Washington's Spies
Page 14
By the early seventeenth century, the French master cryptographer Antoine Rossignol had worked out how to confound his peers by using a two-part nomenclator. The basic principle—turning words into numbers and back again—remained the same, but Rossignol partly randomized the system so that “apple” might be represented by 389 and “zoo” by 25.48 He did this by giving the encoder a book listing alphabetically all the words that would ever need to be masked, with a random number in the next column. The decoder received a book with the digits in numerical order and the words they represented next to them, rather like a bilingual dictionary. A two-parter made codebreaking more difficult, because if, say, “the” was given the number 262, it did not necessarily mean that numbers 1 through 261 belonged to words running from “a” to “that.” Even so, it still suffered from the same fundamental defect as its more primitive ancestor: A cryptanalyst could detect 262 occurring suspiciously often, thereby indicating a lurking “the,” and so on down the line of frequency.
One way of getting around the problem was relatively simple: Don’t encode common words like “the.” Instead, encipher only the more important, but rarer, ones, such as “muskets,” “France,” and “negotiations.” Nevertheless, if the cryptanalyst possessed several encoded messages, he could try to guess the meaning of a number from its context in a sentence by cross-referencing it with the other messages in which the same number was used.
In any case, French superiority in the field had eroded by the early 1700s. Other countries learned to baffle codebreakers by increasing the number of nomenclators from the traditional several hundred to two thousand or three thousand elements—so that each word could have several random numbers assigned to it—and including scores of nulls to throw decipherers off the scent. The French responded by using half a dozen sets of nomenclators simultaneously, a decision that caused immense confusion among recipients of enciphered letters who may not have received the newest update.
Despite the complexity of this new breed of cipher, it was often broken thanks to a willingness to cheat. Europe’s Black Chambers dug deep in the diplomatic bags to find the freshly issued two-part lists destined for outlying embassies. By the time the local ambassador was perusing his new decrypted mail over breakfast, the local Black Chamber had already read it.
Human error was another major factor. At great expense, England’s Decyphering Branch in the early 1700s developed an impregnable four-digit code for diplomatic use only to discover years later that the Foreign Office, thinking it would save some money, had gone on to use the same code in embassies from Gibraltar to Stockholm for more than a decade. Changing the code numbers at least three or four times a year, and allotting each embassy a different version, should have been automatic. As it was, the code had been leaked or broken so long before, the Foreign Office might as well have sent its messages directly to the enemy in plaintext.49 Then there were the diplomats who, after receiving a document written in plaintext from their home government, sent their reply in cipher, so providing enemy codebreakers with a crib sheet of the newest version when they steamed it open about an hour later and compared it with their copy of the original document.
By the eve of the Revolution, it was clear that for the most sensitive diplomatic communications, two-part nomenclators were insecure. To the Americans’ rescue rode Charles William Frederic Dumas, a German in his fifties living in the Netherlands, who had turned into a zealous Patriot partly owing to his friendship with Benjamin Franklin. Employed by the Americans as their agent in Holland, and acquainted with the practices of the Black Chambers, Dumas devised a robust system combining an early version of what is known as a “stream cipher” (encrypter and decrypter share the same text for their correspondence) and a randomized key in which any of maybe a dozen numbers could represent the same letter.
The concept is easier to grasp by looking at the first couple of sentences from the patriotic, 682-character-long passage Dumas and Franklin used as their shared source-text for their correspondence: “Voulez-vous sentir la différence? Jettez les yeux sur le continent septentrional de l’Amérique.” (Translation: “Do you want to feel the difference? Glance at the continent of North America.”)
Then Dumas numbered each character consecutively, e.g.:
Once the entire passage is numbered thus, the encoder could choose between no fewer than 128 different numbers for “e” (or 63 for “r,” 60 for “s,” or 50 for “a”). In the sentence above, for example, “e” was represented by 5, 13, 24, 26, and 29. When it came to “w,” however, or “k,” he was out of luck, for the original passage did not contain them: Franklin and Dumas substituted “uu” and “c” for them. There was also but one number each for “y,” “?,” and “&”—flaws that no veteran cryptanalyst scouring for such regularities would overlook. Nor would he overlook the Dumas Cipher’s other failing—that the encoder must choose between the 128 numbers for “e” as randomly as he can; what Franklin had a lazy habit of doing was repeatedly choosing the first couple of numbers that appeared. In the case of “e”—see example above—he would use 5 or 13 far more often than, say, 29. Again, the frequency of “e” was the dead giveaway.
As far as it went, however, the Dumas Cipher was simple to use and provided fairly good security. The same could not be said of the polyalphabetic system invented by the leading American cryptographer of the time, James Lovell (born in 1737 and Harvard educated), whose cipher was brilliant, yet so confusing that no one could work out how to use it, least of all John Adams, whom Lovell persisted in subjecting to his cipher, much to the former’s dismay. For years, he couldn’t read the missives the cryptographer sent him.
A boiled-down version follows, using just the first two letters of the keyword CRANCH (the name of a friend of Lovell’s and Adams’s).
1 C R
2 D S
3 E T
4 F U
5 G V
6 H W
7 I X
8 J Y
9 K Z
10 L &
11 M A
12 N B
13 O C
14 P D
15 Q E
16 R F
17 S G
18 T H
19 U I
20 V J
21 W K
22 X L
23 Y M
24 Z N
25 & O
26 A P
27 B Q
To encipher “bankrupt,” find “b” in the first column and write the number to its left; for the second letter, “a,” look for it in the second column and write down the number; then for the third letter, return to the first column and continue to alternate. Thus, “bankrupt” becomes “27.11.12.21.16.4.14.3.” If one isn’t paying attention—and Adams often wasn’t—it was easy to miscount and use the wrong column to substitute, thereby garbling the rest of the message. Indeed, Lovell himself used to miscount occasionally, leaving poor Adams still more befuddled than before. And this is all before accounting for Lovell’s arcane “rules,” such as the one laying down that “29.38” was a signal to the reader to begin deciphering in reverse, let alone his fondness for three- and four-letter keys (rather than the above two-letter), which made decipherment even more prone to miscounting. As Abigail Adams told Lovell, her husband was not “adept in investigating ciphers and hates to be puzzled for a meaning.” Consequently, Adams switched to the basic, tried-and-true alphanumeric system that was highly insecure but comprehensible, at least. Thus, to his friend Francis Dana, Adams wrote in 1782: “19 presses 18 to come to him, and he thinks of going in ten days”—18 stood for Adams, and 19, John Jay.50
Tallmadge couldn’t afford the luxury of a top-class diplomatic code—two reasons being the difficulty of training each Culper member in encipherment and ensuring he had the latest key-book—and was obliged to find a simpler, Adams-style solution. He began researching and compiling one in earnest in July 1779. However, almost from the very beginning, the Culper Ring had occasionally employed a
stripped-down code consisting of precisely one element: Thus, Woodhull dated his letter of March 17, 1779, “10 March 17 1779 [italics added].”51 Subsequently, in April, Woodhull clarified that:
“No. 10 represents N. York
20 Setauket
30 and 40 2 Post Riders.”52
The “Post Riders” were Jonas Hawkins and Austin Roe, and until July, the Ring used just those four numbers. By the end of that month, the Ring had switched to a full-scale Code Dictionary.53
The “Dictionary” was the sourcebook of the Culper Cipher, a creaky one-part nomenclator developed by Tallmadge. From his copy of the 1777 London edition of Entick’s Spelling Dictionary, Tallmadge chose 710 words he thought most useful for his agents—county, Congress, advise, gun, intrigue, longitude, navy, Tory, war, and so on—and wrote them alphabetically in the left-hand column; in the next, he numbered the words consecutively. Thus:
many 384
mercy 385
moment 386
murder 387
measure 388
He added 53 numbers, running from 711 to 763, to represent proper names. These included the names of his agents, leading actors, and places. The most important ones, for our purposes at the moment, were:
General Washington 711
General Clinton 712
Tallmadge (“John Bolton”) 721
Samuel Culper 722
Austin Roe 724
Caleb Brewster 725
New York 727
Long Island 728
Setauket 729
For those words not listed in the Dictionary, and for digits, Tallmadge added a bare-bones, mixed-alphabet scheme, which was as follows:
Original: 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
Code: e f g h i j a b c d o m n p q r k l u v w x y z s t
Original: 1 2 3 4 5 6 7 8 9 0
Code: e f g i k m n o q u
Thinking ahead, Tallmadge included a few rules to simplify decryption. Transposed digits were to have a double line drawn underneath to distinguish them from enciphered words; and for words in the past and future tenses and for plurals, a small squiggle, or “flourish,” was added above the code number.54
Tallmadge’s one-part nomenclator is a clone of early French cryptographic practice. Even so, in the American theater, it was counted as cutting-edge: It would not be until the fall of 1781 that this type of numerical code came into more widespread use.55 Tallmadge’s inspiration probably came, indirectly, from the Marquis de Lafayette, who would certainly have been au fait with elementary cryptography after his time at court and in the army of a country once the world leader in ciphering.
Lafayette arrived in America for the first time on June 13, 1777, and met Washington on July 31, 1777. The two became firm friends. He stayed until the beginning of 1779, and then returned from France in the spring of 1780. The Culpers began using Tallmadge’s Dictionary in mid-1779—during Lafayette’s absence. It might have been Washington—never averse to making proposals to improve the Ring’s effectiveness—who, remembering a conversation with Lafayette, described for Tallmadge the rudiments of a one-part nomenclator and code dictionary.
After a few false starts, the Ring quickly got accustomed to this strange new language. On August 6, for instance, part of one of their letters read:
Sorry 626.280 cannot give 707 an exact account 431.625 situation 431.625.635—707.373. think 626.280.249 not taken sufficient pains 634.442.284. I assure 707.626.280.249.190.284 more 146 than 280 expected. It is 282 some measure owing 683[?].379.414 having got 287.1.573 line 431.216 intelligence. To depend 668.80 reports 683.[?].183—I 537.5. conversed 680 two qjjcgilw 431 different 76 from 730 from 419.431 which 280 could 442.2 account 431.625 situation 431.625. army 630. I was afraid 430 being too 526.
The translation, made by one of Washington’s aides, is:
Sorry that I cannot give you an exact account of the situation of the troops. You may think that I have not taken sufficient pains to obtain it. I assure you that I have, and find it more difficult than I expected. It is in some measure owing to my not having got into a regular line of getting intelligence. To depend upon common reports would not do. I saw and conversed with two officers of different corps from Kings-bridge from neither of whom I could obtain an account of the situation of the army there. I was afraid of being too particular.56
Tallmadge’s cipher would not have withstood a Black Chamber specialist for long: The words and numbers were not randomized (so that “artillery” was represented by 46, and “troops” by 635); there were but 760-odd elements, whereas the minimum a one-parter should have had to delay breaking was between 1,500 and 3,000 (including nulls and multiples); and it contained 22 of the 27 most frequently used words—“the,” “an,” “at,” “I”—in the English language. Even a novice cryptologist knew better than to hand out such free clues. Note, also, that the writer of the above letter has transposed “of,” “that,” and “with” but left “intelligence” and “army” in plaintext. This bad habit soon ceased when it was realized that keeping common words as plaintext while enciphering the rarer, less easily guessable ones, made messages far more secure. Tallmadge, of course, could hardly have been expected to know all this, and one error of his stands out: the omission of an obviously pertinent word, “officers,” from his Dictionary, which obliged the sender of the letter above to transpose “qjjcgilw” (Tallmadge’s oversight is inexplicable, though he did include “office”).
Despite the Culper Cipher’s weaknesses, it did the job it was meant for: turning plaintext into code strong enough to baffle an ordinary reader while avoiding the fiendish complexity of Lovell’s keyword system. It was never intended to do more than provide backup security, for no matter how innocuous its contents, an encrypted letter is always bound to raise suspicions that something murky is afoot.
Armed with their code and ink, the Culper Ring was already the most “professional” of all Washington’s spies. By the summer, having found another candidate for investiture, it would be the most effective.
Pleased as Washington was with the Ring’s progress in cipher and ink, Abraham Woodhull’s failure to place a man in New York nagged at him. Amos Underhill was fine, so far as he went, but his role was to shelter Woodhull at the “safe house” during his stays, not to serve as a bona fide agent digging up material. For the time being, Woodhull would have to do until he could find a suitable candidate for investiture in the Ring. Washington’s counsel in the meantime was that “C——had better reside at New York, mix with, and put on the airs of a Tory to cover his real character, and avoid suspicion.” Indeed, as the “temper and expectation of the Tories and [Loyalist] Refugees is worthy of consideration,” Woodhull might find it “political and advantageous” to form “an intimacy with some well informed Refugee.”1
During the spring of 1779, Woodhull heeded his chief’s advice and submitted a nice amount of useful intelligence. On March 17, Woodhull followed up his January warning that Clinton was building transports with the news that the general was touring eastern Long Island to hire crews of Loyalists in preparation for a strike to “plunder and distress” the Connecticut seaboard.2 Washington pondered this. A week later, he informed General Joseph Reed that “the enemy have some enterprize in view.” According to “one of my most intelligent correspondants,” he should be on his guard everywhere, for “General Clinton (under pretence of visiting the troops) is now at the east end of Long Island,” and will doubtless “attempt something that will give éclat to his arms.”3 Washington advised General Israel Putnam in Connecticut to reinforce the militia and have them man the fortifications along the coast.4 On April 1, Washington instructed his soldiers to come off full alert, as the raid now looked less likely: “Sir Henry Clinton is returned to New York … and accounts from New York mention that troops have been relanded upon Long Island which are thought to be those which went eastward.” Even so, “except there should be certain intelligence obtained from Long Island [i.e., from Woodhu
ll] that the matter which has been in agitation is entirely over,” commanders must not take their troops fully off guard.5
Apart from his success in passing on that intelligence, April was a cruel month for Woodhull. Never comfortable as a spy, the stress of his double life was stretching Woodhull’s nerves, and he confessed to Tallmadge that the only reason he remained in the business was because he hoped “it may be of some service towards alleviating the misery of our distressed country, nothing but that could have induced me to undertake it, for you must readily think it is a life of anxiety to be within … the lines of a cruel and mistrustful enemy that I most ardently wish and impatiently wait for their departure.”