4 ms·
I can't believe it's 2008 and we're still reading new stories about MD5 collision-resistance. We get it, anybody can manufacture N files with the same hash and
by aston 19y ago
I can't believe it's 2008 and we're still reading new stories about MD5 collision-resistance. We get it, anybody can manufacture N files with the same hash and different content. Don't use MD5 as a piece of your vote commitment scheme and move on. All of this was news years ago.
edit: MD5 is still good for random other stuff. If you MD5 something, it's not that easy for someone to find a collision with it. I'm anxiously awaiting the revelatory article on rainbow tables in The Economist.
- tptacek 19y agoWhat is MD5 good for? It's slower than a CRC and less secure than SHA1. It's a crappy universal hash (bad constant factors) and it's usually much too large for a unique identifier.
- aston 19y agoMy intent isn't to bat for MD5, really. But since you asked me, I'll point out that MD5 is actually not that bad if you want a deterministic string of randomly distributed bits. If you don't need all 128 bits, nobody's forcing you to keep the extra ones. That said, you might as well use SHA1.
- tptacek 19y agoFor what it's worth, almost any random number generator is better than MD5 at giving a deterministic random bit string, given a seed. You might start with OpenBSD arc4random.
- mattmaroon 19y agoYou have to give them credit for explaining it well to non-hackers.
- mechanical_fish 19y agoAn article which explains that a lock is broken, but doesn't tell you what the lock is protecting, how to get it changed, and what alternative locks are available, is not explaining the situation well.