On Wed, Sep 25, 2013 at 10:11:33AM -0400, John Young wrote: > Is this conclusion still valid? If so, what could be done to restrict traffic > volume to assure unbreakablility? And how to sufficiently test that.
You need to be able to estimate the rate of information leakage. This seems to be related to measuring RNG entropy, and is considered a hard (perhaps hopeless?) problem. > Presuming that NSA and cohorts have investigated this effect. It seems to be possible to construct a family of cyphers based on PRNGs with Very Large Internal State (the shared key is the state) that asymptotically approach (in a special/edge case are exactly equivalent to) one-time pads. You'd tap them for XOR with cleartext through a relatively small (=plenty of hidden state) window (not necessarily contiguous) and use enough iteration rounds to make sure the information has has a chance to propagate through the computational volume. Edge cases are low-dimensional CAs with a suitable rule, which should be easiest to attack. Higher-dimensional CA analoga have a lot of neighborhood cells, and their map to address space looks like a small world network, so state mixes quite rapidly, requiring fewer rounds. Whether making neighborhood itself random versus orthogonal is helping or hindering things is not obvious. Whether to make the neighborhood itself subject to change at each or N rounds is helping or hindering things is not obvious. The actual problem is to build them provably hard to reverse, and rekey (though a secure channel, natch) before they leak enough information about their inner state to be attackable.
signature.asc
Description: Digital signature
_______________________________________________ cryptography mailing list cryptography@randombit.net http://lists.randombit.net/mailman/listinfo/cryptography
