4 ms·

I wouldn't say its easy, but the obvious known example is the forged microsoft certificates on the Stuxnet/Flame/Gauss viruses (widely believed to be developed

• lawnchair_larry 14y ago
  That example does not apply here. A collision attack doesn't help. You need a second preimage attack, of which a practical one doesn't exist for MD5. The difference is with the collision attack, the attacker controls the inputs and has to find any two valid messages with the same hash - any hash. That's what was done with forged certificates. In this case, the hash you need to match against is fixed. You have one preimage, which is also fixed of course. You have to find a second preimage that also matches that specific hash. If you can do this at all, let alone "easily", you will significantly advance the field of cryptography. (To raise the bar even higher, your second preimage has to be valid javascript that executes your malicious action.)
  • kelnos 14y ago
    Fortunately, even the output of /dev/random is usually valid javascript, so that part isn't much of a concern ;)
  • drostie 14y ago
    That last point isn't particularly deep; the same techniques used in http://www.win.tue.nl/hashclash/rogue-ca/ http://www.win.tue.nl/hashclash/rogue-ca/ can be used for two files which are UTF-16 encoded and contains sections which are commented out. With this said, that is a collision attack and not a preimage attack and so your first point is crucial.