[Cryptography] let's kill md5sum!

[Sampo Syreeni]

On 2015-06-08, Peter Gutmann wrote:

> There are a lot of cases where it's used as a kind of super-CRC32, to 
> detect data corruption on storage media rather than malicious 
> alterations, and it's perfectly adequate for that.

Except that it's not. Of course you do have a stochastic guarantee, with 
md5sum, but no hard one like you do with CRC32.

That *can* fuck you up, when you pass lots of short messages, like you 
do over a constrained radio channel. And it *will* fuck you up with 
current bandwidths, sooner or later. Without you knowing what then 
happens, because you haven't quantified the error-floor to coding 
bandwidth staircase.

Likely the problem isn't grave. But I don't much like the idea of 
mistaking ECC for cryptographic checksums. Especially in any application 
where the coding space is less than the square of log of all the bits 
likely to be produced in the lifetime of the protocol.
-- 
Sampo Syreeni, aka decoy - decoy at iki.fi, http://decoy.iki.fi/front
+358-40-3255353, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2