I need to know the cost of succeeding with a Preimage attack ("In cryptography, a preimage attack on a cryptographic hash is an attempt to find a message that has a specific hash value.", Wikipedia).
The message I want to hash consists of six digits (the date of birth), then four random digits. This is a social security number.
Is there also a possibility to hash something using a specific password. This would introduce another layer of security as one would have to know the password in order to produce the same hash values for a message.
I am thinking about using SHA-2.
If you want to know how expensive it is to find a preimage for the string you're describing, you need to figure out how many possible strings there are. Since the first 6 digits are a date of birth, their value is even more restricted than the naive assumption of 10^6 - we have an upper bound of 366*100 (every day of the year, plus the two digit year).
The remaining 4 'random' digits permit another 10^4 possibilities, giving a total number of distinct hashes of 366 * 100 * 10^4 = 366,000,000 hashes.
With that few possibilities, it ought to be possible to find a preimage in a fraction of a second on a modern computer - or, for that matter, to build a lookup table for every possible hash.
Using a salt, as Tom suggests, will make a lookup table impractical, but with such a restricted range of valid values, a brute force attack is still eminently practical, so it alone is not sufficient to make the attack impractical.
One way to make things more expensive is to use iterative hashing - that is, hash the hash, and hash that, repeatedly. You have to do a lot less hashing than your attacker does, so increases in cost affect them more than they do you. This is still likely to be only a stopgap given the small search space, however.
As far as "using a password" goes, it sounds like you're looking for an HMAC - a construction that uses a hash, but can only be verified if you have the key. If you can keep the key secret - no easy task if you're assuming the hashes can only be obtained if your system is compromised in the first place - this is a practical system.
Edit: Okay, so 'fractions of a second' may have been a slight exaggeration, at least with my trivial Python test. It's still perfectly tractable to bruteforce on a single computer in a short timeframe, however.
SHA-2, salts, preimage atttacks, brute forcing a restricted, 6-digit number - man it would be awesome if we have a dial we could turn that would let us adjust the security. Something like this:
Time to compute a hash of an input:
SHA-2, salted Better security!
| |
\|/ \|/
|-----------------------------------------------------|
.01 seconds 3 seconds
If we could do this, your application, when verifying that the user entered data matches what you have hashed, would in fact be a few seconds slower.
But imagine being the attacker!
Awesome, he's hashing stuff using a salt, but there's only 366,000,000 possible hashes, I'm gonna blaze through this at 10,000 a second and finish in ~10 hours!
Wait, what's going on! I can only do 1 every 2.5 seconds?! This is going to take me 29 years!!
That would be awesome, wouldn't it?
Sure would.
I present unto you: scrypt and bcrypt. They give you that dial. Want to spend a whole minute hashing a password? They can do that. (Just make sure you remember the salt!)
I'm unsure what your question is exactly, but to make your encrypted value more secure, use salt values.
Edit: I think you are sort of describing salt values in your question.
