Fighting malware and spam

Security of hash functions called into question

The encryption field was thrown into a frenzy at the end of last month when the security of hash functions MD5, SHA-0 and SHA-1 was called into question.

The encryption field was thrown into a frenzy at the end of last month when the security of hash functions MD5, SHA-0 and SHA-1 was called into question. First, a collision in SHA-0 was uncovered by Antoine Joux; then a group of Chinese researchers released a paper which outlined methods of finding collisions in the MD4, MD5, HAVEL-128 and RIPEMD algorithms; finally, researcher Eli Biham of the Israel Institute of Technology reported at the Crypto 2004 conference preliminary research findings that indicate the presence of vulnerabilities in SHA-1.

In principle it is not possible to design a hashing algorithm that prevents the production of duplicate fingerprints (hash collisions), but the hashing algorithms are designed to make it very difficult to generate duplicate hash codes. It seems that, for MD5 at least, it is easier to do so than originally hoped. While there currently does not seem to be an easy way of faking an arbitrary hash code - thus limiting the usefulness of an attack - it does call into question the usefulness of these hashes as digital signatures.

A similar situation is true of SHA-0, but the evidence that the more widely used SHA-1 is likewise broken is not currently conclusive. However, the possibility that SHA-1 may be flawed is a cause for concern, since SHA-1 has become a legal standard for document signing - it is currently embedded in PGP and SSL and is the only signing algorithm approved for use in the US Government’s Digital Signature Standard.

[Next month’s issue of VB will contain a more detailed look at the security flaws in these hashing algorithms and the implications for the anti-virus industry.]

23 August 2004

Tags:    del.icio.us   digg this

Poll

VB100 certification