[rc5] rc5 crack schemes

Eric Gindrup gindrup at okway.okstate.edu
Mon Aug 25 12:12:41 EDT 1997

        You are probably right that there are effective and subtle methods 
     to prune the likely search keyspace.  I think neural netes would be 
     good at guessing likely candidates but that doesn't preclude them 
     guessing the entire keyspace in some convoluted order only to arrive 
     at the correct key just as the entire keyspace is exhausted...
        Further, I thought the RC5 effort wasn't about tricky techniques 
     that would only work for this encryption technique.  I thought it was 
     about proving that it isn't too hard to brute-force a decrypt of *any* 
     56bit-key encryption.  Because, hey, if a couple thousand geeks 
     running in their spare time can crack it in something approaching 
     reasonable time, dedicated budgets and hardware can do *much* better.
        I'd be more interested in bumping up the client so that it checks 
     two or three structurally similar keys at once.  The sort of thing 
     where the rotate sequence doesn't alter much between decrypts using 
     two different keys.  Probably though, the computational effort to find 
     such similar sets of keys would be prohibitive and representing the 
     sets would be "hard".
            -- Eric Gindrup ! gindrup at Okway.okstate.edu

______________________________ Reply Separator _________________________________
Subject: [rc5] rc5 crack schemes
Author:  <rc5 at llamas.net> at SMTP
Date:    1997/08/25 11:53

how about a variation on brute-force?
We shouldn't be doing a purely sequential scan, but using a bisection 
At the very least, I think that we should assign blocks from both ends 
of the keyspace, and go sequentially from there, instead of starting at 
the beginning.
Also, there are some other schemes. The military uses a scheme which can 
crack a lesser generation 128-bit crypt in approx. 25 minutes. (the gov 
thinks only the US mil has this technology so it likes the DES-56 for 
I've heard that it uses some sort of neural net + fuzzy logic in 
determining key probabilities, and testing the keys that are calculated 
to have a high probability of being correct.
Am I right or dead wrong?
To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the

To unsubscribe, send email to majordomo at llamas.net with 'unsubscribe rc5' in the body.

More information about the rc5 mailing list