Consider the figuring-out to be over.

MD5 (or Message Digest 5) is a 128-bit cryptographic hash function developed by Ron Rivest (the 'R' in RSA) which produces a 128-bit summary (or digest) of a file.  It was intended and always assumed that locating two source files which produced the same 128-bit hash was cryptographically 'difficult' to achieve, thus allowing the hash to act as some kind of proof that the file was unaltered.

Extending the Chinese research, it became obvious that all that was needed was to isolate a small portion of the second file and 'fiddle' with that section until a hash collision was achieved.  Obviously, that would require fiddling with either unimportant or non-obvious portions of the file – suggestions include random padding or (much smarter) graphics images where seemingly invisible changes can be made without altering the appearance of the picture.

All well and good.  But how might that be used?

Enter Alexander Sotirov and his fellow presenters at last night’s CCC Conference presentation.

They have found an easy way to duplicate an intermediate Certificate Authority’s (CA) certificate and masquerade as a legitimate (but actually bogus) Intermediate Certificate Authority which will be trusted by ALL major browsers.  This is achieved by a rapid computation of the bogus information to match the MD5 hash of the valid data.  Note that (at the moment) rapid = roughly a day of computation on a cluster of 200 PS3s.

In case you’re wondering, that have actually done it; they have a fully-configured, seemingly valid CA which is entirely fake.

Read on for why the internet isn’t quite as broken as this all might suggest, and further on for why it really is seriously broken.