Angela shivered. ‘So if we’d taken a later flight somewhere, or the men in Iraq had moved quicker, the Italian killers might have been waiting for us at the airport?’
‘Yes. That was why I was so keen to get out of Kuwait as quickly as possible.’
For a few minutes, they were both silent, the gravity of the situation weighing heavily on them. Then Bronson glanced across at his former wife who was still staring straight ahead through the windscreen at the unwinding road in front of them.
‘I’m starting to have second thoughts,’ he said.
‘What about?’
‘What we do next. I think it’s time for Plan B.’
29
Khaled took off his reading glasses and rubbed his eyes. He had been studying both the photographs of the inscription found in the underground temple and the written copy he had made of the text. He felt as if he had been looking at the enigmatic characters for the entire day, though in fact he’d only been sitting there for about three hours. But the text — such as it was — still made absolutely no sense to him. It was clearly encrypted, because no combination of letters in the text formed a recognizable Latin word, and he was reasonably certain that it was Latin.
He tried character shifting, a refinement of basic Atbash, which meant starting the first letter of the reversed alphabet at some random point, which gave a further twenty-six possible ciphertexts to try, none of which had worked. He’d then again tried character shifting but using the ciphertext alphabet written forwards rather than backwards. That produced another twenty-six possible combinations, and once more none of them had worked.
All he’d so far managed to establish was that something rather more complex than Atbash had been used to encrypt the message. And, realistically, about the only two possibilities were that a code word of some sort had been included within the ciphertext, or perhaps the cipher itself was entirely reliant upon a number of code words instead of a reversed alphabet.
Both techniques were known to have been used as early as the mediaeval period, and because of one single piece of information — the date of an event that took place in Western Europe — which had been included within the text on the parchment in his possession, Khaled knew that the inscription could not possibly have been written prior to the early part of the fourteenth century. So a cipher that utilized code words was entirely feasible.
The problem, obviously, was working out which word, or combination of words, had been used.
Khaled was by no means an expert cryptographer. Because of his job he had become familiar with the basic techniques, and rather than trying to peer into the mind of a mediaeval scribe to guess what names or words he might have chosen almost half a millennium earlier, he decided to approach the problem from a different direction, by using frequency analysis.
In English the six commonest letters in order are E T A I O N, the list including four of the five vowels. So if a piece of English-language text, encrypted using a simple letter-substitution code, is analysed, whatever letter occurs most frequently in that piece of text is most likely to represent the letter ‘E’, and the second commonest the letter ‘T’, and so on.
Latin was obviously going to be different, but Khaled had very quickly located an analysis that had been performed on a number of pieces of mixed-genre Latin texts amounting to nearly 300,000 words. The results, posted on the Internet, were in fact not that dissimilar to English, the commonest letters being I E A U T S, these six letters together amounting to over 55 per cent of the text, and the first three accounting for almost 32 per cent of the total. Obviously different types of text would have produced different results, but he felt this information was a good guide.
He took a fresh sheet of paper, printed a photograph of the entire inscription and worked his way methodically through the first half-dozen lines, crossing out each letter on the photograph as he noted it down on the paper, and putting an additional line beside each letter as a simple counting mechanism each time it reoccurred. That produced a kind of table of popularity for the letters used on the inscription, which he cross-referred to the list on the Internet and pencilled in substitutes accordingly.
That, infuriatingly, didn’t make any better sense, and for a couple of minutes after he had completed the letter substitution Khaled just stared blankly at the paper in front of him.
And then it was as if a light began to dawn, and he suddenly realized that as well as encrypting the original text, the man who carved the inscription had added one other simple layer of complication. A complication that in fact Khaled should have guessed a lot earlier, purely because it was an inscription.
Once he saw how the original text — and he now knew that it definitely was Latin — had been encrypted, deciphering it was simply a matter of time, and within a little over half an hour Khaled was looking at the original plaintext. Or at least, he was looking at the original plaintext of roughly half the inscription.
But the lower half of the encrypted text stubbornly refused to yield to anything he tried, even frequency analysis, and he guessed that some entirely different encryption method must have been used. This was interesting in its own right, and attracted his professional attention, but on a personal note it was just extremely frustrating. There had to be, he assumed, some clue or indication in the section he had managed to crack that would indicate how to decipher the rest of it.
And then he saw something that he hadn’t noticed before. There was a piece of the inscription that he had ignored because he’d assumed it was simply decoration, placed there by the sculptor in an attempt to make a rather uninteresting-looking inscription slightly more attractive. And he’d been wrong.
Between the section of the inscription that he had managed to decrypt and translate and the part that had so far resisted his efforts was an incised line made up of a series of small square crosses, all apparently identical. But now he could see how important the line was because it separated the two parts of the text. It was another indication — if any was needed — to show that a different decryption technique would be required to display the plaintext of the characters that had been chiselled into the rock to form the lower part of the inscription.
As a final check, he then effectively reverse-engineered the part of the cipher he’d successfully cracked and confirmed that it was Atbash, but with the addition of two proper names, one at the beginning of the reversed alphabet and the other at the end, giving multiple options for the encryption. And the names the author had chosen, with hindsight, were entirely predictable, given what he already knew — or at least guessed — about the origin of the inscription and the men who had ordered it to be carved.
He spent another hour trying every combination of names and variants he could think of, attempting to crack the lower section of the encrypted text, but got absolutely nowhere. And there appeared to be no clue within the plaintext he had written out to suggest what word or words might have been used to form the cipher. It looked very much as if the second section needed something else to allow it to be decrypted, and that did make sense of a phrase that appeared towards the end of the plaintext.
But he wondered if he even needed to decrypt the remainder of the text, because the information contained within the inscription seemed enough in its own right, within certain limitations. Although the statement he had decoded was clear enough, it was also in one sense frustratingly obscure.