14 ms·

> (even if limited to 26 letters and 8 characters that's over 200 billion different passwords) Of course, if you're using a fast cryptographic hash function th

> (even if limited to 26 letters and 8 characters that's over 200 billion different passwords) Of course, if you're using a fast cryptographic hash function then someone with a good GPU can throw a surprising amount of brute force at the problem. A Tesla M60, which is five years old, can do about 1.4 billion SHA-256 hashes per second; Amazon will rent you a server with one of these for $0.75/hour, and cracking a single 8-character lowercase alphabetic password should take about 2.5 minutes at most. Half that on average. (Functions like bcrypt help with this by being designed for slowness, and newer ones like scrypt also try to use large amounts of memory with unpredictable access patterns so they're harder to accelerate with GPU/FPGA/ASIC hardware.)

Tainnor 6y ago
This is all true, and this why we have bcrypt etc., but it was just an example for illustration. In real-life situations, the input space is much larger still (if you account for longer passwords, case sensitivity, special characters etc.), as long as your passwords are truly unguessable. And at some point, even a very fast hashing algorithm won't be able to keep up. I don't know the exact calculations, but I would expect MD5 to still be hard to break for truly strong passwords (say, 20 random characters). The problem is more that people don't actually choose strong passwords most of the time.