Russell Holt writes:
BTW, does anyone have references to statistical comparisons of key searches in "real world" settings? Eg, random starting locations versus random keys .. etc..
E(time) with all trial keys distinct: 1/N Sigma^N i = N/2 = 2^55. E(time) for random trials: 1/N + (N-1)/N (1 + E(time)) so E(time) = N = 2^56. Centralized keyspace handling is a clear lose, IMO. The RC4-40 project had server problems; scaling up by another 2^16 would be non-trivial. Then there are problems with malicious searchers, corrupt servers, flaky networks... all this for a factor of two? (And, admittedly, the reduction in variance.) To rain on the parade a bit, what this discussion says to me is that most applications of DES are safe from net cracking for a few years yet... -- Eli Brandt eli+@cs.cmu.edu