Re: Is MD5 outdated ?

From: Mxsmanic (mxsmanic_at_hotmail.com)
Date: 10/06/03 

Date: Mon, 06 Oct 2003 01:41:56 +0200

David Taylor writes:

> Your statement contradicts itself. If a single-bit change of the message
> results in a change typically changes 50% of the bits in the hash, how can
> "any change in the hash" (say 50% of the bits changing) typically
> correspond to a change of about 50% of the bits in the message?

Any change in a message--be it one bit or a trillion bits--will produce
a random hash bearing no relation to the hash of the previous message.
Given this, the two hashes will inevitably differ in about half of their
bits, on average.

Conversely, any two hashes--whether they differ in one bit or in all
bits--will typically correspond to two random messages, and two random
messages will typically differ in about half their bits.

So there is no contradiction. 

-- 
Transpose hotmail and mxsmanic in my e-mail address to reach me directly.

Relevant Pages

  • Re: Is MD5 outdated ?
    ... a properly-designed message digest algorithm will make ... every bit in the hash depend on every bit in the original message. ... correspond to two messages that differ in about half of their bits. ...
    (sci.crypt)
  • Re: Is MD5 outdated ?
    ... > every bit in the hash depend on every bit in the original message. ... > correspond to two messages that differ in about half of their bits. ... Utter nonsense, and different nonsense from ...
    (sci.crypt)
  • Re: Is MD5 outdated ?
    ... Bryan Olson writes: ... Each message digest is completely independent of every other ... the hash is no different from a n/2-bit change. ... the hash may correspond to any other message, ...
    (sci.crypt)
  • Re: Is MD5 outdated ?
    ... > How many digests have you generated that differ from each other ... My proof considered *all* one bit changes of the digests. ... changes and checked whether they correspond to some bit-changed ... "And a one-bit change in the hash will typically correspond to a ...
    (sci.crypt)
  • Re: Is MD5 outdated ?
    ... > I'm surprised that you argue with it, though, since it is logically ... When the hash is a random function of the message, ... > correspond to a change of about 50% of the bits in the message. ...
    (sci.crypt)