fast hashes are killing our security.

But let’s get a bit of context here – do we really need to be able to generate 4.7 billion password hashes per second? That’s like taking the entire Facebook population of over 900 million and sequentially hashing every single password 5 times every second. And that’s on consumer hardware. No, we don’t need to get anywhere even close to that.

The problem is that algorithms like MD5 and SHA were designed to demonstrate data integrity at high computational speed rather than to provide a password storage mechanism; cryptographic hash functions are not password hash functions. Even if they were “safe” for password storage when designed, MD5 goes back 20 years now so by Moore’s Law we now have processors that are now eight thousand times faster.

There are various approaches for breathing life back into old algorithms; key stretching, for example, where an algorithm which is too fast is “slowed” by repeating it over and over again, perhaps thousands of times. But the guidance around the likes of MD5 and SHA is clear and OWASP summarises it quite succinctly:

General hashing algorithms (eg, MD5, SHA-1/256/512) are not recommended for password storage. Instead an algorithm specifically designed for the purpose should be used.

What we need is a hashing algorithm which was designed from the ground up with speed in mind, not fast speed, but slow speed.
The concept of increasing the effort required to execute the hash function is one that is frequently implemented by key stretching and indeed this is modus operandi of PBKDF2. In fact PBKDF2 may then be applied to an algorithm such as SHA so strictly speaking, SHA is still being used, just not as we know it in its single iteration form. //

the thing about algorithms like these is that they’re adaptive:

Over time it can be made slower and slower so it remains resistant to specific brute-force search attacks against the hash and the salt.

Clearly the ability to increase the workload is important if we don’t want to be caught by Moore’s law again in the near future. All we need is a way to integrate this into our existing work.