Construct MD5 Collisions Using Just A Single Block Of Message

So far, all the differential attacks on MD5 were constructed through multi-block collision method. Can collisions for MD5 be found using just a single block of message (i.e. 512-bit)? This has been an open problem since the first 2-block collision attack was given. However, a paper titled “How To Find Weak Input Differences For MD5 Collision Attacks” (Cryptology ePrint Archive (2009/223), http://eprint.iacr.org/) listed all the practically possible weak differences that can be used to make attacks on MD5, in the Table1 of that paper, only an 1-block message difference ( 10 , 5 m , 31 , 10 m ) was included in and suggested in the end of the paper to be able to be exploited to construct a practical collision attack on MD5. A hint was later given in EUROCRYPT2009’s poster paper tilted “Could The 1-MSB Input Difference Be The Fastest Collision Attack For MD5? ” (LNCS 5479, the poster session of EUROCRYPT 2009. Cryptology ePrint Archive (2008/391), http://eprint.iacr.org/) that, 1-block collision attack on MD5 is possible if a more efficient searching algorithm can be developed using evolutionary approaches. Today, in the last month (Dec,) of 2010, we have to make public a result of our 1-block collision attacks on MD5 in Table 1 as below, which was actually obtained at the beginning of 2010, but for security reasons, the techniques are not allowed to be disclosed at the moment.