The present article describes US War Department Telegraph Codes and other codes between the Civil War and World War II.
During the peacetime after the Civil War, there was not much demand for military codes/ciphers. Still, in 1885, "Telegraphic Code to Insure Secrecy in the Transmission of Telegrams" was compiled by Lieutenant Colonel J.F. Gregory, A.D.C, and issued the next year. (ASA p.112)
However, this was only a minor adaptation of a widely used commercial code by Slater, Telegraphic Code, to Ensure Secrecy in the Transmission of Telegrams (see another article). It assigned numbers 00001-24000 to words in alphabetical order and 24001-25000 to places, persons, forts, etc.
The numbers were not intended as code numbers to be transmitted but were for the process of replacing a plaintext word with another. Specifically, the number adjacent the plaintext word is manipulated (e.g., by adding 5555) and the word adjacent to the resulting number is transmitted. Slater's Codebook presents nine examples of such manipulations by combining additions, subtractions, transpositions (e.g., interchanging the 3rd and 4th digits), and regrouping (e.g., regrouping five 5-digit numbers plus three padding 0s into twelve 4-digit numbers) with an example sentence "The Queen is the supreme power in the Realm." The War Department code incorporated the same examples with a different example sentence: "War is punishment whereof death is the maximum." (ASA p.139)
The War Department undertook preparation of a new telegraphic code in January 1899 (ASA p.118-). It was occasioned by extraordinary telegraphic expenses lately incurred. The task devolved on the Chief Signal Officer of the Army, who had a practical knowledge of telegraphy. (For the complexity of regulations concerning code words, see another article.) Since there was an urgent need of reducing cable cost, the Chief Signal Officer at first prepared a "Preliminary War Department Telegraphic Code" in sixteen days. It adopted a commercial code, "Western Union Telegraphic Code" (see another article) to be used temporarily and further provided nearly 4000 sentences likely to be used in military correspondence.
The permanent version of the "War Department Telegraphic Code" appeared either late in 1899 (from the date on the title page) or on 16 January 1900 (as indicated by a stamp on the leather binding). (ASA p.119-120) It still incorporated the main body of the Western Union Telegraphic Code (the first 471 pages, corresponding to serial numbers 1 to about 78200), followed by the War Department Telegraphic Code proper from about 78200 to 100,000.
The War Department Telegraphic Code proper had the following sections.
From time to time, appendices were issued to increase the vocabulary. Appendix No. 3 of a list of Army officers was issued during the 1904 fiscal year (ASA p.124). While the main purpose of the code was economy, the Report of the Chief Signal Officer for 1900 pointed out that secrecy might be attained by enciphering the serial numbers attached to the code words.
In 1902, a new code, prepared by W.H. Allensworth and W.G. Spottswood, was issued by the Adjutant-General's office (ASA p.122). (This "cipher", which is actually a "code" in modern terminology, is the only one made under this office rather than the Chief Signal Officer.)
This is a one-part code (i.e., the entries are alphabetically arranged). It had 751 pages numbered 100 to 850, each containing one hundred entries 00-99. Each entry was given a five-digit code number as well as a code word. For example, the word "abandoning" could be encoded as either 10077 or ABERRANCE. The main body ends with "Zouaves" (plaintext) numbered 77138 and the following blanks up to 77999. Thereafter is an appendix dated 1 March 1902, which contains lists of arsenals, banks, days of the year, companies, headquarters, units, and organizations down to companies, stations, officers of the Army, railroads, phrases of reference, steamship companies, surety companies, telegraph and cable companies, and transports.
In order to encipher words/names not in the code, a table for polyalphabetic substitution was attached. It was similar to a Vigenere table but had a reversed standard alphabet (known as Beaufort cipher).
Five-digit code numbers (as opposed to variable-length serial number of the 1899 code by the Chief Signal Officer) were considered preferable because "if the word code is used, vexatious and other serious mutilations of original messages are committed by operators ignorant of each other's language." (pp.iii-iv, quoted in ASA p.123) This is reminiscent of the statement in the Directions page of the Red Code (1876) of the State Department (see another article) that Chinese, French, German, Italian, Japanese, Russian, Spanish, and Turkish telegraph operators, ignorant of English, "constantly commit vexatious and often serious mutilations of original messages" (Masked Dispatches (2013), p.173). This, as well as the five-digit code number made of page and line numbers, use of the term "Cipher" in the title, the polyalphabetic cipher for spelling unlisted words, etc. suggest influence of the Red Code (1867) and the Blue Code (1899) of the State Department.
Apparently, individual departments had their own codes.
The Report of the Bureau of Insular Affair of the War Department dated 31 October 1905 (Internet Archive p.34) describes consolidation of cable codes in use with the addition of a large number of new phrases. It used "code words" carefully selected to be less error prone, rather than "figure groups."
The "Insular Bureau Supplement" in typescript, prepared by the Insular Bureau of the War Department in October 1909, included numbers 84025 to 107244 (ASA p.126).
The supplement for the Insular Bureau prepared in 1926, intended for use with the War Department Telegraph code, had page numbers 850 to 1067, with each page containing 100 groups, each assigned "line symbols of vowel-consonant form" as well as a two digit number (ASA3 p.25).
The 1902 code, prepared by General Greely, of the Adjutant-General's office was not a replacement of the 1899 code, since the report for the year 1904 of the Chief Signal Officer was still concerned with the 1899 code (ASA p.124-125). There was a major change in the international telegraph regulations effective from July 1904. Now, any pronounceable words could be used as code words, regardless of whether they are real words in some language (see another article).
With a slight change in the title, the War Department Telegraph Code was issued in 1906 (ASA p.124-). This assigned a six-letter code word and a five-digit code number to each entry. (Apparently, the compiler was yet unaware of the economy of five-letter code words as used in commercial codes such as Bentley (1906), whereby two five-letter code words (which might better be called code groups or half-words) are combined to form one ten-letter code word, thus reducing the cost by half.)
The main body after classified sections was alphabetical (i.e., one-part code) and began on page 89 with the indefinite article a or an assigned 10467=AXIFIK and ended with 62000=TEBADU (blank for addition). The classified sections were for names of officers, military organizations, posts and stations, numerals, arrivals and departures, dates, indorsements, letter acknowledgements, requisitions, telegram acknowledgements, mails, shipments, transports, miscellaneous items, vessels, geographical, and radio stations.
The introduction now prescribed that secret messages should employ superencipherment but did not provide how it should be done.
The War Department Telegraph Code of 1915 was prepared under the direction of the Chief Signal Officer of the Army (ASA p.128-). It was printed in a commercial printer in Ohio, despite bearing "Government Printing Office" on the title page (Cleveland in ASA p.128; Cincinnati in Friedman p.1).
Now, the code words consisted of five letters instead of six. The code had 114,000 entries, the last (a blank for addition) being USGKH. As shown by this example, the code words are not pronounceable (at least upon first sight), failing to meet the requirements of the international telegraph regulations. So it had to be used with a table for converting the serial number of each group into a pronounceable five-letter group (ASA3 p.5). The "Table for Forming Five-Letter Code Words" was issued sometime in 1916 and was discontinued in April 1918, when it was decided that secret message should be enciphered with secreta tables to be changed frequenty furnished by MI-8 (see below) (ASA2 p.34-35, 37).
It had advanced features of error prevention: two-letter differential and a mutilation table (see another article).
Apart from the main body, there were classified sections for blanks for emergencies; organizations; departments, etc.; radio and telegraphic stations; vessels; Naval officers; geographical names; Mexican geographical names; special persons and firms; dates, etc.; hours and minutes; numerals; railroads; cables and cable companies.
In March 1918, the Code Compilation Section of the American Expeditionary Force in France (see another article) compiled an addenda sheet for providing a number of omitted words as well as transports and French cities and towns (Friedman p.19, 272).
The Code Compilation Section of the American Expeditionary Force also compiled a comprehensive code for headquarters, since the War Department Code was "intended primarily for cable work and not for active operations in a foreign country." In May 1918, the Staff Code, containing about 30000 words and phrases, went to press and was completed one month later. (Friedman p.19; p.213, Appendix 18)
Each page has 100 entries in two columns. One notable feature is that the Lefthand Column and the Righthand Column form separate series, with the former for place names, persons, spelling combinations, numerals, the less commonly used articles of equipment and supply, and army organizations in France and the latter for commonly used words and phrases.
The serial number (20000-50399) is assigned throughout both columns consecutively (e.g., the first page contains 20000 to 20099) but these were usually not meant for transmission. The four-letter code groups of the Lefthand Column contain two vowels and two consonants arranged in logical sequence, while only consonants are used for the Righthand Column.
Variants code groups are provided for most commonly used words.
This was basically a one-part code but a series of distortion tables for superencipherment were made in July. The distortion tables are for replacing a two-letter combination to another two-letter combination.
An encoding table for indicating the date and time is also provided. The first letter indicates the month, the second and third indicate the day, the fourth and fifth indicate the hour, and the sixth and seventh indicate the minutes. After (redundantly) appending to these seven letters "O" (if before 12 noon) or "Y" (after 12 noon), two four-letter groups are ready for transmission. For example, BOVO FYJY can be parsed as B (January) OV (29) OF (22) YJ (53) Y and means January 29, 22:53. By specifying the time down to the minutes, the message can be referenced as, e.g., "Your BOVO FYJV."
In case of transmission error, it is advised that alternative letters for each letter of the code group should be tried to find the one that fits the context. While the logical arrangement of the letters in the code group provides some help, the process would be cumbersome compared to the use of a mutilation table provided in the War Department Telegraph Code of 1915.
The Military Intelligence Section (later also known as the Military Intelligence Branch or the Military Intelligence Division (ASA2 p.31)) under the General Staff of the army was also involved in code compilation during World War I.
In May 1917, soon after the United States enetered World War I, the Military intelligence Division, which had been merged under the War College Division, was given a new footing under Van Deeman (1865-1952), sometimes called the Father of American Military Intelligence, and was ultimately reorganized into sections MI-1 to MI-12 (Finnegan p.22, ASA2 p.11, The Evolution of American Military Intelligence (1973) p.20, Wikipedia). Of these, the eighth section, MI-8, for codes and ciphers was established in July 1917 for Herbert O. Yardley, who had been a code clerk in the State Department and had seen weakness of the American information security (Kahn (2004), p.20-21, ASA2 p.14). After the war, the task of MI-8 would be taken over by the Cipher Bureau, inheriting some of the members of MI-8 including its chief, Yardley (Kahn (2004), p.53, 55, 59). (Even then, the Cipher Bureau was occasionally called MI-8 (ASA2 p.78 etc.).)
Though code compilation had generally been done under the Chief Signal Officer before World War I, MI-8 took on code compilation on its own when it was informed soon after its establishment that the War Department's code was not secure (ASA2 p.30-33). (The American Expeditionary Force in France also had a Code Compilation Section.)
The Code Compilation subsection of MI-8 worked on compilation of Military Intelligence Code No. 5, which was completed in July 1918 (ASA2 p.37 ff.). (There were no Nos.1-4 or 6-8 (ASA2 p.33).)
The five-letter code words were pronounceable groups in the form of either VCVCV, VCCVC, or CVCVC (V: vowel, C: consonant). Two-letter differential was employed for the code words. For each of these patterns, a mutilation table was provided for error correction. (ASA2 p.38)
Five-digit groups 01000-99999 (not consecutive) were also given to the entries. The main body (p.1-495), with each page containing two culumns of 48 entries each, had a total of 495x96=47,520 groups, which is the maximum number possible by the three mutilaiton tables.
At the bottom of each page were code groups for ten frequent entries: "comma", "ing", "ing" (sic), "paragraph(s)", "period", "quote", "semicolon", "spell", "spelling ends", "unquote." (A similar idea was adoped for field codes, for which see another article.) Again, to facilitate decoding, these code groups were repeated in their respective places in the normal code group sequence.
Moreover, many frequent entries were given variant code groups. For example, the letter "A" could be represented by "ABACE" or any of the variants "ABHIK", "ACEHO", "HAHFO", "QACEN", or "ZABJY" printed in the blank space between the columns.
The main body was followed by three classified sections for the 365 days of the year, cardinal numbers, and ordinal numbers. The code groups for the classified sections were randomly ordered. To facilitate decoding, those code groups were repeated in the main body. That is, essentially, the classified sections formed a two-part code.
Whereas the War Department Telegraph Code of 1915 was used only for confidential information, the Military Intelligence Code No. 5 was used for secret information (ASA3 p.4, 6, 19).
The Military Intelligence Code No. 9, employing the two-part principle throughout, was ready on 2 December 1918 and held in reserve (ASA2 p.41 ff., ASA3 p.9). Since it now had separate encoding and decoding sections, there was no need of entries at the bottom of each page or between columns. Apart from these, the structure was similar to the Military Intelligence Code No. 5 (495 pages, each with two columns of 48 groups each, the same pronounceable patterns, and the same kinds of three classified sections).
As noted above, the American Expeditionary Force in France had its own Code Compilation Section but lack of local place names in the code used in France led to a proposal in June 1918 for a list of code words taken from a French map. It was the Code Compilation subsection of the Military Intelligence Division that prepared the French Geographical Code, finished on 1 October 1918 (ASA2 p.42 ff.).
It contained 10,800 entries, each with five-letter code groups (ABAAB to VUIKY) and five-digit (non-consecutive) code groups (00000 to 35997), of which about 1,000 were blanks for later addition. All the entries were for French place names "taken from a strip of the map of France (on a scale of 1:200,000) extending twenty-five miles on either side of the battle front as it existed on July 10, 1918, as well as from the Amiens, Paris, Lille, and Chalons areas."
The five-letter groups had either VCVVC, CVVCV, or CVCVV, with the double vowel allowing easy distinction from the Military Intelligence Code No. 5. For each pattern a mutilation table was provided.
Change in the military situation led to preparation of Goegraphical Code No. 2 (ASA2 p.45 ff.). The work was completed as early as 17 October 1918 and was ready for distribution in mid-November. Its entries covered "the area beginning with the Battle Line as it existed September 15, 1918 (the day on which ended the Battle of Saint-Mihiel), north and east taking in all of Belgium, a small portion of Holland bordering on the Belgian about 25 miles beyond the Rhine."
It had a substantially enlarged vocabulary of 26,885 entries, each with five-letter code groups (BABAL to XOZYP) and five-digit code groups (05700 to 83448), of which 138 were left blank and 5 were reserved as indicators.
The letter groups followed the pattern CVCCV or CVCVC with a two-letter differential. Both patterns were covered by a single mutilation table. Although the pattern "CVCCV" was also used in the Military Intelligence Code No. 5, the indicator code groups (93758, 93812, 93973, 93985, and 93987) were used to mark the use of this code.
A "pocket code" was prepared in December 1918 (ASA2 p.49ff, ASA3 p.4) for use by military attachés and other agents, particularly when accompanying the Army of Occupation into former enemy territories. Although it was prepared by the Military Intelligence Division, it bore the name of a fictitious firm "Ideal Code Company" on the title page.
It had 148 full pages, each having two columns each with 42 entries, plus one page with 25 entries, totaling 148*42*2+25=12457 entries. In addition, 532 code groups were provided for ordinal and cardinal numbers. As an example, WERUN represented "Commander, s, in Chief, Amex Forces."
This two-part code was unique in its physical format in which the encoding section was printed solely on the right-hand pages, whereas the decoding section was printed on the left-hand pages upside down. This allowed a user to always hold the left-hand page with his left hand without blocking his view.
The code words, formed on the basis of a mutilation table, followed a CVCVC pattern and observed a two-letter differential.
A distortion table (a monoalphabetic substitution table for encipherment) was included in the introduction.
The latest of the few known use was as late as March 1921 (ASA3 p.4, ASA2 p.51).
(See the image of the first page of Introductions in Christos (1))
A new code was compiled by Major Barnes, who had led the Code Compilation Section of the American Expeditionary Force, on the basis of the Staff Code and was completed in April 1920 (ASA3 p.13, 14). It was issued in September 1921 to replace the War Department Telegraph Code of 1915 (ASA3 p.8).
In the compilation work, the 1915 Code, the 1906 Greely Code, and three or four largest commercial codes were consulted. In addition to actual telegrams and various other documents, spelling combinations of English, French, German, Spanish, and Japanese were collected. (ASA3 p.14-15) (The similarity suggests that the Military Intelligence Code No. 5 was also consulted.)
In War Department Telegraph Code (1919), assigned a short title SIGRIM, the alphabetically arranged main body is preceded by four classified sections: Table I for the days of the months; Table II for cardinal numbers beginning with fractions and ending with 6,000,000; Table III for Army organizations and ordinal numbers; Table IV for ordnance, ammunition, and equipment and certain numerical designations used in ordering supplies. The five-letter and five-figure groups for the entries of the tables were selected at random from the main body in order to make them distinct adjacent entries. To facilitate decoding, the assignments in these four tables are repeated in the main body. For example, 50066=MOSIP, representing "1st Field Signal Battalion" in Table III, could be looked up in the logically arranged main body. So, essentially, the classified tables adopted two-part code.
The main body began on page 59 with 10009 =ABAMI for "A" and ends with 83590, a blank (ASA3 p.13).
The five-letter code words were pronounceable groups in the form of either VCVCV, VCCVC, or CVCVC (V: vowel, C: consonant). The letter "H", often mistaken for "S", is never used and no groups begin with "Y" or "Z". With each "root" consisting of the first three letters, 15 two-letter "terminals" are combined. In the 15 combinations, all 6 vowels are used twice and 3 vowels, selected according to rules, are used three times. A mutilation table is provided (ASA3 p.14) for error correction, presumably based on such systematic construction of code groups.
This code was used for administrative traffic for many years until 1943/44 (Christos (2)).
At the end of World War I, Military Intelligence Code No. 5 was in use (ASA3 p.6). It was used with enciphering tables issued every two weeks (later two months). After the war, it was regarded as the War Department Staff Code, with the Military Intelligence Code No. 9 (1918), then held in reserve, was to serve as its reserve edition (ASA3 p.9).
The Military Intelligence Coe No. 9 was recalled about 1923 but it was reissued in April 1933 with a title page pasted over by a new one with a title "War Department Staff Code No. 2 (formerly Military Intelligence Code No. 9)", still for use in secret communication (ASA2 p.42). Now, it was given short titles SIGSYG (encode) and SIGPIK (decode).
In 1934, the Military Intelligence Code No. 1, demoted for use for confidential information, was reissued with a new title page pasted: "War Department Confidential Code No. 1 (formerly Military Intelligence Code No. 5)" and a short title SIGCOT (ASA2 p.40).
The function of code compilation of the Cipher Bureau (MI-8) of the Military Intelligence Division, separate from that performed by the Chief Signal Officer, was transferred to the latter as a result of a conference in October 1920, while the Adjutant General remained responsible for reproduction, distribution, storage, and accounting (ASA3 p.9-10). The Cipher Bureau, led by Herbert O. Yardley, continued its codebreaking activities (in New York from August 1919).
In the Office of the Chief Signal Officer, William F. Friedman (Wikipedia) performed all code compilation work with the assistance of a single clerk-typist until 1930 (ASA3 p.20-21, 41).
Friedman, already a cryptographer before the war, served as a first lieutenant in the Military Intelligence Division at the headquarters of the American Expeditionary Force in France and was engaged in codebreaking. After the war, he was contracted to revise the Staff Code in December 1920. Subsequently, he was further trusted with the task of compilation and revision of field codes as well as preparation of cipher tables for the Staff Code. He entered on permanent duty in the Office of the Chief Signal Officer on 31 December 1921. (ASA3 p.16-17, 19-20)
By the beginning of 1924, a new edition of the War Department Staff Code, containing 60000 groups, was in the final stages (ASA3 p.23).
Friedman and an officer from the Navy prepared an inter-departmental code for intermediate-level communication between the Army and the Navy in 1927-1928 but apparently it never went to press because of lack of funds (ASA3 p.35-36). For higher-level communications, it was decided to exchange liaison officers given the codes of their own departments because it was considered undesirable to trust their codes to officers of the other department. For low-level communications, cipher cylinder device M-94 was distributed. (ASA3 p.32-34)
The Millitary Intelligence Code No. 10, compiled by the Code and Cipher Section, was completed in 1927 (ASA3 p.41).
In 1941, the Germans were supplied with photographed codebooks and substitution tables of the Military Intelligence Code No. 11. These allowed them to read encrypted reports from Cairo to Washington, which provided valuable information to Rommel. (Accounts differ as to German success before obtaining these materials.) Before long, the compromised was discovered (accounts differ as to how) and, in June 1942, the code was superceded by either the M-138 strip cipher or a SIGABA machine.
This code was called the TELWA Code by German codebreakers by its indicator. The Germans found that the letters in five-letter groups were systematically chosen and they identified the relation between the letters. (See the case of the 1919 code above.) Then, studying the pattern at the beginning and end of the radio messages, the meaning of five-letter groups were identified one by one. For example, RYKFI was an opening parenthesis, UZUSP was the word "signed." (Schmeh, Christos (1))
Compared with the longevity of the 1919 code, the 1942 code was short-lived, being replaced by Cryptographic System No. 999 (SIGUCC) in May 1945 (Christos (1)).
Army Security Agency [ASA], (ed.) Wayne G. Barker, The History of Codes and Ciphers in the United States Prior to World War I, A Cryptographic Series (1946, 1978)
Army Security Agency [ASA2], (ed.) Wayne G. Barker, The History of Codes and Ciphers in the United States during World War I, A Cryptographic Series (1946, 1979)
Army Security Agency [ASA3], (ed.) Wayne G. Barker, The History of Codes and Ciphers in the United States during the Peiord between the World Wars. Part I. 1919-1929, A Cryptographic Series (1946, 1979)
William F. Friedman, "American Army Field Codes in the American Expeditionary Forces during the First World War" (pdf) (1942) (See another article for a list of appendices.)
T. Christos (1), "The US TELWA Code", Christos military and intelligence corner (2012)
T. Christos (2), "United States cryptologic security failures in WWII", Christos military and intelligence corner (2014)
T. Christos (3), "US military attaché codes of WWII", Christos military and intelligence corner (2012)
Klaus Schmeh, "Als deutscher Code-Knacker im Zweiten Weltkrieg", Telepolis (2004)
Klaus Schmeh, "Wie ein Rätsel der Cryptologie-Geshichte nach 70 Jahren gelöst wurde", Klausis Krypto Kollumne (2015)
David Kahn (2004), The Reader of Gentlemen's Mail, Herbert O. Yardley and the Birth of American Codebreaking
John Patrick Finnegan (1998), Military Intelligence (Google)