On Thu, Dec 26, 2013 at 04:28:06PM -0300, andrew cooke wrote: > the way that you use the plaintext to avoid short cycles (the "output > number" etc) is worrying - it might open you up to a chosen plaintext > attack in some way. > > and thinking about chosen plaintexts - if you encode a message that is > all zeroes, what does that reveal? it seems like it might leak > information about the board you are using.
Ah, I hadn't thought about that. With the board example I use in my post, the 'random numbers' would be a constant stream of '38', as the bishop can't move out of the corner. Although a plaintext of all zeores isn't practical, it does demonstrate a serious flaw with the algorithm. > these aren't attacks, but they are obvious places where i (with, admittedly, > very little experience of attacking ciphers) would start. This is valueable information, that I'll take back to the drawing board, and see how I can address that problem. My biggest concern, and one I'm finding more and more difficult to escape, is the possibility of repeating cycles in the bishop's "walk". If the board values changed at the end of each algorithm, then it seems those cycles would not be an issue. Because the board is static, it's difficult, maybe impossible?, to come up with an algorithm that still produces a pseudorandom stream of numbers. I've thought of incorporating Blum Blum Shub into the algorithm, but then the cipher is getting decidedly difficult to execute by hand. -- . o . o . o . . o o . . . o . . . o . o o o . o . o o . . o o o o . o . . o o o o . o o o
pgpraalL7vkF3.pgp
Description: PGP signature
_______________________________________________ cryptography mailing list cryptography@randombit.net http://lists.randombit.net/mailman/listinfo/cryptography