Spring 2015

Chris Christensen
CSS/MAT 483

Columnar Transposition

Classically ciphers that rearranged the letters of plaintext were called transposition
ciphers. They can be recognized because ciphertext letter frequencies are the same
as plaintext letter frequencies.

Columnar transposition is probably the most commonly studied transposition

cipher. We will use that method to encrypt the following "pilot's saying:"

The nose is pointing down and the houses are getting bigger.

There are 49 letters in the message. We want to place the letters of the message in
a rectangular array. In this case, because we would like the rectangular array to
have 49 cells, a 7 × 7 array may be used. We also need a keyword having its
length the same as the number of columns – we will use analyst.

A N A L Y S T
1 4 2 3 7 5 6
t h e n o s e
i s p o i n t
i n g d o w n
a n d t h e h
o u s e s a r
e g e t t i n
g b i g g e r

The ciphertext is obtained by reading down the columns in the order of the
numbered columns (which are alphabetically ordered).

TIIAOEGEPGDSEINODTETGHSNNUGBSNWEAIEETNHRNROIOHSTG
Our message exactly fit the rectangular array. If the message does not completely
fill the array, nulls (i.e., padding) may be added to fill the array (this is the easier
cipher to break) or not (this is harder to break because the columns do not all have
the same length). In the latter case, the length of the keyword determines the
number of columns, and the number of letters in the message determines the
number of complete and partial rows.

The transposition should be applied several times if the plaintext message were
longer than 49 letters.

Remember, for encrypting, “in by rows and out by columns.”

Decrypting the columnar transposition

Here is a message that was encrypted using a rectangular array with keyword
analyst.

TRLEELIGCIGEHALANTNCTECYENEN

Because the keyword has 7 letters, we know that the rectangular array has 7
columns. The message has 28 letters; therefore, the array must be 4 × 7 . Each
column must have 4 entries.

First, we place the letters of the keyword in alphabetical order: aalnsty. Then place
the ciphertext letters in columns.

A A L N S T Y
t e c h n t e
r l i a t e n
l i g l n c e
e g e a c y n
Now rearrange the letters of the keyword to form analyst.

A N A L Y S T
t h e c e n t
r a l i n t e
l l i g e n c
e a g e n c y

The plaintext message is the central intelligence agency. (Notice that

there could be some ambiguity about which "A" column comes first. We have
used the convention that the first "A" column will correspond to the first a in
analyst.)

Remember, for decrypting, “in by columns and out by rows.”

Cryptanalysis of the columnar transposition

We will do only "the easy case;" i.e., we will assume that the columnar
transposition uses a rectangular array that was completely filled.

Here is the ciphertext:

ASAIR ITFNM IMTKL SOIEE M

The “key” to cryptanalyzing the ciphertext is to determine the number of columns;

i.e., the length of the keyword. There are 21 letters in the ciphertext. Because we
know that the message completely fills the rectangle, this suggests either a 3 × 7 or
a 7 × 3 array.
We arrange the ciphertext in columns.

A F L
S N S
A I T M T S E A M O
Either S R F I K O E or I I I .
A I N M L I M R M E
I T E
T K M

The solution is by anagramming (making a word or portion(s) of word(s) by

rearranging letters) a row.

The 7 × 3 arrangement seems unlikely because it has a string TKM with no vowels
that is unlikely. Also, the III is unlikely. So, let us try the 3 × 7 arrangement.
Notice that there are 7! = 5040 arrangements of the columns. We would like to not
have to try all of them!

A I T M T S E
S R F I K O E
A I N M L I M

In the first row, MATE seems to leap out. This leaves ITS. Perhaps, a slightly
wrong guess – ESTIMAT- seems to be a possibility.

Let us rearrange the columns.

E S T I M A T
E O K R I S F
M I L I M A N
Not quite, but there are two Ts in the first row. Let us swap those columns.

E S T I M A T
E O F R I S K
M I N I M A L

This works. Notice that because we have multiple rows that are permuted the same
way, we can use multiple anagramming for cryptanalysis.

It is often worthwhile to write the ciphertext in columns, cut out the columns, and
rearrange the columns to do the anagramming.

Determining the dimension of the rectangle

Frequencies can help to determine the dimensions of the rectangle. In English

approximately 40% of plaintext consists of vowels. Therefore, for the correct
dimension, each row of the rectangle should be approximately 40% vowels.
Consider our choice between 3 × 7 and 7 × 3 .

For a 3 × 7 rectangle, each row should contain approximately 2.8 vowels. Let us
note the difference between this estimate and the actual count:

Number of vowels Difference

A I T M T S E 3 0.2
S R F I K O E 3 0.2
A I N M L I M 3 0.2

The sum of the differences is 0.6.

For a 7 × 3 rectangle:

Number of vowels Difference

A F L 1 0.2
S N S 0 1.2
A M O 2 0.8
I I I 3 1.8
R M E 1 0.2
I T E 2 0.8
T K M 0 1.2

The sum of the differences is 6.2. It appears that the 3 × 7 rectangle is more likely.

Using digraph frequencies to arrange the columns

Digraph frequencies can be used to help in the cryptanalysis in place of just

looking for reasonable pairings of the columns. For example, consider our
ciphertext above ASAIR ITFNM IMTKL SOIEE M. Again, we’ll assume that a
3 × 7 rectangle is appropriate.

A I T M T S E
S R F I K O E
A I N M L I M
We will pair the first column with each of the other columns on the right and
consider how likely it is that such digraphs will occur in English. The frequencies
we will use come from Sinkov (see the appendix). Recall that there are
26 × 26=676 digraph frequencies.
AI 311 AT 1019 AM 182 AT 1019 AS 648 AE 13
SR 9 SF 8 SI 390 SK 30 SO 234 SE 595
AI 311 AN 1216 AM 182 AL 681 AI 311 AM 182
631 2243 754 1730 1193 790

The most likely pairing is

AT
SF
AN

Oops! We know that this is not the correct pairing, but the second most likely
pairing is correct. (During cryptanalysis, we don’t always get the correct result on
the first try.)

Once we have a pairing, we could then continue using digraph frequencies to select
columns to add on the left and on the right. Etc.
 More columnar transposition

It would be harder to do the cryptanalysis if the rectangle were not completely

filled. For example, let’s use a columnar transposition with keyword norse to
encrypt the message Germany seeks an alliance. The message contains
22 letters; so, we need 4 complete rows and one partial row.

In by rows:

n o r s e
g e r m a
n y s e e
k s a n a
l l i a n
c e

Out by columns:

AEANG NKLCE YSLER SAIME NA

Because the columns do not have the same length, this would not be as easy to
cryptanalyze. It would not be obvious how many columns were used. (The size of
the rectangle would be either 2 ×11 or 11× 2 if we knew that a full rectangle had
been used; i.e., the keyword would have length either 11 or 2.)

However, if we know the keyword, decrypting is no problem. Try decrypting the

ciphertext

IMYRA CBILM AANIE NSBNR ESE

which was encrypted with columnar transposition with keyword norse.