TOP SECRET.
ULTRA
RIP ]7}
Subtract the enciphered code group ( in the given position) from three or four other groups in the s&rfl.e olumn, choosing the groups from messages c which have sholvn libtle or no tendency to garble. Add these differences successively to the cJear cod.e groups, one of which you suspect may be presentt and where potential cod.e groups resuft from these addition.s- nroceed as in the known-word nethod.. say Thus the parbicul-ar cJear code group for, I/l/S (r?Begin i-rss&g Hererr) may be seLected. from among the tvrenty-odd. clear code groups rvhich st:nd for this in texb val-ue by obser.ving whether its substitution the given positlon would. produce potential code groups in several other places in the colunn. d. IllusbraLions of Additive Recovrf,r page is found. a sample porbion 0n the following of a work sheet with one vacant col-umn. The basio cod.e of the traffic in this example 1s a scanning cod.e" 2
I -l r
I+
----
Ad.d.itive E n c i p h e r e d g ro u p Cod.e group Pl-ain text
ofn
vuvt
22435 06009
a / a a a
irr
I
I I
i3735t+
$7L3 53LI5
80t+69 M/S
6gzlS
2 B 6z 87597 Begin' Nf
?843t+
Kana
\TT
va l u e
9285r
r_4zoo
7475L
00 t+265t+ Roman R
Roman t+5 0 8 5 67 t+LO H,TISU 8985t+
ot+668 Roman
88702
r77o 4
l+t+o 58 X Unit
479)7 ](sHINSHOO
18934 77L69
322L t+2456
tTn
7o5LB
929b3
Rpt A
78330
70723
07a77
Num
0463lx
808
8Ir4
t+5t+68
I
R,AAA
v v v J _
37L98
v-A-6
ORIGINAL
TOP SECRXT-ULTzu.
RIP ]7I
N o te th at 0463I ( colum n l) is a gar bled gr o up and t h e . t n o r r e a n i - n gi.s y e t r e c o v e r e d . . f o r 5 1 6 1 8 . ( c o l u r a n ) ) , 1)
Recover y by the known- wor d nethod, In l i ne I of the se.rnple sheet, the Japanese work SHfNSTiU is str <,ng1y to be expected. The cLea.r c ode gr oup for ttr is text ve,fue ( in Code !) l s 58131+ The trie.I e,dditive 1s, therefore, lSllf,rilnua $7I3 or L5l+2L. +pplied to the 6olur,n, tne followlng ser les r .r f cc.r de oups is c btal ned gr with their nr eanings: 58l-3t+ 89l-72 9t+275 SIIINSHUTSU Roman E ArlT Foll 2 Groups
8)75r o3787
#t4z
7L466
( Garble ) NIJTYOO
These meenings, tluo of which ar e v ex y hl gh frequency, fit the context. exactly ln the severr messages. There is thus very high probability that the e.dditive is L5421,
The gar ble m ay be disr egar cled in v i r tue of t h e e x c e l l e n t t e x t o b t a i n e d . f n f e . c t, 0 ) 7 8 7 m a be consider ed. e gar ble for 0j777 who s e tex t v el is itParen I Groupr't particulerrly since O46jI, ( tne pr eceding Foll gr oup) ungc,ibles to 04632. YOBID+,SHI FUG00. Unlarbling requires a very s kill- fu} he.nd but in thls..ci..se the pr obabi l i ti es the ungar bling is eccur ate ar e v er y hi gh, !l.t The sequence i", indeed, so strong that entiy ir to the colurr:r nright have been l,raclvlr., this grc,u The pr ocedur e would be as f ollows: H av i n! i decid.ed the t Ot+6jI we.s a ga,rble for 046)2, aid having found. thr ^t this r iiust.lilcely tex t v al ue, I' Par en Foll- I gr oup" ( 0) 777) , gave, when s ub_ stituted, a pr eponder ence of ge.r bles , the w or k er wouLd exari.ine the col-uLlLn ihu:s obteined. The tr ie.l- a^dt,itive wc.r uldbe O) 77/ r .inus 9S) 66r or 15411.
. 1 :
lf
t.
a-
ORTGTM,L
TOP SECRET-ULTzu. 58121+ 89l.62 g t+2 6 5 garble ) ger ble ) gerble ) ( ger ble ) 1 in per en ( gar ble )
RJr'17r
837+t o)777 6r466
l ti th the aid of the Rlpr he would next examine these values to see whether a series of cl e a r co d e gr oups could be found. ee,ch of which differed. from its garble by an id.enticel difference. This difference ls eesily seen to be one in the fourth digit. Therefore, the initial trieil- additive should be lncreased by one in the fourth dlgit to read I5l+2I which, when applied, g i ve s th e column r ecover ed by the known- wor d method. using SHII,ISI{UTSU. * In b o th the eval,uation and r ecovefy of add,i tives, the ability to deal with garbles is extre rrre l y i rnpor te.nt. This epplies not nr er ely to the probe.bi-tity', si.X, trie.t 49090 is e garble for 4 9 0 8 0 w h e re e D00 0NI.JI is expected., It is frequently possiblo, as here, to go into a colu n n o r a ge.r bled gr oup, by obser ving whether th e su b stl tution of the clear cod.e gr oup pr od u ce s a col- unr n uhich gar bles conslstentLy r throughout. l.n expert in additlve recovery is a b l e , th rr.iugh the le"r ge nunber of goocl code g ro u p s l rti tii which he is fer nil_ic.r , to exteno the sco p e <-,fal.l- thr ee of the ba.sic nethoos. S e ve ref other points of entr ;z r i.ight heve b e e n u se d .. Lir ie; h is e. por tion of e SHO0GOIC H f n e s s g e . T h e s e r i e s o f v e i l - u e s( p r l o r t o t h e s a i i p l e g i v e n a , b o v e w n i c h er e r s i o c i c t e c i w i t h )
t?Repee t Art e.re :
Repeat l,
Repee ./. #L35 Nurrr.sep. t Repeert #I38 Nun. scp. t#LI+O Num. sep.
#n tl>
l.r0
T h e v a " l u es l f t t + z a n d . i l i 3 c o u t c ' r . ,t h e r e f o r e , be tried in the spee l,rhere encipirered group is located, e.nd #t{Z wc,uld have brought 783)O out the colurn. l.gain, the ce.lI NIJRK is E. very cuf_rironone elnd the col-urnn r,right he.ve been entered via Ronnn b, V-A- 8
0RIGIN/J,
TOP SECRET-ULTR,. 2) method woul-d be ar)p1ied to The difference we.Vo (In the excc,lur.n 2 in the following emple as given, this fliethod vrculd be quite unnecessary since sc mtny v{i.ys uf entering by the known-vuord nethod exist e.nd since the In technique is vr.stly tticre ra.pid. Iatter method. is not the difference actual practice, used to obte.irt an adclitive adjacent ordinarily to e.n r,lready recovered coluian, but is used to break into pages or portions c,f pags where no Ilowever, other additives have been recovered. e.ll xe.rrrpl of for tire purposes of afforoiiig the di-fference ft,etnod., assume that the r.dciitives for adje.cent colunns he.ve nct yet been r e c u v e r e d , r r l c i t h e " s b e e n c l e c l c L e dt o d i f f e r ence colu.riin 2 es . &e-I1s uf bre;.i;i-itg into the \ r ,l
Pc.6E
ffi#
jits
the loca,tion of the As L general rule, hrgh frequency groups is nct knov,rn. In such ciiscs, it is necess.ry to cbtein a.11 of the d , i f f e r e n c e s i n t h e c c l u l n n e . n c rt o c o x t p e r e in the dif these l'rith the diff erences Iistcd dc thrs, any deIn crder tc ference tt,ble. sire d, group is zeroized (thrt is ec.ch of its f ive digits is subtracted f rr-,ii zero ) ernd ad.ded to the cther groups i-n the col-unn. Thi-s proto subtractcedure is, of course, equivilent Zeroizing ing the given gr( up frcm i:11 c;thers. t]'re top enciphcred grcup in the cclunn, the differences i re: Difference s l10l+8 46141
))o.
-./r^ ^. / ^n
Corresp itdrng cc,de grvup
tc )4Y
57r47
7919t+ ) 9) 36 i .nd 61+827
ati
>)o) ) 2)3t*2
T i i e t r i t " l - r d d i t i v e s i . , L r c, , b t ; . i n e d b y i . d d i n g the cLoe groups t,' the zet.:-ze i vr.Iue. Thus tii f rrst trri.l r.dclitive s . re 67397 plus 26)49, 67)97 plus 57Il+7, etc. IJcnc r.'i tire additive s f r c r , , t l i i s s e t p r o v e s g o c , J , s , . t h e s e c c , n c lc o d e grLup is zyToLzed.inct i.dde,i t-, Lrte grLups belolv s It is unlr.ece sLry tc, i.cjcit,rIt- v i.lue t..; it. s 437L3 (the group t-lrei.ciy zeroized.) becr.use L + ) 7 L 3 h a s e . I r e i u y b e e t i e l r r , i r n a , t e C r. . s t h c e n crpherect vi lua f cr iny higli frecllteiicy group vrhrch lve.sused in r:iaking the drff erence table, p r r , v i d c d , ' f c eu r s , t i , i t t i r e c o l u t , , n c o n t r . i n s
i rry v + r l. c 1F' Lh
. i jr . J - d h] I g + q! r^r I ' n C v ^ rU1e l n ( - r) y . ^.r 8I . u-p.
V-",-9
ORIGIN/,L
TOF SLCRJT-ULTR.,
RrP ul
Since this is the r ssurrption crn whj_ch the d i f f e r e n c e m et h < . d c p e r r t e s , t h e g r c u p i l r e e d y tried. is regulirly excluded froi,r f urther con_ siderition, t.dding 67t9? tc the cther enciphered ccde groups, the fcllolving cc.Lumn results:
fl
Difference s I 5103 04689 246L5 82)04 Under t-510j in the difference tr.ble appear _ 8 9 1 7 2 e .n d 01488, 0i. a p:r ge clevr td i r f p e c o v e r i e s , there is no way to tell-ii a c l v a n c el v h i c h c , f t h e o o d e v a l u e s i s n o r e p r o b a b l e , ? h e wc,rker nright s e try 01486 flrst, , f l dw c u l c i o b t a i n t h e f o 1 l c - r w i n g col-u:,.::: *.dditive; )77)7
7Ol+l+0 l_t_rne_r^ning; frequency nc 01488 SHIi\iKruU 165Sr #z
Oj o 67 25093 83782
no r i,elning j RC,fr equency ( g e . r b t e) ( g er b l e )
T h l s c clur :n ]ia.s fuUr . spinning gr cups e.nd t w o g a r b J - e . r t h i . s .f r e q u e n c y g r c u p s s onry in the cese o f t h e c n e s r r h i c h w e i ^ eu = b d * t o b u t . . i r r " t r r e differ_ enc:, Hence, the edditive )T7jZ is-nui the cor_ r99t one. using 89172, hcv/cver, the correct a d d i t i v e l - 5 t + Z Ii s i n r . , e d 1 e . t e l _c b t a i n e d . y . T !" c(,r iplete dif f er ence cf, the cel_uiiin( whi c h a ctu e l l y c onte ins cnly t,,^r c gh h: gr oup s ) i-s i s follclvs: "r ;q;;;;y 11048
l+6Lt+I
)5627 0l+689 5 5 6 5 3 1 4 OL '
L)3t+2
I 5+03
82304
99586 19512
/ /zvr
20a36 88725
68lgg
V-.r.-10
0Rf Gf Nr.L
TOP SECRET-I-,ILTfu,
RrP 171
If thc differencing he.d been carried L,ut f r o r , t h e b et t o r i i v / u r k i - t r g u p , e : d i s t i n c t but equj-velent set i,f dil'ferences rrvuulrl he.ve beerr obtE"inect nalrrely: ,
97768 28706 )3809 22)85 4?)LL
5 5 4 5 7 7 548) 86t+95 o 6 4 z t 9L598 11 \)), 80074
6t*g6g 79062 95907
by which entry 95947 is the other difference intc, the colunn ct-,utd have been effected since it indexcs 9l+?75, ,,t,/T Fotl 2 Gr.cups.'r O n e p c i n t r , , r i : i c hi s e s p e c r r 1 1 y l , c r t h y c " t note rni.y nolr be illustre ted. Suppcse the colu r : r l i n q u e s t i u n c o n t a i n e d n o t t v , , i ob u t t h r e e o r n : 1 c r ch i g h f r e q u e n c y c l e i r c r d e g r - u p s . The s u b t : : e c t i c n s t v c , u l d h ev e p r o d u c e d c l i f f e r e n c e s , tite ccrrespcnding clei r cr cie grc urrs f cr whi ch wrr..1d heve pri-,ciuced the sarrtegri;11p several tines. Thus if r-o B. C, cnd D cre clear cod.e groups usc,d in c(nstructing tire difference tc.ble , and if (l^-X) , (B-X) , etc. are the c rrLsponding ericiphered c.Ce 6r. ups in some c u l u r - n L . n t h e l v t l ' k s h et t , t h e n t h e d . i f f e r . e n c e s enu thc cLr.resp-nding rr^rnrr gr.ups (indexed f ror,r the di f f erenc e ta ble ) are : Difference s \ . . . - , -- H - A (r'--t) - C-.r"
L \ trv\ rl ,/ r n U.tL I 1r \ t r,-,\ I
CIec.r Cocle Grc-ups frorri Table
t, A
The gruup r, i,ppars three ti.i,-es. In pre:ctic,, then if r: set c,f differences generited b y o n e z e r Q i z i n g r , f t h e c c , tu _ r , i p r o c u c e s d u p l i _ i c a t i o n , t r i p l i c a t i t n , et c . , , . ; f c : c l e i , r c o d . e g r u u p ' , v h e nr e f e r e n c e i s r , \ , . a 1t.la t h e d i f f e r e n c e table, c.nsidereble weight is givcn t. the i . s s u r , . L p t i - nt n i t the repel.te-1 clc:.r c,cLe group di,6.s, in f ect, underlie the zerctzed. encipheicO c.Ce grLup.
v-A-11
0RIGfl'ir.L
