THE CODEBREAKERS
Page 37
Page 75 of the Baravelli commercial code, used in the Panizzardi telegram
The cryptanalysts recognized Panizzardi’s intermixture of one-, two-, three-and four-digit groups in a single message as that of an Italian commercial code published earlier that year in Turin by Paolo Baravelli, an engineer. This code, entitled Dizionario per corrispondenze in cifra, was constructed in four sections: table I, in which vowels and punctuation marks were represented by single digits from 0 to 9; table II, in which consonants, grammatical forms, and auxiliary verbs were represented by pairs of digits; table III, consisting of syllables indicated by three-digit groups; and a vocabulary proper, in which words and phrases were represented by four-digit groups. Some four-digit groups were left blank so the user could insert terms he found necessary.
The codebreakers furthermore remembered the Baravelli code from an amusing incident of a few months earlier. In June, a mysterious correspondence had been daily burning up the wires between the Count of Turin, a nephew of the king of Italy, and the Duchess Grazioli, a tall and voluptuous Italian living at the Hotel Windsor in Paris. Colonel Jean Sandherr, the dull and stolid head of French army intelligence, thought it smelled of espionage; Maurice Paléologue, an assistant to the Foreign Minister, whose duties included overseeing the cryptanalytic office, said that it gave off only the perfume of romance. (Paléologue was to become France’s World War I ambassador to Russia and a member of the Académie Française.) Soon Sandherr burst into Paléologue’s office with a small, flat, highly scented book. It was a Baravelli code. One of Sandherr’s agents had stolen it from beneath a packet of the duchess’s handkerchiefs while she was at the races. Two days later, Paléologue brought over the translations. They expressed, he said, nothing but “simple, elemental, natural feelings. However, one four-figure sequence which recurred in most of the telegrams remained indecipherable [presumably because it stood for a blank that the couple had filled in themselves]. All that our decoders were able to suggest was that the apocalyptic number stood for something extraordinary, unforgettable and sublime!”
This experience—in its own way extraordinary, unforgettable and sublime—had taught the cryptanalysts that, to achieve secrecy in a volume that was on public sale, the Baravelli employed an artifice common among commercial codes of the day. On each of the 100 pages devoted to the vocabulary, the words, phrases, and blank spaces were distributed in two columns of 50; each was assigned a printed number from 00 to 99. Each page was also numbered at its lower outside corner by two small printed digits that ran from 00 on the first page of the vocabulary to 99 on the last. The user could either superencipher these, or he could fill in his own page number following the large printed “Pag.” at the top of each page. He was to combine these two digits with the two digits for the words into the four-digit figure of the vocabulary section.
Similar arrangements were provided in the other sections. Table III (syllables) was set up in precisely the same way, except that the ten pages were given single instead of double digits. Table II (grammatical forms, consonants) was divided into ten groups of ten elements; the first number, indicating the group, was provided with a dotted line on which a substitute could be written; the second number, indicating the element in that group, was printed. The ten numbers of Table I (vowels, punctuation) were each preceded by a dotted line.
The Foreign Ministry cryptanalysts undoubtedly attempted to read Paniz-zardi’s message using all printed numbers—in other words, without any substitutes for the page or group numbers. The attempt yielded:
This gibberish showed that Panizzardi had made use of a superencipherment. It was the task of the French cryptanalysts to determine it—a task made more difficult by the fact that this message was the first sent by Panizzardi in this particular system.
They were abetted, however, by the peculiar construction of the Baravelli code, in which the portion of each placode number representing the line remains invariable because it is printed. With this as a start, and with the agitations of the Dreyfus disclosures, the cryptanalysts had little trouble in determining that the arrested man’s name figured in the cryptogram. The plaintext elements available in the Baravelli code permit the word Dreyfus to be broken up for encoding in only one way: dr, e, y, fus. Both dr and fus are found in Table III: page 2, line 27 for dr; page 3, line 06 for fus. The y is found in Table II, group 9, line 8; and the e in Table I, line 1. In placode form, then, Dreyfus would be 227 1 98 306.
Now the Panizzardi telegram includes a similar sequence of codegroups composed of three, one, two, and three digits: 527 3 88 706. Furthermore, the numbers in this sequence that presumably represent the lines—27, 8, and 06 (omitting the single digit from Table I)—are identical with those for Dreyfus. Obviously, then, the sequence 527 3 88 706 represented Panizzardi’s encicode for Dreyfus. From this, the cryptanalysts could see that the digits representing lines were not enciphered. They also had ascertained the encipherment of two of the Table III pages, of one group in Table II, and of a line of Table I.
With this as a start, the Foreign Ministry cryptanalysts produced—perhaps by the very next day—a preliminary decryptment that read: “Arrested [is] Captain Dreyfus who has not had relations with Germany…. ” This text, highly hypothetical, and in which the only certain word was Dreyfus, was shown to Sandherr, who was in intimate and frequent contact with the Foreign Ministry cryptanalysts. He was immediately interested, for the telegram bore on the guilt or innocence of the central figure of a sensational scandal involving his service. By Tuesday, November 6, the cryptanalysts had perceived that the group 913, which they had translated as part of arrestato in their first trial, was just Panizzardi’s serial number, and they had reached a solution which they considered exact, except for the ending: “If Captain Dreyfus has not had relations with you, it would be wise to have the ambassador deny it officially. Our emissary is warned.” These last words seemed to hint darkly at Dreyfus’ guilt, and, though that was the very part that was conjectural, Sandherr, who was disposed to think Dreyfus a traitor, borrowed the cryptanalyst’s worksheet, with its successive hypotheses stacked up beneath each codegroup and with question marks advertising the conjectural nature of the final four words, uffiziale rimane prevenuto emissario. He reported on it to his superiors, remarking to Charles Le Mouton de Boisdeffre, the chief of staff, “Well, General, here’s another proof of Dreyfus’ guilt.” Sandherr had a copy made of the worksheet, which du Paty de Clam studied with interest, and then returned the original to the Ministry of Foreign Affairs.
By the following Saturday, November 10, the cryptanalysts had finally cracked the system of encipherment used by Panizzardi, recovered the placode equivalents, and established the definitive text of the cryptogram. It read, with the placode equivalents:
first placode digit 0 1 2 3 4 5 6 7 89
second encicode digit 9 8 7 6 5 4 3 2 1 0
second placode digit 0 1 2 3 4 5 6 7 8 9
first encicode digit 1 3 5 7 9 0 2 4 68
Or, in English: “If Captain Dreyfus has not had relations with you, it would be wise to have the ambassador deny it officially, to avoid press comment.”
Panizzardi proved to have used a relatively simple system. Line digits were not touched. The two digits of the vocabulary’s printed page numbers were transposed and given substitutes according to these alphabets:
Thus the placode 1336 of capitano became 7836, and the 3306 of evitare became 7606. The second of the above two alphabets also served to encipher the page, group, and line numbers of Tables III, II, and I, respectively. Thus the placode page 3 of the 306 of fus became 7 to give encicode 706; the placode of 98 of y became 88; and the placode 1 for e became encicode 3.
This version, which in no way implicated Dreyfus, was communicated to Sandherr by Paul-Henri-Phillipe-Horace Delaroche-Vernet, a 28-year-old subordinate of Paléologue’s who served as liaison between the cryptanalysts and the Army (and who, in 1908, became chief of the Bureau du Chiffre, holding the post for five years). Sa
ndherr was not pleased with this new version, and he transmitted it to his chiefs with the observation that “with Foreign Affairs, you can’t always be certain about these things—they lack a little precision.” Then one of his subordinates, Commandant Pierre-Ernest Matton, a 39-year-old artilleryman who was the army liaison with the Foreign Ministry, had an idea that would lay all skepticism to rest once and for all. He would trick Panizzardi into sending a telegram whose contents were known to the French; the solution of this would verify or refute the cryptanalysis of the Dreyfus message. Sandherr approved.
The Panizzardi telegram, with correct solution inserted
Matton framed a message with words chosen from pages in the Baravelli code whose numbers were critical in the encipherment and with proper names that could be divided in only one way. He artfully composed it to be so important that Panizzardi could not ignore it, and so perishable that he would have to telegraph it to Rome. It told how “A certain Y, who is now at X, will leave within a few days for Paris. He is carrying some documents relative to the mobilization of the army, which he obtained in the offices of the general staff. This person lives on Z street.” One of the proper names was Schlissenfurt, which the Baravelli can handle in only one way. Matton had a double agent slip it to the Italian attache. Panizzardi fell for the ruse, encoded the message almost verbatim and wired it to Rome at 8:10 a.m. November 13. The usual routine brought the onionskin copy to the cryptanalysts of the Quai d’Orsay. They, not knowing that the Army had the plaintext, solved the message and passed the translation to the Army because of its military import. When Delaroche-Vernet brought it over, Matton says he interrupted the young official and said:
“ ‘Will you permit me? I am going to get the original.’ I went into my office and got out the piece I had written. It was word for word the dispatch they had deciphered. I told him, ‘You may be sure now that you have the encipherment.’ ” The text that exonerated Dreyfus was irrefutably correct.
So conclusive a proof of the solution’s validity would seem to have been unchallengeable. But this would be to reckon without the persistence and tenacity of those who felt Dreyfus guilty, or who thought it better to convict Dreyfus wrongfully than to admit the Army had erred and open it to criticism. Boisdeffre and his fellow generals refused to allow the Panizzardi telegram into evidence at Dreyfus’ first trial, telling the prosecutor that the variations of the progressively more accurate solutions negated the telegram’s value as evidence. Dreyfus was found guilty of treason and interned on Devil’s Island.
But knowledge of the telegram could not be suppressed, and the anti-Dreyfus officers finally slipped a false and highly condemnatory version into the subsequent trials and appeals of the case: “Captain Dreyfus arrested; the Minister of War has proofs of his relations with Germany. Parties informed in the greatest secrecy. My emissary is warned.” This text appeared as No. 44 in the so-called Secret File; it had been dictated from memory by du Paty de Clam, who seems to have fabricated it from among the various hypotheses he saw on the original worksheet borrowed by Sandherr. This version invalidated itself by the simultaneous presence in it of both proofs and relations. Both of these stood on line 88 of different pages of the Baravelli code (provi on page 71, relazione on page 75), and so both obviously could not be the plaintext equivalent for the telegram’s encicode 0288.
Faced with such difficulties, Sandherr checked secretly with Commandant Munier, a former secretary of the Military Cryptography Commission, who obliged with some obscurantist professional cant to indicate that the wrong version was cryptologically correct. Finally, on April 27, 1899, Paléologue and two officers deciphered an authenticated copy of the original Panizzardi telegram for the Cour de Cassation. The result was, of course, the same as the final version decrypted by the Foreign Ministry cryptanalysts—the version in which Panizzardi by implication disclaimed any contact with Dreyfus. At last the correct solution entered the record. This alone did not exonerate Dreyfus; it was to take seven more years before he was to receive justice, reinstatement, and the Legion of Honor. (In the interim, the true author of the bordereau was found to be Major Ferdinand Walsin Esterhazy; among his papers, seized on his arrest, were a number of cardboard grilles, presumably for cipher communications with the German military attaché.) But the demonstration of how the false solutions had been used to bolster the trumped-up anti-Dreyfus case helped clear the man on Devil’s Island. Du Paty de Clam, the officer who long ago had arrested Dreyfus, himself exclaimed on the importance of the telegram, though he intended his words as a condemnation. “This telegram,” he declared, “is, for me, the pivot of the affair.”
When the Dreyfus case was finally closed in 1906, holocaust was less than a decade away. During those years, as the tensions increased and the world girded for battle, cryptology received degrees of attention that varied from country to country according to their individual cryptologic traditions. France had the strongest tradition. The published literature from Kerckhoffs on reflected that nation’s profound understanding of the subject. Practical applications were amply demonstrated by the Panizzardi telegram, by French solution of the Italian Foreign Office’s most secret codes three and five years after the Panizzardi solution, and by France’s possession of the German diplomatic code, which enabled her to read critical German messages on the very eve of World War I.
Army cryptology was better than Bazeries’ rough tongue gave it credit for. The Military Cryptography Commission, which consisted of approximately ten officers chosen from among all arms who had shown an aptitude for cryptanalysis, tested systems proposed for use by the Army and studied cipher systems used by other nations, particularly Germany. The commission’s president was, in 1900, the inspector-general of the military telegraph services, General François Penel, and in that year a 37-year-old engineer, a graduate of the école Polytechnique, was attached to the general staff as adjutant to Penel and secretary to the commission. This was Captain François Cartier, who was to become the chief of the French military cryptologic bureau in World War I. Before that war, Cartier had drafted a memorandum on how to solve German Army cryptograms on the basis of the drill messages, prefixed ÜBCHI, that French radio stations had intercepted during German maneuvers. The commission obtained other information from spies, deserters, and recruits to the Foreign Legion. The members, who did their cryptologic work in their spare time and received extra pay for it, formed a core of cryptologists with valuable experience. All this gave France a preponderant cryptologic superiority in 1914.
The Germans, on the other hand, seem to have disdained studying cryptology. The Junkers felt that their armies could, as in 1870, overrun the French by sheer force of arms. Cryptanalysis played a minor role in intelligence, since it required tapping telegraph wires to intercept texts. The Germans failed to foresee how much radio would be used and how much information would flow in its channels. Hence the general staff obtained what little it knew of other nations’ cryptography from its intelligence service and did not waste manpower on such frills as cryptanalysis. As for their own ciphers—were they not German? Which ended the discussion. And so German cryptology goose-stepped toward war with a top-heavy cryptography and no cryptanalysis.
Marching with them in the parade of cryptologic ignorance were most of the other armies of Europe. England had done little more than distribute field ciphers to its tiny Army; Italy was about as interested in cryptology as it was in, say, social reform. There was no organized military cryptanalytic bureau in any country except France—and Austria-Hungary.
Perhaps the Hapsburg background in cryptanalysis, stretching to the Geheime Kabinets-Kanzlei, had conditioned Austria-Hungary to think in those terms. In 1908, she intercepted Italian radiograms, and again in 1911, when they “rained” from the sky at the outbreak of the Italo-Turkish conflict over Tripoli. The alert Colonel Max Ronge, later head of the Nachrichtendienst, or intelligence organization of the general staff, saw the opportunity. In November of 1911, he instituted a cryptanalytic
bureau with Captain Andreas Figl as its chief. The staff analyzed Russian cryptograms, which proved very difficult under peacetime conditions, and Ronge purchased some Italian ciphers as a headache-preventative for his cryptanalysts.
He was not the only buyer. In the E. Phillips Oppenheim world of prewar Eastern Europe, codes and ciphers were bid up and up like speculative shares in a stock-market boom. Heading the list were those of Austria-Hungary, which, as the crossroads of Europe, was a virtual ants’ nest of espionage.
According to one story, a young and remarkably attractive Italian “countess,” who had become friendly with a lieutenant in the Austro-Hungarian headquarters, sneaked a copy of a red-bound code from an open safe there and replaced it with a book that looked remarkably like it—but that had only blank pages! Some time later a code clerk pulled the book out to use it and discovered the substitution. Shock waves of consternation shook the general staff. Frenzied manhunts began. Not until the Russian attaché had laughingly told one of the staff officers that he had been offered the code, but had turned it down because the 400,000 rubles asked was too much, did the Austrians trace who had taken it.