Cheat sheet update/refactor proposal: Issue on the PBKDF2 iteration counter specified in the Password Storage Cheat Sheet

[Caerostris]

The password storage cheat sheet recommends an iteration count of 10.000 for PBKDF2 key derivation, referencing a 2012 Apple report using these numbers. I believe it should be clear that a 2012 recommendation is out of place in a current security cheat sheet.

The 2018 version of the report states that 10 million iterations are now used for iOS backups. Other reports (although not quite up to date either) mention 100.000 and 200.000 iterations.

Considering how quickly these recommendations seem to be changing at the moment, does it even make sense to include a fixed number at all?

[righettod]

Hi,

Thanks you very much for the feedback.

I agree on this. In the CS we recommand to use Argon2 as first choice (in 2019 I think it's now possible).

I think that is still necessary to add a hint about a minimal number of iteration in order to allow the CS reader to have directly usable information. But, on our side, it need to maintains this hint up to date.

I think to add a hint like this:

return [salt] + pbkdf2([salt], [credential], c=[iteration_count]);

As computation time depend on the target system,
"iteration_count" must have a number implying that the computation time 
on the target system must take at least 1 second. Like 1000.000 for example.

What do you think?

[Caerostris]

This is exactly the kind of wording I was hoping for!
Giving a target computation time rather than a fixed iteration count is far less likely to become outdated in the future.

[righettod]

OK, thanks you very much , I update the CS in this way 😃

[Caerostris]

Perfect, thanks!

[righettod]

I have updated the CS, ping me if it is not OK...