File validation and hash collisions?

C

cyclone3d

[H]F Junkie
Joined
Aug 16, 2004
Messages
16,305
This topic has come to mind every once in a while and it always makes me ponder what the huge deal is with worrying about hashing algorithms such as md5 for executables.

Even if you were able to reliably change the content of a file to return the same hash as the original file, other than making the program crash, how extremely hard / impossible would it be to make a modified executable return the same exact hash and also still run? Then think about how impossible it would be to do that AND be able to have malicious code run from that same executable.

Anybody have a single example of a hash collisioned executable being able to even run?
 
cyclone3d said:
This topic has come to mind every once in a while and it always makes me ponder what the huge deal is with worrying about hashing algorithms such as md5 for executables.

Even if you were able to reliably change the content of a file to return the same hash as the original file, other than making the program crash, how extremely hard / impossible would it be to make a modified executable return the same exact hash and also still run? Then think about how impossible it would be to do that AND be able to have malicious code run from that same executable.

Anybody have a single example of a hash collisioned executable being able to even run?
Click to expand...
I don't have an example, but I know it has been done. It's especially "easy" with low resolution hashes, because a rounding error is all you need to have the same hash. Theoretically you could change a large chunk of code and do an asm jump to it from the beginning of the program if conditions are right, and still maintain the same hash.
 
Nobu said:
I don't have an example, but I know it has been done. It's especially "easy" with low resolution hashes, because a rounding error is all you need to have the same hash. Theoretically you could change a large chunk of code and do an asm jump to it from the beginning of the program if conditions are right, and still maintain the same hash.
Click to expand...
You could also hijack an existing jump and send it to some other place that has malicious code, thinking about it. Would be easier and even less detectable.
 
If you were to go to the trouble of writing a compiler specifically to target a specific hash, it could be done rather trivially. MD5 is just a mathematical equation.

However, it would be far easier to pad the executable with junk at the expense of increases file size. You'd just need to write the tool to target a specific hash.

Hashing algorithms such as MD5 are not meant to solve security issues. But you can leverage them for their mathematical improbability of detecting bit flips and truncated files. These are general solutions to these simple problems.

Exploitations such as executable manipulations require more complex solutions such as digital signatures and code signing. Those are much more difficult to beat.
 
You must log in or register to reply here.
Back
Top