Re: advice sought on key/data histogram analysis of rijndael/128 and serpent
From: lkcl (lkcl_at_lkcl.net)
Date: 17 Oct 2005 16:25:03 -0700
okay, well, every 16 inputs i change the set of keys used, by doing
the following in the outer loop:
for (k = 0 to 2048000)
if (k % 16 == 0)
for i=0-127, j=0-127
hist[i][j] += analyse(make_input(k), keyset[i], 1<<j);
where create_keyset_with_seed(k) does this:
sha1_update(ctx, key, 16)
sha1_update(ctx, (int*)&k, 4)
keyset[i] = sha1_final(ctx)
yes i know i could simplify the key create function by just
calling sha1_update with the ints 0..127 and k :) heck
even a _byte_ 0..127 would do.
testing with 2048 loops (yes, very quick, i know) and therefore
2048/16 sets of keys still shows correlations.
you implicitly ask what am i trying to prove by doing this test?
well, what i suspect is that, if my test is correct rather than
based on some mistakes or turns out to have false assumptions,
that the test could be used as the basis for an alternative
type of differential analysis.
i only have some ideas on what to try out, so far on that score:
i haven't quite thought out the full implications, yet.