CHAPTER 5

MONOALPHABETIC MULTILITERAL
SUBSTITUTION SYSTEMS




Section I
Characteristics and Types

Characteristics of Multilateral Systems
5-1.
As explained in Chapter 3, monoalphabetic unilateral systems are those in which the
ciphertext unit is always one character long. Multilateral systems are those in which
the ciphertext unit is more than one character in length. The ciphertext characters
may be letters, numbers, or special characters.

a. Security of Multilateral Systems. By using more than one character of ciphertext
for each character of plaintext, encipherment is no longer limited to the same num-
ber of different cipher units as there are plaintext units. Although there is still only
one alphabet used in multilateral systems, the alphabet can have more than one
ciphertext value for each plaintext value. These variant ciphertext values provide
increased security. Additionally, the plaintext component of alphabets can be
expanded easily to include numbers, punctuation, and common syllables as well as
the basic 26 letters. When used, the variation in encipherment and the reduced
spelling of numbers, punctuation, and common syllables minimize the exact
weaknesses that we used in Chapter 4 to break into unilateral systems.
b. Advantages and Disadvantages. The increased security possible with variant
multilateral systems is the major advantage. The major disadvantage is that by
substituting more than one character of ciphertext for each plaintext value, the
length of messages and resulting transmission times are increased. A second disad-
vantage is that more training and discipline are required to take advantage of the
increased security. If training and discipline are inadequate, the security advan-
tages are lost easily.



5-0
Types of Multilateral Systems
5-2.
Multiliteral systems are further categorized by the type of substitution used. The
major types are”
Biliteral systems, which replace each plaintext value with two letters of ciphertext.
Dinomic systems, which replace each plaintext value with two numbers of cipher-
text.
Trilateral and trinomic systems, which replace each plaintext value with three
letters or numbers of ciphertext.
Monome-dinome systems, which replace plaintext values with one number for some
values and two numbers for other values.
Biliteral with variants and dinomic with variants systems, which provide more than
one ciphertext value for each plaintext value.
Syllabary squares, which may be biliteral or dinomic, and which include syllables as
well as single characters as plaintext values.




Cryptography of Multilateral Systems
5-3.
The cryptography of each type of multilateral system, including some of the odd varia-
tions is illustrated in the following paragraphs. Most of these systems are coordinate
matrix systems in which the plaintext values are found inside a rectangular matrix and
the ciphertext values consist of the row and column coordinates of the matrix.

a. Simple Biliterals and Dinomics. The simplest multilateral systems use no varia-
tion. They typically use a small rectangular matrix large enough to contain the
letters of the alphabet and any other characters the system designer wants to use as
plaintext values.
(1) The plaintext values are the internals of the matrix. They may be entered
alphabetically, follow a systematic sequence, or they may be random. They
may be entered in rows, in columns, or by any other route.

(2) The row and column coordinates are the externals. Conventionally, the row
coordinates are placed at the left outside the matrix, and the column coor-
dinates are placed at the top. As with the internals, the coordinates may be
selected randomly or produced systematically.
(3) A ciphertext value is created by finding the plaintext value inside the matrix
and then combining the coordinate of the row with the coordinate of the column
for that plaintext value. Either can be placed first, although placing the row
coordinate before the column coordinate is more common.



5-1
(4) Five by five is a common size for a simple system (Figure 5-1). The 26 letters are
fitted into the 25 positions in the matrix by combining two letters. The usual
combinations are I and J or U and V. It is up to the deciphering cryptographer to
determine which of the two is the correct value. There are few, if any, words in
common usage in which good words can be formed using either letter of the I/J
or U/V combinations. Other common sizes are 6 by 6 (which gives room for the
10 digits), 4 by 7, and 3 by 10. Many other sizes are possible.




(5) Example A in Figure 5-1 is a simple 5 by 5 matrix with I and J in the same plain-
text cell of the square. The coordinates and the sequence within are in
alphabetic order.



5-2
(6) Example B is a simple 3 by 10 matrix with orderly coordinates and a keyword
mixed sequence inscribed within. The four extra cells are used for punctuation
marks.

(7) Example C is a 6 by 6 matrix with a spiral alphabetic sequence followed in the
spiral with the 10 digits. The coordinates in this case are related words.

(8) Example D is a 5 by 5 matrix with numeric coordinates. The plaintext sequence
is keyword mixed entered diagonally. In this case, there is deliberately no
repetition between the row and column coordinates. This allows the coordinates
to be read either in row-column order or in column-row order without any
ambiguity, as in the sample enciphered text. This is unusual, but you should be
alert to such possibilities.

b. Triliterals and Trinomics. Trilateral and trinomic systems are essentially the
same as biliteral and dinomic systems. The difference is that either the row coor-
dinates or the column coordinates consist of two characters instead of one, creating
a three-for-one substitution. Such systems offer no real advantage except to provide
a slightly different challenge to the cryptanalyst, and have the distinct disadvan-
tage of tripling the length of messages. They are easily recognized, and offer no
increase in security.




c. Monome-Dinomes. Monome-dinomes are coordinate matrix systems constructed
so that one row has no coordinate. The values from that row are enciphered with the
column coordinate only. This means that some ciphertext values are two characters
in length (dinomes) and others are only one (monomes). If the values used as row



5-3
coordinates are also used as column coordinates, no plaintext values are placed in
the monome row under those repeated column coordinates. The blanking of cells in
the monome row is shown in the example below.




Resulting message:
25720 67463 63485 69575 40000



(1) If the cells corresponding to the row coordinates in the monome row are not
blanked, the deciphering cryptographer will have difficulty. Decipherment
proceeds left to right, and when a 5 or a 6 is encountered in the matrix shown, it
will always be a row coordinate or combine with a preceding row coordinate. It
will never stand alone as a monome. If the 5 and 6 cells were not blanked, the
deciphering cryptographer could not tell if a 5 or 6 were a monome or the begin-
ning of a dinome. The cryptographer would have to rely on context to figure out
which was intended, and that could lead to errors.

(2) The additional examples of monome-dinomes shown below demonstrate the
various ways they can be constructed. The last example (top of page 5-5) is a
monome-dinome-trinome.




5-4
Resulting message:
31323 12331 3023271318 90000
d. Variant Systems. Variants in a multiliteral system allow plaintext characters to
be enciphered in more than one way. Variants can be external or internal.
(1) External variant systems have a choice of coordinates. Either row coordinates
or column coordinates or both can have variants. Examples A and B in
Figure 5-2 provide two ways to encipher every letter.




5-5
Example C provides four ways to encipher every letter. Example D was con-
structed to provide the most variants for the most common letters. The letters
E, T, and O can all be enciphered in eight different ways. R, N, and I can be en-
ciphered in six different ways. A, S, D, L, U, H, and M can be enciphered in
four different ways. Q, X, Z, and the comma can only be enciphered one way.
When any of the systems are conscientiously used, repeated words in the text
will not produce repeated ciphertext segments.

(2) Internal variant systems use larger matrices to provide variants inside the
matrix. Each common plaintext letter appears more than once. Here are two
examples of internal variant systems.




The first example above places the letters in the matrix according to their
expected frequency in plaintext. If their use is well balanced, all letters in the
square will be used with about the same frequency. The second square achieves
the same effect by using 10 words or phrases in the rows, which use all the
letters. The first letters of the column spell out an eleventh word”logarithms.




5-6
e. Syllabary Squares. Another type of internal variant system is the syllabary
square. This type includes common syllables as well as single letters. When these
are used, the same square may be used for a period, changing the coordinates more
frequently than the square itself.




The two sample encipherments of REINFORCEMENTS show that a syllabary
square suppresses repeats in ciphertext just as single letter variant systems do. It
also has the advantage of producing shorter text than single letter multilateral
systems.
f. Sum Checks. It is very easy for errors to occur when messages are transmitted and
received, whatever means of transmission are used. Because of this, some users
introduce an error detection feature into traffic known as sum checking.

(1) In its simplest form, a sum-check digit is added to every pair of digits in numeric
messages. The digit is produced by adding the pair of digits to produce the



5-7
third. If the result is larger than 9, only the second digit is used, dropping the
10™s digit, for example 8 plus 9 equals 7 instead of 17. This is also known as
modulo 10 arithmetic.




(2) Whenever the first two digits do not add up to the third, the receiving cryp-
tographer is alerted that an error has occurred. The cryptographer then tries to
figure out the correct digit from context or by assuming that two of the digits are
correct and determining what the third should be.

(3) There are many variations on the simple system of sum checking described
here. Sometimes the sum-check digit will be placed first or second in each
resulting group of three. Sometimes a sum check will be applied to a larger
group than two numbers. Sometimes a different rule of arithmetic will be used,
such as adding the sum-check digit so that the resulting three always add to the
same total. Sometimes a more complex system will be used that provides
enough information to resolve many errors as well as detect them, particularly
when computers are used in data and text transmissions.

(4) Computer produced sum checks can be used with any characters, not just num-
bers. Computer produced sum checks will normally be invisible to the user, as
they are automatically stripped out when a message is received. They may or
may not be invisible to the cryptanalyst. Recovery of computer produced sum
checks is well beyond the scope of this text, but you should be alert to their
existence.



Section II
Analysis of Simple Multilateral Systems

Techniques of Analysis
5-4.
The first steps in solving any multilateral system are to identify the system and
establish the coordinates. It makes little difference whether the system uses numbers
or letters for coordinates. The techniques are the same in either case. Once the system
is identified and the coordinates set up, a solution of the simpler systems is the same as
with unilateral systems. Variant systems require additional steps. Each type is con-
sidered in the following paragraphs.



5-8
Identification of Simple Biliteral and Dinomic
5-5.
Systems
Simple biliteral and dinomic systems are very easy to recognize and solve.

a. First, the two-for-one nature of the system will usually be apparent. The message
will be even in length. The majority of repeated segments will be even in length,
although when an adjacent row or column coordinate is the same, a repeat may
appear odd in length. The distance between repeats, counted from the first letter of
one to the first letter of the next, will be even in length.

b. Second, unless the identical letters or numbers are used for row and column coor-
dinates, there will be limitation by position. One set will appear in the row coor-
dinate position, and the other set will appear in the column coordinate position.
Even in the case where all coordinates are different and either the row or column
coordinate character may be placed first, each pair will be limited to one from one
set and one from the other. If you do not recognize it right away, charting contacts
will make it obvious.
c. For systems with letters as coordinates, not more than half the alphabet will be used
as coordinates. This severe limitation in letters used is the most obvious charac-
teristic, since only very short unilateral messages are ever that limited. A phi index
of coincidence will reflect that limitation, always appearing much higher than
expected for a unilateral system.

d. Dinomic systems, since they are limited to the 10 digits anyway, are not quite as
obvious. Simple systems should still show positional limitation, however.




Sample Solution of a Dinomic System
5-6.
The next problem shows the steps in solution of a sample dinomic system. These steps
apply equally to biliteral systems.




5-9
a. The most obvious thing about this cryptogram is that every pair of numbers begins
with 2, 4, 6, or 8. The final pair begins with 0, but since it appears nowhere else, it is
probably a filler. This suggests that we are dealing with a matrix with four rows.

b. Scanning the second digit of every pair, we see that there is some limitation in the
column position, also. All digits are used except 8. The matrix appears to have nine
columns, although it is possible that a column for 8 exists, but no values from it
were used. Four by nine is a reasonable size for a matrix.

c. Next, we check for repeats and underline them. We also prepare a dinomic
frequency count by setting up a 4 by 9 matrix and checking off each dinome that
appears.




d. The two longer repeats both include patterns of repeated values. Word patterns can
be constructed on repeated dinomes just as they were for repeated single letters.
The word patterns for the two longer repeats are shown below.




5-10
e. The word pattern lists in Appendix D show only one possibility for each pattern as
shown. The two are consistent with each other. Using these recoveries, we can set up
a matrix and place the values in it and the cryptogram.




f. The plaintext words ENEMY and AIRSTRIKE are now obvious. Placing the M
from ENEMY shows COMMANDING at the end of the message. Most of the
remaining plaintext letters are easily recovered.



5-11
g. The letters in the second row precede all the letters in the third row alphabetically.
This suggests an alphabetic structure, although the columns are clearly not in the
correct order. The first row probably contains a keyword. If we rearrange the
columns so the letters in the second and third rows fall in alphabetical order, we see
the next structure.




h. The plaintext letters area keyword mixed sequence based on INCOME TAX. After
placing the remaining letters, there are still 10 blank cells in the matrix. Seven of
them are used in the cryptogram, and they cluster together in segments of three or
four dinomes. They show the typical pattern of numbers. In particular, the four



5-12
plaintext values of groups 50 and 51 of the message indicate time, and 66 is
probably a 0. More likely than not, the remaining numbers fill the bottom row of
the matrix in numerical order, but these recoveries cannot be confirmed without
more information. If hill numbers could be compared to known numbers from an
enemy map sheet, we could accept the values with more confidence. At this point,
we are reasonably confident of the letter arrangement and the number 0, but the
remaining numbers are only a possibility. However, if this were a current real life
situation and the enemy referred to by the text is our own forces, we would certainly
consider reporting the likelihood of air strikes on our artillery positions.


Analysis of Monome-Dinome Systems
5-7.
The characteristics of biliteral and dinomic systems that stand out most are the
divisibility by two and the positional limitation that makes it easy to determine matrix
coordinates. By changing the length of the plaintext unit from character to character,
monome-dinome systems avoid both of these characteristics. In their place, however,
the frequency of the numbers (or occasionally, letters) used as row coordinates tends to
be higher than the other coordinates. Choosing the highest frequency numbers as row
coordinates gives a starting point to reconstruct a monome-dinome system. Consider
the next example.




a. Repeats are underlined and the number frequencies are shown in the example. A
dinomic system can be ruled out, because the repeats are an odd interval apart. The
distance between the repeats is 153 characters, counting from the first character of
one to the first character of the next. A three-for-one substitution is possible from
the position of the repeats, but no patterns or positional limitations appear when
divided into threes. The very high frequency of the numbers 0 and 9 in relation to



5-13
the other numbers suggests that the system is monome-dinome. The most likely
row coordinates are 0 and 9. Other row coordinates are possible, but at this point it
is best to start with the most likely candidates only.


b. Begin by breaking the message into monomes and dinomes using only the 0 and 9 as
row coordinates. Mark off the divisions in pencil, keeping in mind that some
changes may be required later. Start with the first character of the message and
work through in order to the end, marking off the monomes and dinomes. Whenever
the first character after a division is a 0 or 9, include it with the next character. If it
is any other character, leave it as a monome.




c. With the divisions in place, we can try a word pattern on the long repeat.




d. We next set up a monome-dinome matrix with row coordinates 0 and 9 and include
the recovered letters. Shown below is the partially recovered matrix and the crypto-
gram with all letters from RECONNAISSANCE placed in the plaintext and the
matrix.



5-14
e. These recoveries suggest additional plaintext, particularly the message beginning
AERIAL RECONNAISSANCE REPORTS ENEMY. Placing these new values
leads to additional recoveries.



5-15
f. Several things remain to be done to complete the solution. The columns can be
rearranged to recover a keyword in the top row and alphabetical progression in the
next two rows. Additionally, there are two unrecovered segments of text. Both of
them include a number of 5s, and the preceding text in each case suggests numbers.
The solution is that there is another row in the matrix with the 5 as its coordinate. It
was not used enough to select from frequency alone, but once enough text was
recovered, the structure can be seen. The added row includes the numbers. The
complete solution appears in the next example, with the recovery of specific num-
bers only tentative.



5-16
Application of Vowel-Consonant Relationships
5-8.
to Multiliterals
Vowel-consonant relationship solutions can be applied to multiliterals, too. As long as
you can determine the coordinates of the matrix, you can set up a dummy matrix with
any sequence of characters inside as a pseudoplain component. You then reduce the
cryptogram to unilateral terms by deciphering with the dummy matrix. Next, solve the
resulting unilateral cryptogram using any of the techniques learned with unilateral
systems, including the use of trilateral frequency counts and the vowel and consonant
lines.



5-17
Solution of Trilateral and Trinomic Systems
5-9.
Trilateral and trinomic systems are solved in exactly the same way as biliterals and
dinomics. The systems are identified by the tendency of messages to break into groups
of three instead of groups of two. With simple triliterals and trinomics, positional
limitation is even more evident than it is for biliterals and dinomics. Look for a limited
set of pairs of characters as either the first pair of characters or the last pair of charac-
ters in every three, Once these are found, set up your coordinates and solve as before.




Section III
Analysis of Variant Multilateral Systems

Identification of Variant Systems
5-10.
As with any coordinate system, analysis of variant multilateral systems begins with
determination of the coordinates. If the product of the row and column coordinates is
50 or more, the system is almost certainly a variant system of some kind.


Analysis of External Variant Systems -
5-10.
Frequency Matching
External variant systems are generally easier to solve than internal variant systems.
Frequency counts can usually be used to determine which coordinates combine with
each other on the same row or column, whenever the text is long enough to give a good
representative sample, as shown in the next problem.




5-18
a. The cryptogram used 10 different letters as row coordinates and 10 different letters
as column coordinates. Using these coordinates, a digraphic frequency count has
been completed as shown. For example, the letter I is paired with itself five times,
so the number 5 appears in the matrix at the point where the row and column of I
intersect.
b. Examining the frequency count, we can see that there are good frequency pattern
matches between certain rows and certain columns. For example, the I row and the
R row are nearly identical. Similarly, the A column and the I column are nearly
identical. Carrying this process further, we can match the row pairs, AU, DP, IR,
MN, and OS. The column pairs are AI, CN, GS, MO, and RU. At this point, we
have no idea in what order the coordinate pairs belong or which letter in each pair
comes first or if it even matters which letter comes first. We have enough informa-
tion, however, to reduce the cryptogram to unilateral terms.
c. To reduce the cryptogram to unilateral terms, we set up a matrix with the combined
coordinates and write any sequence of letters within it, for example, A through Y.




5-19
d. We see that repeats appear in the pseudotext that results from our trial decipher-
ment. The repeats that were suppressed by the variants are now visible with the
variants combined. The recovery of the plaintext is like any of the previous
problems. When we recover the plaintext and enter the recovered values in the
matrix in place of the trial sequence, we reach the solution shown below.




5-20
e. With the plaintext values filled into the matrix, we can see in what order the rows
and columns belong. Starting with the last row of the internals, we rearrange the
columns of the matrix in alphabetic order.




5-21
The first row of the internals should follow alphabetically after the third
row”scdef, gikln.




f. All that remains is to fill in the missing letters H, J, and Q in the plaintext sequence,
and to try to recognize how the coordinates were constructed. As mentioned earlier,
it is common practice to couple I with J or U with V when using a 5 by 5 matrix.
Since J did not appear in the plaintext, we may assume it occupies an alphabetical
position within the I block. The Q clearly belongs between the P and T, leaving the
H in the top row. The plaintext keyword is BRAHMS (the classical composer). With
that as a clue, the letters in the coordinates are shifted to their correct positions,
revealing the keywords PIANO, DRUMS, MUSIC, and ORGAN.




Analysis of Variants - Isologs
5-12.
Two or more encrypted messages with different encrypted text, but the same underly-
ing plaintext are called isologs. When isologs are encountered, your job is much easier.
Isologs are particularly useful in solving variant multilateral systems, either external or
internal.
a. Isologs can be recognized by one or more of these characteristics”
Identical message lengths.
Similar characteristics in the text, such as repeated segments or characters
occurring in the same position in each message.



5-22
External indications, such as identical times of file or identical message numbers
included in the header for each message. Normally, no two different messages
from the same sender receive the same file time or message number. When you
see the same time of file on the same date originating from the same unit, the
messages are likely to be isologs.
b. Two messages that showed the same time of file in the message header appear in
Figure 5-3.




c. Each message shows positional limitations. Message 1 has the letters
ADFGLNQRTX in the row coordinate position and BCHKLMPSVZ in the column
coordinate position. Message 2 has AEFGKLOQVZ in the row coordinate position
and BDHMPRSTWY in the column coordinate position. The two messages are not
encrypted in the same system, but they appear to be isologs.
d. The initial step in solving these isologs is to see what values equate to each other in
the two messages. Pick one of the most frequent digraphs in either message as a
starting point. For example, FH occurs four times in the first message. A frequency
count, while not strictly necessary, may be helpful in spotting the most common
values. The digraphs that occur in the same positions in message 2 as FH in
message 1 are OS, GW, GS, and another OS.

e. The next step is to find each of the digraphs in message 2 that equated to FH from
message 1. The letters OS, GW, and GS in message 2 and the digraphs in the same
position in message 1 are underlined in Figure 5-3.



5-23
f. We now see that RH, RP, FP, and FH in message 1 equate to GS, GW, and OS in
message 2. A check of the new values in message 1 adds the additional digraph OW
in message 2, completing the equations for that set. It appears that R and F are
variant row coordinates and P and H are variant column coordinates in message 1.
Similarly, the message 2 variants are G and O on the rows and W and S on the
columns.

g. Continue the process by picking additional repeated values. Complete the equa-
tions for each, working back and forth between the two messages, just as we did for
the initial digraph FH. Continue until all coordinates have been combined, or you
run out of digraphs to compare. You can set up a plot to keep track of the equations
as shown in the next example.




h. Other combinations could have been selected than the ones shown, but these are
sufficient to show all the variants in both matrices. From this point, either message
can be reduced to unilateral terms and solved. Then the recovered plaintext can be
applied to the other message to complete the recovery of the second matrix. Note
that if the same matrix was used in both messages, the similarity should be quickly
recognized and the solution accomplished more easily. The next paragraph shows
the simpler technique when the same matrix is used.

Solution Using Isologous Segments
5-13.
Segments of ciphertext which have the same underlying plaintext are known as
isologous segments. A technique similar to the one used in isolog solution can be used
any time repeated plaintext can be identified. This is likely to occur with repeated
beginnings and endings to messages or with long repeated words and phrases.
a. Recognizing repeated plaintext in variant systems requires painstaking inspection
of the ciphertext. Computer indexes of repeated plaintext, which show repeated
text on consecutive lines along with the preceding and following text makes repeats



5-24
easier to recognize. In any long plaintext repeat, some of the ciphertext digraphs or
dinomes are likely to repeat. Other ciphertext digraphs or dinomes are likely to
show common row or column coordinates. Pairs with neither row nor column coor-
dinates in common will generally be in the minority. Therefore, although a lot of
trial and error may be involved, the longer repeated plaintext segments can often be
identified. Consider the two message beginnings shown below.




b. The similarities of the text make it quite clear that the underlying plaintext is the
same in both cases, and the same matrix is used for both. Proceeding on the
assumption that the plaintext and matrix are the same, it is easy to match the
remaining values to determine the variants. For example, from the first dinome in
each message, 3 and 4 are column variants. From the second dinome in each
message, 8 and 9 are column variants. All the variants can be combined from this
short example, and the remainder of the solution is routine.




Analysis of Internal Variant Systems
5-14.
Internal variant systems are generally more difficult to solve than external variant
systems. With no coordinates to combine, frequency counts do not provide immediate
clues to variants. Similarly, isologous segments are harder to recognize. Some charac-
ters are likely to repeat in isologous segments with internal variant systems, but the
partial repeats caused by common row or column coordinates are much less likely to
occur. Still, given enough messages from a single system to produce repeats; given
operator carelessness in encryption; or given stereotyped traffic, these systems can
readily be solved, too. Once a plaintext entry is found, the remainder of a solution is
not difficult. When you find isologs or isologous segments, you can equate ciphertext
values just as was demonstrated in the internal variant examples. The only difference
is that you do not combine coordinates through this process, but instead find all cells
in the matrix that have the same plaintext value.



5-25
Analysis of Syllabary Squares
5-15.
Syllabary squares are closely related to small code charts, and the solution of both
types of systems is similar. The analysis of syllabary squares produces some distinct
differences.

a. Isologs or isologous segments are not necessarily the same length in each case. The
encipherment examples below are repeated from paragraph 5-3e.




b. Isologous segments can often still be recognized by the plaintext values which have
no variation. In the example, there is only one way to encipher the letters M and S.
When REINFORCEMENTS is enciphered, the ciphertext equivalents of M and S
will always be the same. Other values are likely to begin with the same row coor-
dinate, since syllables beginning with the same letter are likely to be on the same
row, such as the R and the RE. Still others will have a possible variation, but the
variation will not be used. The repeated CE syllable in both segments is an example
of this. As a result of all these considerations, isologous segments are often
recognizable and provide a point of entry to the system.

c. Solution of syllabary spelling will be further explained in Part Six, Analysis of Code
Systems.



5-26