Re:Django-bcrypt
by ylem on Thursday February 03, 2011 @06:52PM
(#2177582)
What were your timings like for bcrypt using 12 rounds? Also, while for the SHA-x algorithms, there have been numerous tests--what about for the pycrypt module?
[ Parent
| on HN ]
Timings are going to be completely dependent on the server. Anything I tell you will be wrong unless you're running on my server.
And I didn't test any SHA-x algorithms. Any algorithm that has a time you can't easily increase by tweaking a number (BCRYPT_ROUNDS) will eventually become insecure as computers get faster.
Re:Django-bcrypt
by JoachimSchipper on Friday February 04, 2011 @06:08AM
(#2179058)
It's not straight SHA-x, I'd wager; glibc includes a version of phk's MD5-based crypt() based on SHA2 (with some bonus insanity thrown in; it's a spawn of MD5 crypt and Drepper, after all.)
Like md5crypt, this new crypt() is not based on established cryptographic principles.