5 ms·

Just a wild guess, but I imagine you would start by designing the gif animation to depict a plausible MD5 hash, then modifying the file in such a way that you m

Just a wild guess, but I imagine you would start by designing the gif animation to depict a plausible MD5 hash, then modifying the file in such a way that you make the file MD5 hash match the one depicted in the animation. Very impressive!

sdegutis 10y ago
Are MD5 hashes that predictable that you can just make a few adjustments to a binary until it's the one you want?
- Filligree 10y ago
  MD5 as a cryptographic hash is thoroughly shredded, yes. It's still fine for non-cryptographic purposes, except that it's far slower than e.g. CRC.
  - hueving 10y ago
    That would be news if that were the case. The preimage resistance for MD5 was pretty strong last time I looked: http://crypto.stackexchange.com/questions/13303/is-md5-second-preimage-resistant-when-used-only-on-fixed-length-messages http://crypto.stackexchange.com/questions/13303/is-md5-secon...
MichaelGG 10y ago
But any 128 bit value is a plausible MD5 hash. Modifying the file to make that value is akin to doing a preimage attack right?
- rhaps0dy 10y ago
  Yes, that's probably the point of the maker. To prove that they can do a preimage attack.
  - kmm 10y ago
    No, it's just some fancy tricks with collisions. A preimage attack is still far beyond our current technologies
    - kevincox 10y ago
      It's important to note that technologies here means not only our hardware technologies but also our cryptographic knowledge. There is a chance that there are unknown flaws in the MD5 algorithm that make preimage attacks far easier to compute. Our compute power is also growing cheeper, which would let us solve harder preimage problems.
moyix 10y ago
No, this almost certainly isn't a preimage attack – it is more likely to be caused by the fact that you can generate MD5 collisions cheaply. Getting the gif animation to show the right frames based on the colliding block is a neat trick though! Looking forward to the writeup. Here's how a similar thing was done for PostScript: https://twitter.com/teh_gerg/status/838422647193157632 https://twitter.com/teh_gerg/status/838422647193157632

sdegutis 10y ago

Are MD5 hashes that predictable that you can just make a few adjustments to a binary until it's the one you want?

Filligree 10y ago
MD5 as a cryptographic hash is thoroughly shredded, yes. It's still fine for non-cryptographic purposes, except that it's far slower than e.g. CRC.
- hueving 10y ago
  That would be news if that were the case. The preimage resistance for MD5 was pretty strong last time I looked: http://crypto.stackexchange.com/questions/13303/is-md5-second-preimage-resistant-when-used-only-on-fixed-length-messages http://crypto.stackexchange.com/questions/13303/is-md5-secon...

MichaelGG 10y ago

But any 128 bit value is a plausible MD5 hash. Modifying the file to make that value is akin to doing a preimage attack right?

rhaps0dy 10y ago
Yes, that's probably the point of the maker. To prove that they can do a preimage attack.
- kmm 10y ago
  No, it's just some fancy tricks with collisions. A preimage attack is still far beyond our current technologies
  - kevincox 10y ago
    It's important to note that technologies here means not only our hardware technologies but also our cryptographic knowledge. There is a chance that there are unknown flaws in the MD5 algorithm that make preimage attacks far easier to compute. Our compute power is also growing cheeper, which would let us solve harder preimage problems.

moyix 10y ago

No, this almost certainly isn't a preimage attack – it is more likely to be caused by the fact that you can generate MD5 collisions cheaply. Getting the gif animation to show the right frames based on the colliding block is a neat trick though! Looking forward to the writeup. Here's how a similar thing was done for PostScript: https://twitter.com/teh_gerg/status/838422647193157632 https://twitter.com/teh_gerg/status/838422647193157632