 11
K U L I
K V V M
0 K M H
W U X L
J J K
M B B Y
A E C L
Z ,'D W
M' D 0
E B C
P W D A
V W R M
C D X
V L M
D E W
H Y F M
V W R M
H V P
A F Y
Y M G X
D H G B
Z 0 N U
P 0 W K
W U E 0
H E Q A
D U E T
P W W B
V F Q J
M H F A
M I Q X
F T T A
F C K K
The next step is to compile individual frequency tables from the columns. There will, of course, be 28 of
them. They have been consolidated in table XIX, and the average frequency per column is given.
TABLE XIX
Column
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
A
1
4
1
1
2
4
1
2
3
3
1
4
4
B
i
2
1
1
3
2
1
1
1
1
1
1
1
C
3
2
1
2
2
1
4
1
D
1
1
4
1
3
2
3
2
3
1
3
1
2
2
2
2
1
1
E
2
3
1
1
2
3
1
1
3
3
2
2
3
6
1
2
2
2
2
4
1
I
1
4
2
2
1
2
3
1
3
2
1
1
1
1
4
2
1
2
1
O
2
1
2
1
2
1
1
1
1
1
1
2
1
2
H
3
1
2
3
4
2
1
2
3
2
3
1
1
1
I
2
2
1
1
4
2
1
1
1
5
1
2
1
1
J
1
1
2
4
2
4
1
1
1
1
2
1
2
K
2
2
4
2
4
1
1
3
4
2
1
1
2
2
2
2
3
2
3
6
1
3
2
c
L
2
3
1
3
4
3
2
4
Ipha
M
3
1
1
1
3
1
3
1
2
1
4
4
1
1
1
2
1
4
2
letu
N
2
2
1
2
2
2
1
2
1
1
3
4
1
4
1
4
2
4
n
0
2
3
3
3
1
2
1
3
1
1
1
1
2
2
4
1
2
1
P
3
3
1
2
2
1
1
5
3
1
5
3
1
1
2
Q
2
1
2
3
1
1
1
1
1
3
1
1
B
2
3
1
1
1
2
1
1
1
2
1
2
8
2
1
3
2
3
4
1
T
1
1
1
1
2
1
U
1
1
5
1
2
3
1
1
1
2
1
V
2
1
4
3
2
3
4
3
2
3
3
2
W
1
3
2
3
3
1
2
2
5
3
2
3
2
3
3
2
1
3
3
X
1
1
2
2
2
1
1
3
2
T
3
1
1
3
1
2
5
1
Z
1
2
2
2
4
1
1
1
2
1
1
3
1
Fre-
quency
26
26
26
26
26
26
26
26
26
26
26
26
26
26
26
26
25
25
25
25
25
25
25
25
25
25
25
25
Number
of
different
letters
12
12
16
14
15
14
13
13
12
11
14
11
12
13
13
14
13
14
11
11
11
13
14
13
14
13
12
16
AvcnfB
frequency
2. 16
2. 16
1.62
1.86
1.75
1.86
2.00
2.00
2.16
2.36
1.86
2.36
2. 16
2.00
2.00
1.86
1.92
1.78
2.27
2.27
2.27
1.92
1.78
1.92
1.78
1.92
2.08
1.56
Showing frequency distributions for the various columns of figure 34.
71
Now, according to table XVIII, a cycle length of 28 requires that the interruption key consist
of 7 numbers, obviously the digits 1 to 7. We do not know how many alphabets are involved,
but they probably do not exceed 7. If only 3 alphabets are involved, then they would be dis-
tributed in 7 irregular groups as follows:
TABLE XX.Basin of S alphabets
Groups Alphabets
1 1
2 12
8 123
4 i_ 231
6 12312
8 123123
7 1231231
Alphabet 1 would be employed twelve times, alphabet 2, nine times, and alphabet 3, seven
times. The following data give the number of times the various alphabets would be employed
on different hypotheses:
TABLE XXI
Basis of 4 alphabets
Groups Alphabets
1 1
2 12
8 123
4 1234
5 123
6 123
7 12
Frequencies
Alphabet 1, 10 times
Alphabet 2, 8 times
Alphabet 3, 6 times
Alphabet 4, 4 times
Basis of 6 alphabets
Groups Alphabet*
1 1
2 12
3 123
4 1234
6 123t56
7 1234561
Frequencies
Alphabet 1, 8 times
Alphabet 2, 6 times
Alphabet 3, 5 times
Alphabet 4, 4 times
Alphabet 5, 3 times
Alphabet 6, 2 times
Basts of 6 alphabets
Groups Alphabet!
1 1
2 12
3 123
4 1231
5 12345
7 1234512
Frequencies
Alphabet 1, 9 times
Alphabet 2, 7 times
Alphabet 3, 5 times
Alphabet 4, 4 times
Alphabet 5, 3 times
Basis of 7 alphabets
Groups Alphabets
1 1
2 12
3 123
4 1234
5 12345
7 123 4 567
Frequencies
Alphabet 1, 7 times
Alphabet 2, 6 times
Alphabet 3, 5 times
Alphabet 4, 4 times
Alphabet 5, 3 times
Alphabet 6, 2 times
Alphabet 7, 1 time
W e shall now try to determine how many alphabets were employed, and how they ro
distributed within the cycle. The basis for this determination rests upon the possibility of
comparing the various frequency distributions produced by each alphabet and determining
which are similar. In other words, we proceed to classify the various frequency tables into
72
several groups, the number of different groups corresponding to the number of alphabets em-
ployed in the message. Accordingly, we select one of the frequency tables as the most closely
approximating that produced by a single mixed alphabet, and try to find others which seem to be
identical with it, as evidenced by the index of coincidence. Since there will be more than one
alphabet in each group, we shall look for more than one high index of coincidence. For example,
if we find three indices which are much higher than the remaining indices, we may conclude that
the alphabets corresponding to these indices are identical with the alphabet which is being
used for comparison.
Returning to table XIX, and following the reasoning given on page 37 of the first part of this
paper, columns 10 and 12 give the closest approximations to a theoretical single frequency table,
and we shall start fitting frequencies with them as bases.
Taking the frequency table for column 10 and applying it to that for every other column, we
get the coincidence data shown herewith:
TABLE XXII
Total occurrences
Coincidences
Differences
Indices of coincidence
i
62
17
-1
-.02
Column..
Total occurrences
Coincidences
Differences . . _
Indices of coincidence
2
10
-22
-.42
18
52
6
-34
-.65
52
9
-25
-.48
17
51
10
-21
-.41
52
7
-31
-.60
18
61
16
-3
-.06
62
6
-34
-.65
i
51
6
-33
-.65
52
5
-37
-.71
30
51
11
-18
-.35
62
10
-22
-.42
21
51
3
-42
-.83
62
11
-19
-.37
33
51
6
-33
-.65
52
9
-25
-.48
23
51
6
-33
-.65
11
52
8
-28
-.54
24
51
8
-27
-.53
12
52
7
-31
-.60
25
61
c
7
-30
-.59
13
52
8
-43
-.83
20
51
6
-33
-.65
14
52
5
-37
-.71
27
51
7
-30
-.59
it
52
8
-28
-.54
29
51
8
-27
-.53
The indices of coincidence for columns 1 and 18 are so high, relative to those for all other
columns, that we may conclude at once that the frequency tables for these two columns belong
to the same alphabet as column 10. To corroborate this calculation, let us consolidate the three
frequency tables to see whether the result of such consolidation will present the appearance of a
single mixed substitution alphabet.
F IUUKE 35
A B E F O H I J K L M N O P Q R S T T T V W X T Z
Consolidated frequency table of columns 1, 10, and 18
W e note at once that this frequency distribution gives every indication of being that of a
single mixed alphabet.
Returning to the table of indices of coincidence for all other columns tried with column 10, we
note that the indices for columns 1 and 18 are in reality so much higher than those for all other
73
columns that we may assume with a fair degree of certainty that the three columns thus classified
constitute all the columns applying to one alphabet. Reference to the diagram (table XX) for a
basis of 3 alphabets shows that alphabet 1 is used 12 times, alphabet 2, 9 times, and alphabet 3,
7 times. Now if we are correct in our assHmption that columns 1, 10, and 18 constitute all
the columns for one alphabet, then this assumption automatically rules out the hypothesis that
the problem involves 3 alphabets because, on the latter hypothesis, there cannot be an alphabet
which is used but three times; and, for the same reasons, a hypothesis of 4 alphabets is eliminated.
It is possible, however, to have an alphabet which is used but three times on hypotheses of 5,
6, or 7 alphabets. In each of these cases, the alphabet to which columns 1, 10, and 18 would
belong is alphabet 5.
Let us proceed now to find another alphabet to which some of the other frequency tables
belong. The table for column 12, having the same average frequency as that for column 10, is
taken as a basis for comparison for the next classification of frequency tables. We may omit
from the calculation the frequency distributions for columns 1, 10, and 18, since they have
already been classified. The data for this test are as follows:
TABLE XXIII
Total occurrences
Coincidences
Differences
Indices of coincidence...
Total occurrences _
Coincidences -
Differences
Indices of coincidence- -
2
52
9
25
-.48
10
52
11
19
-.37
3
52
13
-13
-.25
17
51
6
33
-.65
4
52
6
34
-.65
19
51
8
27
-.53
s
52
8
28
-.54
20
51
15
6
-. 12
< >
52
10
22
-.42
21
51
0
24
-.46
7
52
16
4
-.08
32
51
10
21
-. 41
8
52
6
34
-.65
J3
51
5
36
-. 70
e
52
7
31
-.60
34
51
11
18
-.35
11
52
9
26
-.48
26
51
6
33
-.65
13
52
4
40
-.77
X
51
9
24
-.46
14
52
9
26
-.48
27
51
5
36
-.70
IK
52
4
40
-.77
28
51
10
21
-. 41
W e may with a fair degree of certainty conclude that the frequency distributions for columns
7, 12, and 20 constitute members of another set. But it is impossible in this system to have two
alphabets which are used exactly the same number of times; and since we have already assumed
that columns 1, 10, and 18 constitute the alphabet which is used three times, we are forced to
conclude that one or mpre additional columns in this test belong with columns 7, 12, and 20.
Now the index for column 3, viz, .25, is the closest to the indices for columns 7 and 20, and
it is perhaps likely that it belongs with these columns. But the index for column 24 is .35,
and that for column 16, .37, so that wo must apply a secondary test. Let us consolidate
the frequency tables for columns 7, 12, and 20, and then make a calculation of index of coin-
cidence for columns 3, 24, and 16.
74
A B C
FlGCRE 36
D E F O H I J K L, M N O P a B S T T T V W X Y Z
Consolidated frequency table of columns 7, 12, and 20
Frequency . .- _
Coincidences .
Differences
Indices of coincidence
s
103
21
40
. 39
1C
103
14
61
. 69
34
102
18
48
. 47
It is very probable that column 16 does not belong in the same alphabet with columns 7, 12,
and 20, but columns 3 and 24 do. To corroborate, let us add their corresponding frequency dis-
tributions to the consolidated table.
FIGURE 37
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z
Consolidated frequency table of columns 3, 7, 12, 20, and 24
There is every indication that we are dealing with a distribution for a single mixed alphabet,
and that no foreign elements have been introduced. If we have succeeded in collecting all the
columns which should belong to the alphabet being studied, it follows that the latter must be
alphabet 3, for it is used five times on a hypothesis of 5, 6, or 7 alphabets.
Let us now consider the sequence of alphabet numbers which results when we set down the
alphabet to which each of the columns now classified belongs.
F IGURE 38
1-2-3-4-6-6-7-8-9-10-11-12-13-14-16-16-17-18-19-20-21-22-23-24-25-26-27-28
5 3 3 5 3 5 3 3
Now it will be remembered that in encipherment the letters in each group were reversed
before grouping into fives for telegraphic transmission. It follows, therefore, that the alphabet
numbers applying to columns when the message is set up as shown in figure 32 will be hi groups
of descending series of numbers and this descent is complete in each case to number 1. Thus,
for example, the series in the illustrative message on page 67 are as follows:
F IGURE 39
J-2-3-4-5-6-7-8-9-10-11-12-13-14-15-16-17-18-19-20-21
1-3-2-1-1-2-1-3-2- 1- 3- 2- 1- 2- 1- 3- 2- 1 -3-2-1
Returning now to figure 38, we see that our classification of columns is in accordance with
the requirement of descending series of numbers, and we may even fill in many numbers without
any further test. Thus, the sequence becomes:
F IGURE 40
1-2-3-4-5-6- 7-8-9-10-11-12-13-14-15-16-17-18-19-20-21-22-23-24-25-26-27-28
5-4-3-2-1 3-2-1- 5-4-3-2-1 5- 4- 3- 2- 1 3-2-1
To corroborate this sequence, let us consolidate the frequency tables for columns 2, 11, and
19 into one table; those for columns 4, 8, 13, 21, and 25 into a second table; and those for columns
75
5, 9, 14, 22, and 26 into a third table. We do this to see if the resulting table in each case
corresponds to that of a single mixed alphabet.
F IGURE 41
Columns 2, 11, and 19
A B C D E F O H I J K L M N O P Q R 8 T U V W X Y Z
Columns 4, 8, 13, 21, and 25
A B C D E F G H I J K L M N O P Q R S T T T V W X Y Z
A B C D E F O H I
Columns 5, 9, 14, 22, and 26
J K L M N O P Q R S T U V W X Y Z
The approximation to single alphabet distributions is, in each case, very close, and we may
assume our work thus far to be correct. The only columns remaining to be assigned to alphabets
are 6, 15, 16, 17, 23, 27, and 28.
Now column 6 can only belong to either alphabet 4 or alphabet 1; and the same alternatives
present themselves with respect to column 23. This is because these columns, being bounded on
either side by columns already distributed into alphabets 1 and 3, must either be members of
the descending sequence 4 3 2 1, or must be the column forming that part < > f the interruption key
which consists of the number 1, thus constituting the isolated column which must always be
enciphered by alphabet 1. In order to determine which of these alternative assignments is the
case, we may make a single calculation by finding the index of coincidence when column C is
grouped with columns 5, 9, 14, 22, and 26, and when grouped with columns 2, 11, and 19. In
order to make the results strictly comparable we should include but three columns belonging to
alphabet 1, because there are only three columns assigned thus far to alphabet 4. Let us make a
special consolidation of columns 5, 9, and 14 for this test.
B C D E F Q H I J K
F IGURE 42
M N O P Q R S T T / V W X Y Z
B O D
Special consolidated frequency table of columns 6, 9, 14
F Q H I J X L M N O P Q R S T T J V W X Y Z
Consolidated frequency table of columns 2, 11, and 19
Total frequency, columns 6, 5, 8, and 14 104 Total frequency, column 6, 2, 11, and 19-
Total coincidences, column 6 with 5, 9, and 14... 26
Difference S(26)-104 -26
Index of coincidence . 25
Total coincidences, column 6 with 2, 11, and 19.
Difference 3(19)-103
Index of coincidence
103
19
-46
-.46
76
There seems to be no doubt that column 6 belongs to alphabet 1 and not to alphabet 4.
This automatically makes it necessary that column 23 be assigned to alphabet 4 since we cannot
have two no. 1 columns in the same interruption key.
Let us set down what we have:
F IGURE 43
1-2-3-4-6-6-7-8-9-10-11-12-13-14-16-16-17-18-19-20-21-22-23-24-25-26-27-28
5-4-3-2-1-1-3-2-1- 5-4-3-2-1 6-4-3-2-1 -4-3-2-1
W e now proceed to compare these sequences with the sequences of alphabets within the
reversed groups on hypotheses of 5, 6, and 7 alphabets.
W e may conclude, by a short analysis, that a hypothesis of five alphabets must be correct,
because the distribution of columns as now made, if it be correct (and we feel fairly certain of
our work so far), allows no room for any alphabet 6 or 7. This analysis is fairly simple. Let
us assume a hypothesis of 6 alphabets.
Consider the two places where alphabet 6 would have to fall:
F IGURE 44
13-14-16-16-17-18-19 .
2-1 6-6-4-.
25-26-27-28-1-2
2- 1 6-5-4-
According to this diagram, column 27, isolated as it stands, would have to bear the number 1.
But we found, already, that column 6 is the column which forms the isolated group containing
only one number. Hence, a hypothesis of 6 alphabets cannot be correct.
Let us assume a hypothesis of 7 alphabets. Note these two places once more:
FIGURE 45
13-14^16-16-17-18-19
2-1 7- 6- 6- 4- . . . .
26-26-27-28-1-2
2- 1- 7- 6-5-4-
Here again we would have an isolated column (15), which for similar reasons cannot bear
the number 1.
W e are left, therefore, only the hypothesis of 5 alphabets. This requires that alphabet 1 be
used nine times, and alphabet 2, seven times. W e have already distributed 6 columns to alphabet
1, and 5 columns to alphabet 2, leaving 3 more columns to be assigned to the former, and 2 more
to the latter. W e may assign columns 27 and 28 immediately to alphabets 2 and 1, respectively.
The sequence thus becomes:
F IGURE 46
1-2-3-4-6-6-7-8-9-10-11-12-13-14-16-16-17-18-19-20-21-22-23-24-25-26-27-28
6-4-3-2-1-1-3-2-1- 5- 4- 3- 2- 1 5-4-3-2-1 -4-3- 2- 1- 2- 1
5 1 3 5 5 4 2
If we make a grouping of columns to correspond with the probable interruption key, as
shown in figure 46, we note that we have 2 extra groups of 5 and no group of G and 7, respec-
tively, as required by a key of the determined length. If, however, we shift the sequence so as
to bring columns 27 and 28 to the left, which is perfectly legitimate since we are deah'ng with
a cyclic key, we have:
F IGURE 47
27-28-1-2-3-4-5-6-7-8-9-10-11-12-13-14-15-16-17-18-19-20-21-22-23-24-25-26
2- 1-6-4-3-2-1-1-3-2-1- 5-4-3-2-1 - 5-4-3-2-1 -4-8 -2-1
7 1 3 5 5 4
77
W e now have our group of 7 columns, but lack the 1 of 6 and 1 of 2; but we have 2 groups
of 5 columns. Obviously, the only thing to do is to insert the sequence 2-1-1- in its proper
place as follows:
FIGURE 48
27-28-1-2-3-4-5-0-7-8-9-10-11-12-13-14-15-16-17-18-19-20-21-22-23-24-25-26
2- 1-5-4-3-2-1-1-3-2-1- 5-4-3-2- 1- 2- 1- 1- 5- 4- 3- 2-1 -4-3-2-1
7 1 3 5 2 6 4
W e have thus reconstructed the complete interruption key, and may now proceed to tran-
scribe the message in groups in accordance therewith. Since the first group contains 7 letters,
and since the letter S, with which wo started our first transcription (fig. 34), is labelled 1, it is
clear that we must start our second transcription with the two letters preceding S, i.e., the
twenty-fifth and twenty-sixth letters of the cipher message. Thus:
7
P M S I E E T
R G M N J G K
i
P
D
D K U
Q K V
2
H W
K E
B R E S A T
N E I H A E
H R K Z
A A M E
FlOURE 49
6
L I S Y Z
V M F 0 B
etc., etc.
After this transcription a third set-up is necessary, in which the sequences of letters are
reversed within groups, in order to bring the cipher letters back into the original arrangement
before transposition. The rearranged set-up is as follows:
1 2
7
1 4
I
N
L
T E E
K G J
R X W
D G X P
K W P E
0 0 D 0
K W I E
C T Z 0
D G K N
E H R L
U K K C
D 0 H H
W N K P
G D F K
F H M H
0 A R P
N R P A
0 A R L
P W V K
N 0
D M
E
P E
W J Q F
B A F H
K X Q I
L A T T
E K K C
1 2
M P
G R
P
J
F H K
J P W
R B A
L V K
E N 0
E K J
X T D
C W M
C O M
S I
N W
K Q Q
K N A
V M N
N E M
C W W
F Z G
L N R
R N K
V P M
M F A
M K W
F M A
F N G
1
i
P
D
K
Z
F
B
W
F
I
0
G
N
B
W
Z
E
P
0
X
G
W
N
E
T
0
S
F IGURE 50
3
i i
U K
V K
K 0
U W
J J
B M
E A
J D
K T
W Z
W J
K M
U E
F M
P C
0 J
U C
V Z
U N
V N
E Y
0 K
M D
N Y
K J
E Y
3
D
Q
E
V
V
F
P
W
H
F
D
D
H
P
T
W
V
W
U
M
X
K
F
H
V
E
1 2
Z Y
B Q
W M
Y A
M E
Q 0
I N
Z W
D G
Z K
B A
Q Y
Z Q
B E
D W
R W
F Y
R W
G M
G H
N 0
W 0
E U
Q E
E U
W W
5
3 4
S I
F M
F H
Z L
P I
W Y
R C
D 0
P N
V H
I L
S I
V J
I K
P P
V V
H L
V V
Y A
D T
Z N
P I
W H
H V
D H
P U
6
L
V
M
X
K
B
L
J
S
F
I
F
L
S
M
J
C
J
M
F
X
L
K
S
V
J
2
1 2
W H
E K
F A
E U
N E
V Y
N H
A U
R U
G Y
C E
Z M
W Z
C C
V A
D M
N M
D M
V X
J B
E U
0 K
D 0
A A
C T
C B
6
i
T
E
E
I
Z
G
N
K
N
K
F
K
N
D
R
W
N
W
F
W
Q
M
D
Z
R
2 S
A S
A H
P S
A H
Y V
D X
Q D
W D
0 D
P N
P S
Q Q
W W
M F
K X
Y F
P V
Y F
K J
0 P
U W
Y F
Y S
P X
W V
4 6
E R
I E
E W
L I
B S
I V
0 M
Z S
M F
P V
G S
I G
L M
L X
P M
A W
H K
A W
0 K
I L
P 0
H B
I L
I V
H M
i
B
N
T
N
F
K
N
E
Q
E
E
K
I
D
L
E
L
E
N
0
V
V
R
E
L
4
1 2
Z K
E M
G D
E A
M D
E H
0 A
B K
Z A
H A
0 K
G Y
D C
0 D
F K
N E
F K
N E
Z B
W P
N J
D M
K Z
I K
N U
3 4
R H
A A
D B
U U
K W
K I
K I
K G
Z A
Q A
P E
W T
U J
0 E
J 0
X H
J 0
X 0
V S
P N
Q Q
P E
V K
Q T
F H
78
W e are now confronted with a rather simple case of the analysis of five reciprocal and
interrelated alphabets. They are composed of the consolidated frequency tables applying to
the columns which belong in the same alphabets and are as follows:
TABLE XXIV
1
Columns 669141617222628
A B C D E F G H I J K L M N O P Q R S T t T V W X Y Z
Columns 481316212527
A B C D E F O H I J K L M N O P Q R S T T T V W X Y Z
Columns 37--1220 24
A B C D E F G H I J K L M N O P Q R S T T T V W X Y Z
Columns 2111923
A B C D E F O H I J K L M N O P Q R 8 T T 7 V W X Y Z
Columns 11018
A B C D E F O H I J K L M N O P Q R S T T T V W X Y Z
It will be unnecessary in this paper to discuss the method of deciphering the message by an
analysis of these five single-frequency tables. Suffice it to indicate the values obtained from the
decipherment. They are given below, where the values obtained from a knowledge of the
reciprocal relation are placed within parentheses. Only four values remain unknown, all in
alphabet .
r
> .
79
(1)
(2)
(3) .
(4)
(5)
A
K
A
E
A
W
A
0
A
TABLE XXV
B C D E F G H I J K L M
L M O N R T V W Y A B C
B C D E F G H I J K L M
X P H A (Q) U D (K) L I (J) S
B C D E F G H I J K L M
(Y)(Z) E D I (J) S F G N (Q) U
B C D E F G H I J K L M
F G ( Z ) I B C N E X R T V
B C D E F G H I J K L M
N
E
N
0
N
K
N
H
N
0 P
D U
0 P
N C
0 P
(V) T
0 P
A S
0 P
Q
't
X
Q
%
F
Q
"
L
Q
't
W
0
R S
F Z
R S
W (M)
R S
X H
R S
(K) P
R S
T
G
T
Y
T
P
T
L
T
u
P
u
G
u
M
u
Y
U
V W
H I
V W
(Z) R
V W
0 A
V W
V W
x y
(Q) J
X Y
B T
X Y
R B
X Y
(J) U
X Y
S
V
2
C
2
D
2
(F) G H (K) V A B (C) X P D R T W S (J) L 0 (M) E N I
We may now attempt a reconstruction of the original, or primary, alphabet, of which these
are secondaries. Note the following values:
In alphabet 1, E=N
In alphabet 2, N=0
In alphabet 3, 0=V
In alphabet 4, V =M
In alphabet 5, M=T
Now if E occurs in the upper half of alphabet 1, the table of alphabets must contain a column
like this:
E
Alphabet 2
Alphabet 1
Alphabet 3
Alphabet 5
M
Alphabet 4
Let us assume this to be correct. In alphabet 3, E will again be in the upper half of the
alphabet. We have these values:
In alphabet 3, E=D
In alphabet 4, D =Z
In alphabet 6, Z =?
But to this we may add two more values since we have the value of E in alphabet 2. Thus:
In alphabet 1, K=A
In alphabet 2, A=E
In alphabet 3, E=D
In alphabet 4, D=Z
In alphabet 5, Z =?
Since the upper line in alphabet 3 is displaced but one letter to the right as compared with
that of alphabet 1, we have two columns in the table as follows:
f E K
Alphabet 1 j
Alphabet 3 |
Alphabet 2
Alphabet 5
W e may continue thus:
Hence we have this:
0 E J
V D 1
., > Alphabet4
M /i J
In alphabet 1, W =I
In alphabet 2, I=K
In alphabet 3, K=N
In alphabet 4, N=H
In alphabet 5, H=C
f E K W
Alphabet 1 {
N A T
Alphabet 3 I
Alphabet 5
0 E K
V D N
M
? U C
> Alphabet 4
n J
The process is very simple and easy to continue. Finally we have this:
Alphabet 1
F X L Y
R Q B J
l W A J . i \ v { D t i I
A
i u u * o
1 1 1 > K
r o E KW F x
L
|
A l p h a b e t 2
Alphabet 3 \
J
1 V D N A I R 0 1
, , , [ M Z H O E K W
A l p h a b e t 5 |
? u c s v D N
W e may now continue from the sequences given already. Thus in the last line we see the
sequence . . . S V D N, in the fourth line, V D N A I R Q . Hence, we may add A I R Q
to the last line. The same process applied to the other lines gives us:
2 3 4 6 6 7 8 9
, , , f E K W
Alphabet 1 {
N A I
Alphabet 3 I
f M Z H
Alphabet 5 (
?

c
F X L Y
R Z B J U
O E K W F X L Y
V D N A I R Q B J
O E K W F X L Y
S V D N A I R Q B J
M Z H 0
C S V D I
M Z H )
A1
P
habet2
U C S
> Alphabet
81
We may fill in the rest from the alphabets themselves, and we have:
FIGURE 51
1 2 3 4 8 7 8 9 10 11 12 13
fE K W
Alphabet l (
N A I
Alphabet 3 j
Alphabet 5 |
F X L Y G P M Z H O
R Q B J T U C S V
O E K W F X L Y G P M Z
V D N A I R Q B J T U C S
M Z H O E K W F X L Y G P
T U C S V D N A I R Q B J
D l
> Alphabet 2
i Alphabet 4
Taking alphabet 1, a speedy reconstruction of the original rectangle is at once effected.
Thus:
FIGURE 62
E X P O R T S
K L M N Q U V
W Y Z A B E D
F G H I J
The keyword is EXPORTS. In conformity with the agreements of the system, the indicators
should be K X Y and indicate the following:
K, the initial letter of the distorted alphabet;
X, the fifth letter in alphabet 1, hence five alphabets;
Y, the seventh letter in alphabet 1, hence seven groups arranged as follows:
E K W F X L Y
1 3 2 0 4 7
The indicators will be after either the fifth letter or the nineteenth (corresponding to the
numerical value of the initial or final letter of the keyword). W e find K X Y after the nineteenth
letter.
We may now proceed to decipher the first few groups of the message and the entire solution
is at hand. It is as follows:
1 3 6
1
K
K
A
1 2
H N
V N
H 0
3
V
H
S
i
L
G
T
2 3
I Q
K Q
I L
4 S
K G
I L
E R
1 2
K N
N K
E I
i
N
E
N
2 3
H K
Q V
F 0
4 S
V Q
K H
R C
i
E
N
E
1 2
E X
0 X
D B
X Y ) X 6
i
P
2 3
M S
S 1 2
E E T
X E T E E I S M P
R I G A D E O C C
1
P
P
U
1
D
U
P
2
K
K
I
3
U
D
E
i
L
Z
S
2
I
Y
T
3
S
S
H
4
Y
I
E
8
Z
L
R
'A hostile reinforced brigade occupies the
82
2. Solution of a series of messages.The solution of a series of messages in this cipher
presents an interesting demonstration of the fact that in cryptography it is often the more
insignificant details of a system that enable the cryptanalyst to solve the enemy's messages
rather than any definite weakness of the method from the cryptographic point of view. In
this case, the solution of a series of messages based upon the same keyword, but involving
all the manifold modifications of alphabets and interruption key, can be achieved without
attempting the decipherment of a single message. The method involves merely an analysis
of the indicators for a series of messages, resulting in a direct and speedy reconstruction of the
various generating rectangles derived from the same keyword.
In the first place, it may be pointed out at once that the various generating rectangles
based upon the same keyword consist of two parts: A constant sequence, consisting of the key-
word proper, making up the first line of the rectangle, and a variable, or revolving sequence,
consisting of the remaining letters of the alphabet, or as we shall term it, the "residual sequence",
making up the remaining lines of the rectangle.
In the second place, the length of the interruption key for each message can be determined
by applying the principles of coincidence as explained on page 68. Each message is then
accompanied by the number thus found.
Given the beginnings of a series of 18 messages, the lengths of whose interruption keys
have been determined and are as indicated below, let us proceed to an analysis of these lengths,
which in a short time will lead to a direct reconstruction of the keyword.
FIOTJBB 58.Beginnings of teriet of mtttaget
MMU
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Lengt h
o(ky
K
P
Q
0
A
F
A
D
V
I
C
Y
D
G
S
I
0
T
I B
0 N
Z M
Z A
A Z
M Y
T N
Z N
R A
A B
T T
L P
T X
J E
B U
B U
Y Z
M R
I Y
Q D
V R
P N
S L
0 T
L B
D J
Y Q
0 Q
0 S
A Y
D R
T M
X L
R T
V A
G V
F
A
P
V
I
M
P
U
E
K
J
N
C
J
R
W
R
I
P P
M N
X V
Z U
G I
M F
T R
B R
R H
U L
X V
L W
G T
B P
Q Y
Y J
G U
E E
L
D
T
L
A
E
Z
U
P
D
X
0
K
P
L
A
Y
D
H
U
W
V
X
R
T
L
0
T
P
F
T
U
L
H
Y
K
GZ O P B M A W F V T B N B I
B N K Y O A W D R T W Q Z N J
H f f J X U H M E Q Z W A X G U
U Y H M K Y L J L W G V F O P
V H G W M Q M Z S I O I T Y E
B B B S F O T P F I A Y K N A
Y W O B A V I K Y H Z U A L X
M F F M H W V R E A A Q T S S
P V J Y O W U N K Q V A C A W
D E Q D M J A U C W A S D K Y
O A D S F J J P Y H V L C K H
G K P D S G N N Q O C U O F K
C K O Z U K Y A Z Y T U K N I
Q T D U F C T C A L C V Z E D
A I K U O L P H Y B Y T A K F
T H O P B K L Q C V O G X L G
B G H W U M E K D S M G H F G
H X F U U V G F I Z S Z K J T
F U
L M
S V
Z G
R G
V Q
V K
Q S
Q N
T P
P D
0 N
A N
A P
G L
N J
N T
Y Y
E B
U R
H Y
A Y
B N
V L
R K
0 N
E U
A D
A U
V F
E G
U Q
H Y
X P
0 G
K 0
6
5
4
8
7
8
8
8
8
7
5
8
4
5
5
6
4
4
Let us now turn our attention to the generating rectangles that are possible, indicating
merely their outlines, for keywords of 5 to 10 letters in length, and numbering the squares
which will contain the first, second, third, to the ninth letters of the first line of the alphabet
table. We will assume that no keyword will consist of more than 10 letters, and no interrup-
tion key of more than 9 numbers. The principles to be elucidated may be extended by
the reader to cover other cases.
FlOtTMD 54
(A) (B) (0)
i a t 4 t
1
2
3
4
8
6
7
8
0

1
2
8
4
5
6
7
8
9
1
2
8
4
5
6
7
8
9
(D)
(B)
1 3
1
2
8
4
6
8
7
8
9 1
2
8
4
0
0
7
8
9
(F)
1
2
8
4
5
8
7
8
9
10
Note now that in (A) of figure 54, the seventh letter of line 1 of the alphabet table based
upon a keyword of 5 letters will be the second letter of the keyword proper and, as such, does
not change its position. This follows from the method of constructing the alphabets from the
generating rectangle. In other words, with a 5-letter keyword, no matter what letters be
chosen as the initial letters of the various possible generating rectangles, all numerical keys
consisting of seven numbers will always be designated by the same letter in the indicator group,
because the first line of all generating rectangles based upon the same keyword is always the
same. Since the letter which designates the length of the interruption key is the third letter
of the indicator group for a keyword of 5 letters, all messages which factor for a key of seven
numbers must show the same letter as the third element of the indicator group. Conversely, if
messages which factor for a key of seven numbers show a constancy with respect to the third
element of the indicator group, then it must follow that the generating rectangle is based upon a
5-letter keyword, and that the corresponding indicator letter is the second letter of the key-
word proper. Similarly, when the third element of the indicator group is constant in messages
which factor for a key of six numbers, a generating rectangle based upon a 6-letter keyword is
indicated and the corresponding indicator letter is the second letter of the keyword proper.
When the third elements of two utt of indicator groups are constant and the interruption keys
for the corresponding sets of messages consist of five and nine numbers, a 7- or 8-letter keyword
is indicated and the corresponding indicator letters are the second and third letters of the key-
84
word proper; when they consist of four and seven numbers, a 9- or 10-letter keyword is indicated
and the corresponding indicator letters are also the second and third letters of the keyword
proper.
The first step is, therefore, to determine the length of the keyword.
Let us arrange the series of messages in accordance with the already determined lengths of
their numerical keys. Thus:
FiauBB 55.Mettaget rearranged
Ma ng* o(
3
13
17
18
2
11
14
15
1
16
5
10
4
6
7
8
9
12
Q Z
D T
0 Y
T H
P 0
C T
G J
S B
K I
I B
A A
I A
0 Z
F M
A T
D Z
V R
Y L
H V
X D
Z V
R G
N Q
T 0
E T
U X
B I
U R
Z S
B 0
A P
Y 0
N L
N D
A Y
P A
R
R
A
V
D
S
M
L
Y
T
L
Q
N
T
B
J
Q
Y
P
C
R
I
A
J
J
R
F
W
I
K
V
M
P
U
E
N
X V
G T
G U
E E
M N
X V
B P
Q Y
P P
Y J
G I
U L
Z U
M F
T R
B R
R H
L W
T W
K T
Y Y
D K
D U
X P
P U
L L
L H
A H
A X
D T
L V
E R
Z T
U L
P Q
0 F
H W J X U H M E Q Z W A X G U
C K O Z U K Y A Z Y T U K N I
B G H W U M E K D S M G H F G
H X F U U V G F I Z S Z K J T
B N K Y O A W D R T W Q Z N J
O A D S F J J P Y H V L C'K H
Q T D U F C T C A L C V Z E D
A I K U 0 L P H Y B Y T A K F
G Z O P B M A W F V T B N B I
T H O P B K L Q C V O G X L G
V H G W M Q M Z S I O I T Y E
D E Q D M J A U C W A S D K Y
U H Y M K Y L J L W G V F O P
B B B S F O T P F I A Y K N A
Y W O B Q V I K Y H Z U A L X
M F F M H W V R E A A Q T S S
P V J Y O W U N K Q V A C A W
G K P D S G N N Q O C U O F K
S
A
N
T
L
P
A
G
F
N
R
T
Z
V
V
Q
fi
0
V
N
T
Y
M
D
P
L
U
J
G
P
G
Q
K
S
N
N
H Y
E G
0 G
K 0
U R
A U
U Q
H Y
E B
X P
B N
A D
A Y
V L
R K
0 N
E U
V F
K.y
4
4
4
4
5
5
5
5
6
6
7
7
8
8
8
8
8
8
Note, now, the repetitions in column 15. W ithin the sections containing the messages whose
Interruption keys consist of four and seven numbers, the letters U and M are constant. This,
according to the principles stated above, indicates a keyword of 9 or 10 letters and that the
second and third letters of the keyword proper are UM. It is true that there are repetitions
within this column other than those involving the letters U and M, but there is no constancy in
such repetitions. For example, the letter F is repeated in column 15 within the sections applying
to keys of five numbers, but the letter 0 is also repeated within the same section. The letter B
is repeated within the section applying to keys of six numbers, but in this case the entire indicator
groups are constant, OPB, showing that messages 1 and 16 are in the same key. The constancy
of the third elements of indicator groups must be definite as regards the two sets of corresponding
interruption keys; it must involve keys of five and nine numbers (for keywords of 7 or 8 letters)
or of four and seven numbers (for keywords of 9 or 10 letters). It cannot involve keys, for
example, of four and five numbers, or, five and seven numbers. W e may regard it as certain,
then, that our assumptions with regard to the keyword, as given above, are correct.
The indicator group is located within columns 13,14, and 15, and we may reasonably assume,
because the indicator groups are placed after the twelfth letter of the messages and since the
twelfth letter of the normal alphabet is L, that the initial or final letter of the keyword is L.
Now let us make a list of all the indicator groups, arranging them in alphabetical order in
accordance with their first letters. Thus:
I
85
TABUS XXVI
Menage number
6
11
14
8
18
6
4
17
9
Indicator group
B S F
D S F
D U F
F M H
F U U
G W U
H H K
H W U
J Y 0
Length of inter-
ruption key
8
5
5
8
4
7
8
4
8
Menage number
3
15
2
7
13
1
16
12
10
Indicator group
J X U
K U 0
K Y 0
0 B Q
0 Z U
0 P B
0 P B
P D S
Q D M
Length of inter-
ruption key
4
6
5
8
4
6
6
8
7
From the arrangement of the messages in figure 55 and this list we conclude the following:
(1) The keyword consists of either 9 or 10 letters.
(2) The letter L is either the initial or final letter of the keyword proper.
(3) The letters UM form the second and third letters of the keyword proper.
(4) The letters B, D, F, G, H, J, K, 0, P, Q are not in the keyword proper.
We now take each indicator group and from'its accompanying data make certain deductions
with respect to the sequence of letters in the keyword alphabet. Thus, for example, the indicator
group B S F, applying to a message, the length of whose interruption key is 8, means that certain
of the letters within the generating rectangle concerned are as shown in figure 56, where both
possibilities as regards the length of the keyword are indicated.
1
B
(A)
UIM _ |
IF |
FIGURE 56
1 B
a
U
3
U
F
(B)
4 8 7 8 t 10
Since there is room for but one letter between B and F, it follows that only one of the inter-
vening letters, C, D, and E, is left in the residual sequence, the other two being in the key word proper.
Now the indicator group D S F shows that D is the initial letter for the generating rectangle for
message 11, and, since a letter which can be an initial letter of an indicator group cannot be a
letter of the keyword proper, it follows that the order in the residual sequence is B D F, and
that the letters C and E are in the keyword proper.
This is confirmed by the indicator group D S F, which accompanying message 11 has an
interruption key consisting of five numbers. Thus:
F IOUKB 57
(A) (B)
1 3 1 4 8 6 7 8 0 1 2 3 4 6 0 7 8 9 1 0
D
U
F
JW

D
U
F
H

8.6
Again, the indicator group F M H, applying to message 8, whose interruption key consists
of eight numbers, shows that the sequence must be B D F G H. Thus:
FIOTJRE 58
(A)
1 2 3 4 8 6 7 8 9
II
I I
(B)
1 2 3 * 5 6 7 8 9 1 0
T
-^
M
H
-
-hi-

There is one space vacant between F and H, which must be occupied by G.

The conclusions furnished by the other indicator groups are given in the diagrams below.
The question as to whether B 9-letter or a 10-letter keyword is involved will be determined later.
i si
HMK 81 H
_U II
K"
F IGURE 59
H (I J) K
Since J is present as an
initial letter the sequence
must b e : B D F G H J K
l i t
JYO8 J
J (K L M) 0
Since K is present as an
initial letter, the sequence
must b e : B D F G H J K O
1 3
OBQ 8 0"
M
Q
0 ( ) Q
Since P falls naturally
between 0 and Q, the sequence
must b e r B D F G H J K O P Q
In the absence of additional indicator groups, with different initial letters, we cannot
continue in the same manner and reconstruct the entire sequence. Let us, however, try to
determine now whether the keyword contains 9 or 10 letters. W ith the residual sequence
as determined above, let us fill in the generating rectangle for the indicator group D S F,
according to both assumed keyword lengths. Thus:
FIGURE 60
(A) (B)
1 J 4 8 7 8 1 2 S 4 6 8 7 8 9 1 0
D
S
U
F
M
G
1
HfJ
1
K 0
P
B
-Q D
U
F
M
G H J K
Now the interruption key for the indicator group D S F consists of five numbers. The
system is such that the number of alphabets employed in a message must be equal to or less
t han the length of the interruption key; it cannot be more than this length. Hence, the letter
S of the indicator group D S F must indicate a number of alphabets equal to five, or less than
five. Upon the basis of a 10-letter keyword, the number of alphabets indicated by the letter S
would be impossible; but upon the basis of a 9-letter keyword, the number of alphabets would
be three, which is very probable. W e may tentatively consider it as established that the key-
word is 9 letters in length.
87
Note now diagram (A) of figure 60. There arc six spaces vacant at the end of the key-
word alphabet. The letters from S to B in the normal alphabet arc T U V W X Y Z A, n
total of eight. Now we know that U is in the keyword proper so that the residual sequence
docs not contain this letter. Of the letters, V W X Y Z A, the one most likely to be present in
the keyword proper is A, leaving the sequence V W X Y Z as the end of the key-word alphabet.
The keyword proper consists of the letters not present in the residual sequence. It must,
therefore, consist of the letters A, C, E, I, L, M, N, R, and U. W e know that the letters U M form
the second and third letters of the word; and the position of the indicator groups in the cipher
text makes L very probable as the initial or final letter of the word. Now, very few words begin-
ning with L U M and containing the other letters A, C, E, I, N, and R can be found. But if we
assume L to be the final letter of the keyword, the most probable ending would be C A L. Given
. U M . . . C A L , the word NUMERICAL soon suggests itself.
W ith the keyword at hand, every generating rectangle can be constructed at once, and the
messages may now be deciphered as rapidly as by the legitimate recipient.
The decipherment of message 1 follows herewith:
FIQ URK 61
M A W F V T B N B I F U E B . . .
R Generating rectangle
N U M E R I C A L
O P Q S T V W X Y
K I B I Y
Alphabet 1
Alphabet 3
Alphabet 5
ff

C
F P P L H
0
G
N
Y
S
A
Z
I
0
T
F
X
U
V
Z
G
R
K
P
H
U
I
N
L
B
C
P
V
0
Y
M
W
B
H
Z
T
G
Q
J
M
C
U
G
Z
D
A
Q
W
P
I
0
E
X
D
J
B
V
P B
S F
K L
E S
A X
M Q
H C
Alphabet 2
K1
\ Alphabet 4
1 2 1 2 3
K I B I Y
I K Y I B
Z E R 0 H
H
1 2 3 4 6 6
F P P L H G
G H L P P F
0 U R W I L
W
1 2 3 4 5
Z M A W F
F W A M Z
L B E A T
Z B D F G H J K
Interruption key
1
V
N
T
4
2
T
B
W
3
B
T
0
N 0
2 3
1
N
V
0
Z U
e 6
1
i
B
B
C
P
4
2
i
I
F
L
B
i
2
F
I
0
"Z ero hour will be at two o'clock . . . "
The fatal defect in this system lies in the fact that a key is used which introduces a frequently
repeated cycle within a message. The determination of the length of this cycle, and its recon-
struction by means of a comparison of alphabets based upon the index of coincidence, enables a
speedy solution to be attained. The insertion of the key indicators within messages makes the
reconstruction of the keyword and the consequent solution of a series of messages very easy.
The many details involved in encipherment, and decipherment, concomitants of an attempt to
make the entire operations dependent upon the knowledge of a single keyword uud the case
with which a solution may bo achieved in the case, of a single message or a scries of messages makes
this cipher unsafe for use in either the field or the more important operations of the larger head-
quarters in the rear.
o
A SPECIAL PROBLEM
This Special Problem consists of 15 ciphertest messages. Each is enciphered in the same system. The
system is known to be a cipher machine whose result is polyalphabetic substitution with a lengthy key between
50 and 100 the period of the system. Although every message uses the same key, each begins at a different
point on the key. Regarding the polyalphabetic substitution, the plaintext alphabet is the standard alphabet
and the ciphertext alphabet is mixed.
By means of analyzing coincidences the cryptanalyst's first task is to determine the length of the key.
Thereafter, the "starting point" on the key of each message should be determined. The first group of each
message is an "indicator" group which shows this "starting point," but only until after the messages have been
correctly placed can the indicator group probably be analyzed correctly.
After determining the "starting point" of each message by means of coincidences the messages may
be placed one beneath another to provide in a sense a giant polyalphabetic substitution problem with mixed
alphabets. For each key there will be a considerable number of letters enciphered with the same setting of
plaintext and ciphertext alphabets.
A certain amount of information is known about the plaintext of the messages. The text is of a military
nature and includes such words as ENEMY , ARTILLERY , BRIGADE, GENERAL, INTELLIGENCE,
PRISONER OF W AR, etc. Most importantly, each message specifically begins with the words MESSAGE
NUMBER COLON and ends with the word STOP. Here an important point must be added. The
cipher machine automatically produces spaces between each word and the letter Z is used for this purpose.
Each message in effect begins, therefore, with the letters MESSAGEZ NUMBERZ (number) ZCOLONZ,
and ends with the letters ZSTOP.
As only one mixed ciphertext alphabet is employed, for each key there will be lateral symmetry within the
ciphertext alphabets.
The problem is not an easy one. But with knowledge of the message "beginnings" and "endings" and
particularly with messages resulting in a large number of letters being enciphered with each key, a "solution" is
not beyond reach.
89
NO. 1
B
E
X X X
B Q T
A K Q A N
F H L G H
VI Y T L
U T VJI
T R Q Z C
S W JP I
JZ X M P
C T W U J
Q T I
C A P E X
VG Z R G
S K W P Y
JM R U Y
Z M
U L.
M B
Y 0 D
JE G
Y JF
R JG F K
JN F U M
I VL B K
K Y I L T C
S F VB O O
0 H M K C M
T E W 0 E X
B VH D R D
H C JG R L
1 Y T N S J
C D D G K
H VC U J I
H P C P A P
B P T M
P F M K
F B N A
B G B U
X X
Z T
I L
D K P W V
G F
W G
D G
R
X
F M
N R
G H S B
N S X S
Z JN B H
G E W V D
W H C M U
N 0 VT F
M R K R H
X Y JG P
Y G A U
P H Y K
C
C
( 253 )
NO. 2
< 21 9 )
NO. 3
I
C
I
B
J
0
N
J
P
O
J
Z
N
I
Q
S
J
B
B
X
K
A
F
C
N
G
G
F
X
X
Z
S
P
W
B
N
Z
X
T
J
S
W
T
V
Q
T
X
D
J
T
U
Z
K
R
Q
B
F
0
I
I
T
T
F
G
B
K
A
D
N
R
K
K
G
I
U
M
U
F
L
F
Y
B
F
S
L
I
M
R
H
U
Q
X
J
H
U
W
G
T
T
T
V
I
Z
N
Y
V
H
H
L
Y
S
G
L
T
R
K
I
U
V
N
W
A
I
V
D
R
M
T
L
A
T
T
0
I
H
L
P
S
X
X
E
G
T
U
N
N
0
G
F
T
A
G
P
A
G
X
S
E
A
P
J
N
T
Z
X
U
Y
J
I
T
Q
Y
V
R
N
S
J
S
N
Z
X
T
C
R
Z
C
I
J
Y
T
H
J
G
I
S
M
F
Z
R
T
F
V
U
L
0
I
R
0
X
L
V
Z
X
N
W
G
X
E
F
V
A F X X
J
J
VF U
X
L
U V Y N
K H D W Q
U Z G Y M
Z G G C B
U R I R T
L S T VK
Y C G JW
N F B G I
C G U U X
Y Z B I A
A I Q B H
L D U O A
X S
S R G C W
Z F D S Y
Q W JI V
JI X W Z
Y F S JW
F A E C N
U N V0 N
N U S F I
< 112 )
91
QN0 1 CDI XQ.J
3 I -CDCDUJ> -I ~ Z QL > CD
xCDUI O-UCCCQI CnUJ
I ZU L U ZO C I C D
N_l _JCDUQ>-<
>
->Ul j L
o
n
j-
m
in
o
Z
X S Y ZU . 3C I NX -^
x c o c n Qi n a > D H H oo
i s:o.uz)3u.z:a cu
o
z
XMUZ CEU. DQCD3 I S : in
r
LL>i-) >>ZnQ.ai-Z CXI
NO. 7
F
T
U
H
H X X X
G Z L
F K U
L T G
Q T
I Ul
H
I
I
U H S
G K Q
M H N D U
F L B T Z
D Y W A K
L N H Z O
S R G U U
Z A P D Ul
C Z G C U
B H U U G
K JLI N C
M T Y F U)
D G W W T
0 A I E
U U N T
H C A V
0 U Z Z
V K B T
0 0 N X
R A
A
M
X
G
X
S
Y
Q
F
C
X
D
Ul
M
J
0
T
Z
F
N
H
T
P
C
E
H
D
E
Q
X
Z
I
J
B
M
X
K
T
M
J
H
A P A Ul X
X
U
J
P T 0 I
Z
E
A W T
0 N
H P F Q
Z G L X
G G W V
F
Z
U
p
( 187 )
NO. a
I
o
S
A
J
F
G
U
K
G
Y
P
R
D
X
V
F
K
X
E
J
F
I
S
I
X
B
X
O
V
M
S
U
Y
I
0
X
T
C
S
X
J
T
P
E
K
Z
J
U
Q
H
K
H
O
G
K
G
S
R
I
L
V
U
G
V
M
T
X
P
T
T
A
R
I
Y
I
A
H
Y
Q
R
Z
X
P
T
D
B
K
R
0
G
B
V
C
Y
R
K
M
H
V
B
W
B
R
L
T
K
J
E
R
V
F
Z
L
L
Y
Y
N
E
A
V
J
X
F
H
G
Z
J
B
I
U
V
I
G
P
E
L
U
R
C
E
K
V
L
R
E
Ul
R
B
G
Q
S
Z
M
U
N
U
T
W
I
U
P
E
P
U
Y
N
R
I
A
J
S
Y
W
T
M
I
R
W
C
N
V
X
C
A
X
M
Z
Y
0
H
T
N
P
F
V
F
U
W
K
Z
P
J
Z
H
G
V
R
H
N
T
F
Y
B
J
F
N
Y
N
< 219 )
NO. 9
F
P
U
H
G
N
X
F
H
I
M
E
R
F
X
Q
X
Z
Ul
O
I
X
J
K
U
V
L
X
R
R
Ul
D
Z
S
K
I
Y
V
M
I
H
K
F
S
C
W
B
E
F
L
T
Ul
J
I
G
F
Ul
U
Y
Y
H
P
I
K
T
M
J
U)
T
A
U
C
M
U
R
J
H
H
R
Q
U
C
P
F
U
J
Z
V
T
U
Y
C
A
C
K
F
T
H
J
L
B
H
E
A
E
X
J
M
Ul
U)
0
K
Q
E
0
D
U
G
L
T
E
T
J
A
G
G
X
I
A
A
H
L
B
Q
M
L
E
Z
G
P
S
D
D
D
B
I
G
U
K
G
U
Z
U
( 153 )
93
i > c n <: M (.4
r o < i n -
x e n -i r N x
r 2: TJ x 3 x
3 Q 3J C x
o
ru
0)
CO
1> Z C N H x
o en H x N 33
x TI o "n H N
r "n "n A tn co
0 "D o M r o
en z r z TJ2:
uzo TI c m
3> en 33 n A <
co n 3 x en N
3> zro tn A c
H c en m 3 <
n z M x i> o
1 O 3 C X A
TJr r en N <
TI m 3 o en -
D x 3 r O <
N N n o i > "n
M 2: CD 3) T 3
< I> CO X
I X T j O
z
o
ru
03
ZTJl TiC<t -iDCQZ
x r c < X T j mt n 3< z
3< No c zt n < e n o c
V
< 88E )
9
3
X
H
0
1
W
f
3
1
I
X
D
3
3
M
5
1
Z
H
9
Z
0
W
I
H
H
M
i
1
0
I
0
1
I
f
n
a
z
8
d
3
A
3
a
i
i
N
9
d
H
A
a
i
i
n
3
d
W
a

A
X
i
y
3
M
9
0
e
z
s
1
y
s
3
1
M
y
n
i
0
a
M
z
X
s
z
A
d
1
0
y
9
y
i
A
X
3

n
3
S
S
3
3
A
a
0
a
M
A
D
N
D
9
9
9
M
S
W
A
W
N
3
A
n
n
d
A
M
a
H
H
M
H
M
A
X
9
0
0
f
M
S
I
W
N
S
3
d
0
1
N
8
d
d
n
A
A
A
9
X
Z
0
1
n
A
i
H
X
A
X
I
X
8
I
W
3
H
I
N
3
Z
N
n
n
n
d
X
d
X
1
9
I
y
d
M
Z
f
n
i
i
0
H
f
y
i
d
H
a
M
X
3
i
d
X
>!
1
N
n
A
d
0
a
M
0
n
n
z
3
d
1
a
a
X
d
y
z
z
n
a
M

9
o
i
n
i
a
3
S
a
9
n
n
d
d
d
8
X
i
0
i
A
I
X
d
3
A
d
S
i
y
0
d
z
r
d
i
H
d
M
a
A

a
0
s
s
X
a
i
d
w
i
i
9
A
n
M

i
r
z
8
Z
a
y
X
r
A
a
n
d
a
9
r
n
f
3
I
N
X
9
d
a
M
y
d
z
d
a
H
N
X
z
w
9
X
H
3
C
W
Z
y
a
d
i
0
y
w
9
a
X
X
X
1
I
w
A
W
H
A
n
1
a
i
y
i
i
H
3
r
3
N
ID
Z
y
A
i
0
d
a
o
x
s
3
ST ' ON
( S6T )
Z
X
9
a
H
X
0
N
f
A
W
M
1
A
a
w
d
H
3
d
W
A
W
f
O
9
X
f
A
I
M
W
a
1
H
y
i
r
y
3
S
"1
9
A
1
i
f
A
X
0
X
9
i
Z
M
X
n
s
X
X
n
s
s
i
n
a
i
H
M
d
A
1
9
d
3
d
S
n
0
X
3
9
A
9
N
H
d
1
a
X
3
d
r

d
A
3
9
f
r
H
1
n
i
9
a
y
a
w
d
X
X
9
N
>
W
M
n
A
f
z
I
n
d
3
3
a
9
I
a
M
i
A
a
n
z
A
d
M
1
9
a
A
a
0
N
i
a
z
n
n
i
z
X
s
9
a
s
a
s
3
Q
X
3
9
H
N
d
H
a
X
d
9
a
a
0
H
A
X
o
X
3
s
a
i
n
d
X
N
X
Z
a
d
d
d
i?T 'ON
X
d
s
H
H
H
r
w
d
f
X
a
9
N
H
I
X
0
A
M
A
y
X
H
z
w
n
N
3
N
M
1
Z
1
H
3
H
H
Z
M
Z
y
d
A
Z
8
a
0
n
n
w
y
H
3
s
i
A
f
d
N
X
9
I
A
f
A
Z
f
9
0
W
y
9
9
A
Z
n
A
a
d
0
y
Q
9
9
y
H
M
S
9
S
W
H
y
M
r
d
A
y
y
X
9
W
X
S
d
3
0
r
z
a
d
H
i
d
i
z
z
M
I
a
3
S
i
a
H
N
1
A
9
Z
A
1
1
W
A
A
X
1
0
a
a
a
a
n
w
n
0
d
N
9
y
d
i
a
i
d
3
d
9
n
n
d
r
9
a
a
X
M
1
H
A
1
0
d
a
N
X
d
f
0
3
S
a
i
9
1
9
X
f
"1
Z
3
A
S
A
a
a
a
X
H
r
9
i
r
9
3
M
A
d
n
3
s
0
M
X
d
0
D
9
a
8
a
z
1
X
w
9
W
3
W
H
A
d
1
d
A
y
i
i
w
H
A
M
W
a
9
1
a
3
9
0
a
d
N
1
n
3
M
Z
Q
Q
a
a
a
9
s
N
n
X
i
M
M
M
A
X
X
Z
A
1
N
y
3
y
d
y
X
3
i
H
d
N
X
1
S
1
H
9
N
A
a
A
d
d
A
9
d
n
Q
3
n
X
a
3
I
y
0
a
0
X
s
s
M
a
A
H
i
s
H
Z
0
1
~l
A
A
n
M
w
X
9
X
a
M
8
X
X
0
A
n
X
n
3
d
y
8
H
X
y
w
z
3
X
H
a
A
y
o
X
H
i
d
9
z
s
A
X
A
d
X
y
9
w
s
Q
1
0
X
d
r
X
A
i
A
a
3
a
n
n
i
3
9
a
o
z
i
X
a
A
i
X
w
H
H
E T 'ON
BOOKS IN THE CRYPTOGRAPHIC SERIES
1 - Manual for the Solution of Military Ciphers, Parker Hitt
2 - Cryptanalysis of the Simple Substitution Cipher with Word Divisions, Wayne G. Barker
4 - Statistical Methods in Cryptanalysis, Solomon Kullback, Ph.D.
5 - Cryptography and Cryptanalysis Articles, Vol. 1, edited by W.F. Friedman
6 - Cryptography and Cryptanalysis Articles, Vol. 2, edited by W.F. Friedman
7 - Elementary Military Cryptography, William F. Friedman
11 - Solving German Codes in World War I, William F. Friedman
13 - The Zimmerman Telegram of January 16, 1717, and its Cryptographic Background,
William F. Friedman & Charles J. Mendelsohn, Ph.D.
17 - Cryptanalysis of the Hagelin Cryptograph, W ayne G. Barker
18 - The Contributions of the Cryptographic Bureaus in the World War [World W ar I],
Yves Gylden
22 - History of Codes and Ciphers in the U.S. During the Period Between the World Wars,
Part I. 1919-1929, ed. Barker
29 - Cryptanalytic Programs for the IBM PC, incl. diskette, C.A. Deavours
35 - The Origin and Development of the National Security Agency, George A. Brownell
36 - Treatise on Cryptography, Andre Lange and E.A. Soudar
39 - Cryptanalysis of Shift-Register Generated Stream Cipher Systems, W ayne G. Barker
40 - Military Cryptanalysis, Part II, William F. Friedman
41 - Elementary Course in Probability for the Cryptanalyst, [Rev. Ed.], Andrew M. Gleason
42 - Military Cryptanalytics, Part I, Vol. 1, W.F. Friedman and L.D. Callimahos
43 - Military Cryptanalytics, Part I, Vol. 2, W.F. Friedman and L.D. Callimahos
44 - Military Cryptanalytics, Part II, Vol. 1, L.D. Callimahos and W .F. Friedman
45 - Military Cryptanalytics, Part II, Vol. 2, L.D. Callimahos and W.F. Friedman
46 - Pattern Words: Three-Letters to Eight-Letters in Length, Sheila Carlisle
47 - Cryptology and the Personal Computer with Programming in Basic, Karl Andreassen
48 - Pattern Words: Nine-Letters in Length, Sheila Carlisle
49 - The Index of Coincidence and its Applications in Cryptanalysis, William F. Friedman
50 - Cryptographic Significance of the Knapsack Problem, Luke J. O'Connor and Jennifer Seberr)
51 - Breakthrough '32, The Polish Solution of the Enigma, with diskette for IBM PC,
C.A. Deavours
52 - The American Black Chamber, Herbert O. Yardley
53 - Traffic Analysis and the Zendian Problem, L.D. Callimahos
54 - History of Codes and Ciphers in the U.S. During the Period Between the World Wars,
Part II, 1930-1939, ed. Barker
AEGEAN PARK PRESS, P.O. Box 2837, Laguna Hills, California 92654-0837