Re: Time Invariant String Comparison

[Philipp Janda <siffiejoe@...>]

Am 16.01.2014 14:01 schröbte Oliver Kroth:
> If the response time is measured, a random delay for the response should
> fix this.

A small random delay won't help much: You can test the same password 
multiple times, calculate the mean response time, and subtract the 
expected delay ...

> One may even add a basic delay for each response that is increased with
> every wrong attempt from same source,. This makes it more and more less
> efficient to hack the credentials.

This is a good idea anyways but not as easy to do if you have a 
multi-process server (you would need some form of shared memory to store 
state between requests).

> If the CPU load, memory access, HF radiation por some other side effects
> are monitored, a custom compare function may provide enough fog.

I think Thjis' suggestion could work, but I would make sure that all 
one-character strings are in the string pool at the time of the 
comparison or else you could measure the time difference between 
creating one or two one-character strings.

> Why not use hashed passwords, which is a better idea anyway as this
> takes a length independent time, and you may store the credentials in a
> safe way?

That would also be my favorite option. And always use salted password 
hashes!

Philipp