Although MD5 was theoretically cracked in 2004 by Chinese professor Wang Xiaoyun its practical attack scenario was not shown to the world which was mainly in my opinion due to the complexity of the method and ongoing research to minimize processing and computational power required and therefore detecting collision was not thought a practical approach. MD5 has already been abandoned by the more security savvy organization as their preferred mean of calculating hashes and digitally signing them.
In 2008 researchers from different parts of world gathered again knowing that enough research has been done to practically present the threat to those still using this weak algorithm and worked on finding collision - different messages having same MD5 hash - on MD5 signed SSL certificates and finally succeeded by the end of the year 2008 to create a fake CA certificate issuer and prove the practical implementation of the attack. The irony of the situation is that despite of the fact that MD5 has been proven to contain weakness in its hashing mechanisms many renowned Root CA's still use it.
The researchers disclosed their work at the 25C3 conference in Berlin on the 30th of December by creating a fake ssl certificate signed by RapidSSL which the researchers thought was the weakest of all. VeriSign, the issuers of RapidSSL certificates stopped using MD5
More details can be found on the links mentioned below. 25C3 has also released the videos of the presentation on their website.
- Detailed explanation
- Slides from the 25c3 presentation
- Demo site (set your system date to August 2004 before clicking)
Although the private key of this was not released due to the danger of being misused by phishers the method itself was elaborated to a point that this could be done in a lesser amount of time.