bcrypt

bcrypt copied to clipboard

It's generally not recommended to use bcrypt on top of another hashing algorithm. While doing it over sha256 is obviously not as bad as doing it over md5, I think it's important to note in the README that OWASP recommends against the practice.

Sure it is. “Hash shucking” remains an issue, and the differences in collision resistance and output size aren’t relevant to this application.

You're absolutely right!

How about recommending pre-hashing with HMAC as a keyed hash instead? Or just bcrypt.kdf?

I've removed the mention of the DDoS attack, replaced sha256 with bcrypt_pbkdf, and adjusted the explanation of why using an unkeyed hash function is not a good idea. What do you think?

I don't have a lot of experience with bcrypt_pbkdf, so I just copied the parameters from the kdf part of the README. Do you think those parameters are readable?

https://flak.tedunangst.com/post/new-openssh-key-format-and-bcrypt-pbkdf has some information about choosing parameters for bcrypt_pbkdf. I've fixed CircleCI so if you push a new commit here CI should pass now and we can get it reviewed 😄

closing/reopening to trigger CI