"what is the input and output of this at a conceptual level?" Our input is the dataset we want to crack, and our output is the passwords that were successfully cracked. In a more conventional scenario, we would have a list of hashses that need to be cracked. So the steps are
1. pick parents at random
2. crossover and maybe mutate
3. hash the child
4. see if the hashed child exists as a password in our list of hashses we want to crack
5. if it does, add the child to the end of the container and pop the oldest organism from the front.
6. goto start
it will genetically run through all ngrams and check if it is in the pass list to determine how to advance the evolution of the algorithm
does this crack passwords genetically? well, yes and sorta
it's a proof of concept against an existing list of real leaked passwords, proving that it could efficiently crack a number of these real passwords real people were using to protect real personal data
but from there you have to extrapolate the effectiveness on all possible passwords..
if you train against the myspace list then the passwords would have to resemble myspace passwords
can you train the algo on the myspace list then try to crack nuclear codes? very unlikely, unless government officials are protecting their access with passes like 'WARMACHINEROX'
I chuckled, imagining how some government official changes his password right now.
Like substrings "pass" and "word" in "password".
One of the examples apparently takes a list of leaked passwords from MySpace - and so what does it "crack" then? I think the phrase "high impact substrings" down in the explanation is the key, but it's not wholly clear to me what the ultimate purpose of this is.
The 'GA without a fitness score' idea seems interesting, but it would help to know what exactly the algorithm is trying to do.