Implementing brute force attacks on hash values in Javascript

Question

I'm working for my bachelor thesis at the end of which I aim to implement a proof-of-concept Javascript-based hash cracker. The idea is to work like this: Users can submit a hash value along with information about the used algorithm. (Other) users can also click on a button on the website in order to participate in the cracking process. The server's task is to accept and split the submitted 'orders' into ranges, depending on the number of available workers. The ranges are then sent to the clients who clicked said button.

I am currently stuck with the big question of how to actually implement this brute force function. So my main problem now is that, frankly, I'm not really that settled in Javascript yet. For starters, I would just use a hardcoded character set: alpha-numeric, lower and upper case, no special characters. Problem is I honestly have absolutely NO clue of how to actually implement the function that would try out al the character combinations, on how to program that. I can imagine using a normal array containing the charset, then two strings. One string would contain the range, the other will contain the tried combinations. So I would somehow have to iterate through the charset array and the strings maybe with cascaded for-loops or something, but I'm really stuck with the question of 'how' exactly :). I don't expect any of you to actually provide me with the full source code for such a function (unless you want to of course), but I'd really appreciate some hints or explanations on how to implement such a brute force function. I'd also not bother about performance or optimized coding at this point, but rather about comprehensive coding, or whatever you might want to call it.

Answer 1

First define the space of potential passwords. Then number them sequentially.

For instance, suppose that you define the possible passwords as "sequences of eight letters and digits". This is an alphabet of 62 signs (26 lowercase letters, 26 uppercase letters, 10 digits). So there are 62⁸ = 218340105584896 passwords to try. Passwords are easily mapped to integers in the 0..218340105584895 range by considering the passwords to be a base-62 representation (i.e. 'a' is 0, 'b' is 1,... 'z' is 25, 'A' is 26,... 'Z' is 51, '0' is 52... and '9' is 61). Now that the potential passwords are just integers in a consecutive sequence, it is easy to distribute ranges.

Of course, performance of hash function implemented in Javascript will be very disappointing. A single PC with a good GPU can try more MD5 passwords than 1000 clients who use Javascript. You might want to make versions which use Java and/or .NET (Silverlight): these programming languages have much more muscle (not quite as good as C, but at least in the same magnitude order).

Edit: from your comment, I understand that you also wish to keep the target hash somewhat private, and not send it to the cracking machines. This forces the cracking machines to send back all their hashes (or part thereof) to the server, and you will have a bandwidth problem. Each cracking machine could send back, say, only the first two bytes of each computed hash; the server then knows which password candidates match the target hash on the first two bytes. This would reduce the problem by a factor 65536, and then the server can finish the work itself (retesting the matching candidates on its own CPU). However, with a 10 MBytes/s incoming bandwidth, you would be able to receive only about 5 millions hashes per second, and if you target a "simple" hash (like a single iteration of MD5), then this is very poor performance (the server alone, on its own CPU, can test many more passwords per second). This would be a valid scheme for a slow hash (like bcrypt) where each cracking machine will be able to compute a hundred hashes per second at most (but then, you really do not want to use Javascript...).

Answer 2

First of all it sounds like you need to formulate a cracking strategy, and to do that you should look at what tools hackers are using. Here is a nice blog post of someone who cracked 122 million hashes. He also links to a large list of real password hashes that you could use for your thesis. I think all of these hashes are unsalted, so that makes things easier.

RainbowCrack and JohnTheRipper are tools that have been around for a long time and are popular tools for breaking hashes. But what you are doing is pretty different, its more cloud-like in that its breaking up a large task to be completed by a large number of nodes. There are cloud cracking tools like Cryptohaze that look promising.