Re: Haval hashing - how many passes?

From: Andy Ritchason (maxace_REMOVE_TEXT_AFTER_MAXACE_at_bellsoth.net)
Date: 06/14/03 

Date: Fri, 13 Jun 2003 19:56:07 -0400

I have to agree that Jimmy should use SHA-256 or SHA-128. MD4 is known to
have collisions., so if speed is an issue, try MD5. If you are just trying
to use the hash as an identifier, a single pass is enough. Info about MD4
collisions can be found at:
http://www.hack.gr/users/dij/crypto/overview/mdx.html

Andy

"Tom St Denis" <tomstdenis@iahu.ca> wrote in message
news:RV2Ga.72196$3Sm.6570@news01.bloor.is.net.cable.rogers.com...
> Jimmy Talbot wrote:
> > Hi,
> >
> > Peer-to-peer sharing networks (such as eDonkey) create a hash of their
> > shared files to make it easy to tell if two files are the same. I'm
> > working on software (not p2p sharing) where we need to do the same
thing.
> > After doing a bit of research (I'm a graphics programmer, not a crypto
guy
> > :) ), I'm thinking of using the Haval hashing algorithm. One of the
> > parameters to the hashing algorithm is how many 'rounds' (or passes) to
go
> > through.
>
> Drop HAVAL and use SHA-256 [or SHA-1] instead. Its faster and a
> standard :-) If you don't care about security why not just use MD4.
> Its faster and not likely to cause collisions.
>
> free source :
> http://libtomcrypt.org
>
> Tom
>

Relevant Pages

  • Re: Haval hashing - how many passes?
    ... MD4 does /have/ collisions, and some have been found, but I don't ... rather than maliciously chosen input. ... MD4 is way faster than MD5, and on large data might easily beat CRC32 too. ...
    (sci.crypt)
  • Re: Article: MD5 breaks on laptop computer in 15 minutes
    ... The experience with MD4 certainly doesn't bear out such predictions. ... collisions in MD4, then later we saw meaningful collisions. ... a MD5 collision that is meaningful when interpreted as a cert. ... an announcement of colliding Postscript documents: ...
    (sci.crypt)
  • Re: Haval hashing - how many passes?
    ... > have collisions., so if speed is an issue, try MD5. ... > to use the hash as an identifier, ... MD4 is still fine. ...
    (sci.crypt)