> I doubt i could do better than just compressing the dataset and extracting it on the fly.
The meta in that field is to extract "rules" (i.e. for hashcat) from datasets. Then you run the rules over the encrypted dumps. Rules can be word^number^number, word^number^word^number, or letter^upper^number^lower... etc. Then you build a dictionary and dict + rules = passwords.
Pretty sure you can extract some nice signals nowadays with embeddings and what not.
The meta in that field is to extract "rules" (i.e. for hashcat) from datasets. Then you run the rules over the encrypted dumps. Rules can be word^number^number, word^number^word^number, or letter^upper^number^lower... etc. Then you build a dictionary and dict + rules = passwords.
Pretty sure you can extract some nice signals nowadays with embeddings and what not.