1997-02-07 - Re: 40-bit RC5 crack meaningless??

Header Data

From: nobody@REPLAY.COM (Anonymous)
To: cypherpunks@toad.com
Message Hash: 7a5d33667e0cc8b2d99218a57171ecfb2a31f390127691d0067a517acf0a6a12
Message ID: <199702071508.QAA05145@basement.replay.com>
Reply To: N/A
UTC Datetime: 1997-02-07 15:19:01 UTC
Raw Date: Fri, 7 Feb 1997 07:19:01 -0800 (PST)

Raw message

From: nobody@REPLAY.COM (Anonymous)
Date: Fri, 7 Feb 1997 07:19:01 -0800 (PST)
To: cypherpunks@toad.com
Subject: Re: 40-bit RC5 crack meaningless??
Message-ID: <199702071508.QAA05145@basement.replay.com>
MIME-Version: 1.0
Content-Type: text/plain


Over on sci.crypt, Paul C. Kocker <pck@netcom.com> gave a clear and
confident response to a query about the statistical difference between a
brute force attack on a known Russian or English text, versus a similar
attack on cyphertext with a known-plaintext sample (Strassmann's tell-tale
Clue #3.)

Said Kocher:

The difference is negligable.  With English text encrypted under
a 64-bit block cipher, you know that the most significant
bit of each of the 8 bytes in the block should be zero.  For a
wrong key, there is a 255/256 probability that at least one of
these bits will be nonzero, allowing immediate rejection of
the key.  Keys which do produce all zero bits get tested on
additional blocks until the key is either deemed correct or
rejected.  The extra overhead per wrong key is the sum from
i=1 to infinity of i*(1/256^i), or under 0.4 percent.

In practice, the slowdown is actually a couple of percent, since
it complicates the skip-the-last-Fiestel-round optimization. Also,
the 1/256 case requires running extra code, which can fill the
microprocessor's cache with code which isn't part of
the main loop, slowing things down a bit when the computer
goes back to the main search.

Cheers,
Paul

____________________________________      http://www.cryptography.com
Paul Kocher (pck@netcom.com)       |     Voicemail: +1-(415)-354-8004
Crypto consultant                  |           FAX: +1-(415)-321-1483







Thread