by James Becker
“There obviously isn’t a photograph of the original document,” he said, pointing at the printed pages. “It was probably a piece, or maybe even several pieces, of parchment or vellum. They probably took it out of the secret compartment at the bottom of the chest and simply transcribed it,” he rationalized. “Perhaps it was far too big, or too awkward a shape, to photocopy.”
“So, what’s on this sheet? Just single letters followed by a basic numbering system?”
“That’s exactly what it is,” Benelli replied. “It looks to me as if they got nowhere with trying to crack the encryption using plain Atbash or word substitution, and this is their stab at frequency analysis.”
“At bash? What’s that?”
“Don’t worry about it. It’s just a basic letter-substitution code that was quite often used during the medieval period. The other pieces of encrypted text relating to this quest all used variants of Atbash, and this probably does as well. It’s just a matter of working out the code words that were used to encipher it in the first place, and that’s why the English couple seems to have been trying frequency analysis. You calculate the most frequently used letters in the ciphertext and then substitute the commonest letters in use in Latin—and I’m quite sure that this is Latin, at this period of the Middle Ages.”
The communications officer didn’t look any the wiser, but just nodded and then left Benelli’s office.
It was late, and he was already tired, but the espresso helped, at least a little. For the moment, he put the frequency-analysis sheets to one side and concentrated instead on what was clearly the transcription of the hidden document or documents, the parchment or whatever the English couple had found. Maybe he could try some different possible code words, the names of people or places most intimately associated with the Knights Templar, names that might not be familiar to Mallory or Jessop. Without the slightest shadow of doubt, the Dominican order knew far more about the order of heretics than anyone else—after all, they had been on the trail of the Templar treasure for over seven centuries—so this seemed to Benelli to be a reasonable idea. And even if it didn’t work out, he could still go the frequency-analysis route.
Just before three o’clock in the morning, he gave up. He was so tired that the letters on the sheets of paper seemed to be blurring and running into one another, and he was constantly making mistakes as his concentration slipped. He knew that he had not achieved anything, and wasn’t likely to until he had managed to get some sleep.
But there seemed little point in returning to his apartment in Rome. He was so tired that driving himself was probably not a good idea, and trying to get a taxi to take him back there at that time of night would probably prove futile. There was a very basic bedroom located in the basement of the building that was intended for exactly this kind of situation, and at that moment the thought of the monastic solitude, bare walls, and hard mattress on the single bed in the room that awaited him seemed infinitely appealing.
Benelli shut down his computer, locked the printed sheets of paper away in his personal safe, went into the communications suite, and told the officer what he intended to do. He also asked the officer to make sure he was awake by seven thirty so that he would be able to brief Silvio Vitale when he arrived at about eight.
He walked into the bedroom, took off all his clothes apart from his underpants, and slid between the sheets, not even bothering to wash or clean his teeth. Less than two minutes after he put out the light, he was sound asleep.
10
Dartmouth, Devon
Things hadn’t gone much better in Devon. Robin and Mallory had continued working on the frequency-analysis sheets until late in the evening, when Robin had virtually lost patience with the tedious process.
“There has got to be an easier way of doing it than this,” she snapped, tossing her pencil down and leaning back in her chair. “Isn’t there some blasted computer program that you can use?”
Mallory stared at her for a few moments.
“Oddly enough,” he admitted somewhat sheepishly, “bearing in mind what I do for a living, I never thought of that. I’ll have a look.”
While Robin went out into the small kitchenette to make yet another coffee, Mallory performed a quick Internet search, and by the time she came back carrying two mugs, he’d downloaded three fairly small and quite simple programs.
“I’m not entirely sure how much help these three programs will actually be,” he said, “but we can certainly give them a try. One of them is specifically designed for cracking a single transposition monoalphabetic substitution cipher, which is just a fancy way of describing a Caesar or Atbash cipher. The second one is a frequency-analysis cracker, and that might be slightly more useful, because although we’re almost certainly looking at a substitution cipher, as I said right at the beginning, it certainly isn’t a plain vanilla Caesar cipher. We’ve already tried every possible shift, with the alphabet reversed as well, and made no progress whatsoever.”
He pointed at the sheaf of papers in front of him.
“Just like before,” he added, “whoever encrypted this text didn’t use the straight alphabet for the decryption code words, but instead picked a combination of words, probably with at least fifty-two letters in total, so there would be at least two solutions to every possible letter in the ciphertext. And that increases the number of possible decodes enormously.”
Robin nodded.
“I know, and that’s what’s been really irritating me. Just pick out any three letters from this lot,” she said, pointing, “like these, for example. G, C, D. Because the letter C occurs quite frequently in the encrypted text—we’ve already established that through the frequency analysis we’ve done so far—it’s most likely to be one of the vowels, which in the Latin alphabet means it’s most probably I or E. If it was a word in the English language, that could decode as almost anything from FIG through to SEX, and until we could work out one or both of the other letters, we’d be going nowhere.”
“You’re right. It’s really frustrating.”
“You said there were three.”
“Sorry?”
“Three programs. You said you’d downloaded three programs. What does the third one do?”
“Oh, yes,” Mallory replied. “That’s something a bit different. It’s a word-pattern analyzer. If you’re trying to break a monoalphabetic ciphertext, and you know a couple of the letters in a particular word, it will do a dictionary search and bring up all the words in the chosen language that match the shape of that word.”
“You’ve kind of lost me now.”
“I’ll give you an example. Suppose you have an encrypted word like KPCCRRF, and you believe that the letter P decodes as A. You feed the encrypted word into the program, which converts it into a numerical equivalent, in this case 1233445, so you have two unique letters, followed by two sets of two repeated letters, and then another unique letter. So, numerically, that is the shape of the word. Then you replace the number 2 with the letter that you believe was represented by the encrypted letter P, and then just look up the result. What the program does is check every possible word in the selected language where the second letter is A and that’s followed by two pairs of letters. In English, there are only four possibilities: BALLOON, BARROOM, BASSOON, and RACCOON, and in this case using that software would have positively confirmed that the ciphertext R is the plaintext O, and given only two possible plaintext decodes for the first and last letters of the word.”
“I see what you mean,” Robin said, “and that’s really rather clever. But it would only be any use to us if it works with languages other than English.”
“That’s not a problem. You can upload a word list or dictionary for almost any language, including Latin. And you’re right: it is clever, but we’ve got two obvious limitations. First, we don’t yet have a single confirmed plaintext letter and, second, and just as important, because
the original text is written without any breaks—the scriptio continua that was used—at the moment we have no idea where any of the words start or end. So we’ll need to do quite a bit more work before this clever program will be any use to us. But now that I’ve downloaded these programs, we might as well give the frequency-analysis calculator a try. The good news is that it’ll produce the results a hell of a lot faster than we’d be able to do using a pencil and paper.”
“And the bad news?” Robin asked. “I assume there is some bad news?”
Mallory nodded.
“Kind of, yes. To get the program to work, I’ll have to type the original ciphertext into my laptop. The handwritten sheets aren’t going to be good enough to scan in and then use an OCR—optical character recognition—program to read them.”
“I do know what OCR stands for, thank you. You’d better get started, then. I’m going to finish this coffee, make sure the outside door’s locked, and have a shower. I feel dusty and grubby. Then I’m going to bed. I expect you to follow me there before too long.”
“Trust me,” he said, “I’ll be as quick as I can.”
Transcribing the ciphertext wasn’t difficult, but it was extremely tedious, because Mallory knew that he had to check every single letter that he typed to make sure that he didn’t miss any out or press the wrong letter on the keyboard. When he heard the shower switch off, he was only about halfway down the first of the three sheets that they had prepared.
He glanced at his watch, saved the document he was working onto the hard drive, and made a separate backup copy on a memory stick, then closed the lid of the computer and headed for the bathroom himself. He’d finish the transcription in the morning.
11
Via di Sant’Alessio, Aventine Hill, Rome, Italy
Roman Benelli was up, washed, shaved, and dressed, albeit in precisely the same set of clothes he’d been wearing the previous evening, by a quarter to seven, and was back at his desk and looking again at the printed images in front of him five minutes later.
As dawn broke gloriously—and invisibly, to Benelli in his underground office—over Rome, the pages didn’t look any better or more hopeful to him than they had before. None of the Templar-specific names and words he had tried had worked, and after inputting another dozen or so as code words and then trying an Atbash decode of the first line of the text—all of which achieved precisely nothing—he came to the reluctant conclusion that the only way he was going to solve the decryption would be to apply frequency analysis, tiresome and time-consuming though that exercise would be.
When Silvio Vitale pushed his way into the office without knocking—as usual—just after eight, Benelli was already hard at it.
“This is the data sent by the tertiary in England?” Vitale asked, as Benelli stood up to greet him.
“Yes. I told the communication section to send a copy to your computer as well.”
“I know. I saw that when I arrived, but I’ve not looked at it yet. What I want to know is what you’re doing about it.”
Benelli explained that he had tried a large number of different code words that were directly relevant to the Knights Templar, but without any success at all.
“There appears to be no indication within the document about how to decode the text. Sometimes a person carrying out an encryption will include a hint in plaintext, or possibly a series of initials that suggests what the decryption key might be.”
“You mean like the letters P, C, C, T, S could mean Pauperes commilitones Christi Templique Salomonici, the original Latin name of the order of heretics? And those could be the code words needed?”
“Exactly,” Benelli replied. “It’s clear from the pictures obtained by the tertiary—or rather by the agent he used—that there was a concealed space inside the wooden chest.” He picked up one of the photographs. “This piece of wood obviously concealed a cavity at the bottom of the chest, which looks to me as if it’s about two centimeters deep. I assume that whatever the British pair found was inside it.”
Vitale nodded.
“But there’s no picture of whatever they recovered from it? Or have I missed that?”
“No, there’s no photograph of it, so we don’t know for sure what it was, but it was most likely a piece of either parchment or vellum—probably vellum, as it was an important document.”
“What’s the difference?”
Benelli looked slightly surprised by the question.
“Very little,” he replied. “In fact, they’re essentially the same thing. The word parchment refers to an animal skin that’s been prepared for writing on, and vellum was the highest-quality skin available. The word is derived from the Latin vitulus, meaning a calf, or more likely vitulinum, which literally translates as ‘made from a calf.’ And the finest-quality vellum of all was uterine vellum, made from the skin of calf fetuses, though there’s some doubt about whether that was ever used during the medieval period. Parchment and vellum were both prepared in the same way. First the skin was flayed. Then the hair was removed with a kind of lime solution, often made from rotting vegetable matter. That part of the process usually took a few days, but the people doing it had to take care with it because if they left the skin for too long in the liquid it would end up being too soft to be usable. Once that dehairing phase had been completed, the skin was washed, mounted on a stretching frame, and left to dry in the open air.”
“Interesting, but not helpful, Benelli. I’m not interested in the mechanics of the process, only in the result, in what was actually on the vellum or parchment or whatever it was. So, you haven’t got a picture of it, but you have got a transcription?”
“Yes. That was what I showed you first.”
“And nothing you’ve tried—none of the possible code words, I mean—has so far cracked the encryption?”
“Exactly,” Benelli said. “But at least we know that English couple is in the same position. They don’t know how to decrypt this, either.”
“How can you be sure of that?”
Benelli pointed at another of the sheets of paper on the desk in front of him.
“Because if they knew or had deduced the code words, there would be no need for them to be doing frequency analysis, and that’s obviously what these pages of letters are. They’re trying to work out which encrypted letters occur most frequently in the text so that they can use a Latin letter-frequency table to work out what the plaintext equivalent might be.”
“I know the theory, Benelli. And that’s what you’re doing as well?”
“Yes. I’ve tried everything else, so as far as I can see it’s the only possible way forward.”
Vitale looked at the pages.
“It’ll be quicker if you split the work. I’ll send in three or four other people to help you. Give each of them one section of the transcription and tell them exactly what you want them to do and how to do it. Any problems, let me know.”
Benelli nodded, and Vitale noticed the look on his face.
“What?”
“Frequency analysis may not provide the information we need, or not immediately.”
“Explain.”
“Looking at the analysis that I’ve done already has not produced the results I would have expected. Even in a small piece of encrypted text, just a couple of sentences, frequency analysis should quickly identify at least half a dozen of the commonest letters, and in Latin there should be two that are far more common than any of the others, the letters I and E. And that isn’t what I see here.”
Vitale leaned forward and looked at the page and the analysis that Benelli had completed on the first section of the text.
“There are two ways of doing this. The English couple obviously decided to analyze the entire piece of text. Noting the first letter of the first word, and then counting every subsequent recurrence of it. That works, but it means that until they reach the end
of the text the analysis remains incomplete. The quicker way to work is to take a much smaller piece of text and carry out the same analysis on that. Then you take the results and apply them to the rest of the ciphertext, and that often works. It may need refining to positively identify every single encrypted letter, but it will certainly generate most of the plaintext alphabet.”
“Presumably not in this case,” Vitale suggested.
“Not unless the section I picked is very different to the rest of it, no. I’m not seeing two letters appearing more commonly than any others. Instead, there are eight letters that seem to be dominating the table. And the most likely explanation for that is that we’re looking at a very complex Atbash cipher, where there weren’t just twenty-three or twenty-six letters in the code word or words corresponding to the Latin alphabet—depending on the period—but more like seventy. Maybe even a hundred. Let me show you what I mean.”
Benelli took a fresh sheet of paper, wrote out the normal alphabet in a single horizontal line, then wrote three horizontal lines of random letters below it.
“Now, if you look at the letter E,” he explained, “instead of there just being one equivalent letter in a single code word below it, you have a choice of three. That obviously increases the possible permutations for each letter, but to make it worse those three letters in the vertical line below the E also occur elsewhere in the code words. They have to, because there are only twenty-six letters in the alphabet and to have code words totaling seventy-eight letters inevitably means that there will be repeats. In this example I’ve made up, the three letters below the E are P, G, and R, and G can also represent the letters K and S. So if you look at the letter G in the ciphertext, the plaintext equivalent could be E, but it could also be K or S.”
“So what you’re telling me is that for every single word there could be three or four possible plaintext equivalents for each letter in the ciphertext.”