# What value for a Rfc2898 (PKBDF2) is overkill?

I am working on a project which needs to share messages, and part of that project is deriving a password, I am currently using a value of 65537, totally arbitrary (it is $$2^{16} + 1$$, so not TOTALLY) and that takes about 1-second to compute. but, I just wondered what value or iteration count is enough?

Note its just a hobby project, not a federal info. transfer thing.

• What is your risk? who is your enemy? – kelalaka Oct 13 at 17:57
• @kelalake no one. its just for fun. but in general, like a big company or something, what would be a good value? – Omer Enes Oct 13 at 18:14
• NIST SP 800-63B recommends at least 10,000 but as long as as you can afford (so measure to hit 100ms - 1s if you can or as long as your users won't notice the latency too much) – SEJPM Oct 13 at 18:30
• Can you pick a passphrase for each user, by a process with >128 bits of entropy, e.g. a string of ten words chosen independently and uniformly at random from a 7776-word list? If so, it doesn't really matter how many iterations of PBKDF2 you use, as long as you use a random salt per user. (If you use a process with 256 bits of entropy, like twenty words, then there's no need for a salt either.) – Squeamish Ossifrage Oct 14 at 0:45