Anthony Ferrara has a new post today looking at password hashing and a type of hashing that's beginning to get more attention in the PHP community - scrypt. However, he doesn't recommend it for production password storage and shares his reasoning why.
Scrypt was not designed for password storage. It was designed as a key derivation function for generating keys from weak material (namely passwords). The prime type of attack that scrypt is designed to defeat is ASIC based attackers. It is not designed to try to favor CPU over GPU (and thereby defeat GPU based attacks). It is this fact that we can leverage to gain an advantage when used as a password hashing mechanism.
He covers some of the basic design decisions that were made when scrypt was created. He also points out that none of the results of these decisions are strictly fatal, they just make it a bit weaker than something like bcrypt for password storage. He goes through the basic inputs scrypt requires and includes a quick snippet of code (not PHP, but easy to understand) showing its use. He talks about its "chain of 4 operations" and gets into what he sees as limitations: loop unrolling and the tune-able reduced memory usages. He finishes off the post mentioning that scrypt is still secure, but despite this he doesn't recommend it for password storage specifically.
Link: http://blog.ircmaxell.com/2014/03/why-i-dont-recommend-scrypt.html
没有评论:
发表评论