I'm building a web application and would like to use the strongest hashing algorithm possible for passwords. What are the differences, if any, between sha512, whirlpool, ripemd160 and tiger192,4? Which one would be considered cryptographically stronger?
+4
A:
bCrypt - Why would be a very long explanation, for which I recommend Enough With The Rainbow Tables: What You Need To Know About Secure Password Schemes
Basically, it's secure, it's slow, it's already implemented.
Tom Ritter
2008-10-08 18:20:57
A:
Mathematically, no idea. Though, I suppose, the general rule is the longer the final hash, the better -- assuming you can store the difference over a smaller hash.
Determining collisions is also helpful. And, with that, I'd say your choice of the algorithms you listed.
Jonathan Lonowski
2008-10-08 18:21:22
"the general rule is the longer the final hash, the better". That's very faulty reasoning. I could hash may password as "drowssap" with a zillion random characters after it -- that's a very poor hash. Strength is better determined by the computational complexity.
TrickyNixon
2008-10-08 18:37:58
A:
Here's a good post on coding horror about storing passwords. In short, he suggests bcrypt or SHA-2 with a random unique salt.
Daniel Plaisted
2008-10-08 18:26:38
+2
A:
David, those are all plenty strong functions. Even the much-ballyhooed MD5 collisions are not of the password-cracking variety, they just generate two different strings with the same MD5 (a very different proposition from finding a string that generates a given MD5 value).
If you are concerned about the security of the passwords, you need to worry about the protocols used to store them, the protocols used to recover passwords forgotten by users, and all the other possible avenues of attack. Those options are used far more often to crack passwords than brute-force crtyptanalysis.
Do use a salt, though.
But first read the article AviewAnew posted
SquareCog
2008-10-08 18:30:10
A:
If you are actually concerned about the security of your system (as opposed to the quite academic strength of algorithms) then you should go with a proven and mature implementation instead of nitpicking algorithms.
I would recommend Ulrich Drepper's SHA-crypt implementation. This implementation uses SHA-512, a 16 character long salt, is peer reviewed and scheduled to go into all major Linux distributions via glibc 2.7.
P.S.: Once you have reached this level of security, you'll be visited by the black helicopters anyways.
David Schmitt
2008-10-08 18:48:27