I am extracting strings from malware samples and I would like to build an algorithm to tell from which family of malware a string is (most probably) coming.
I thought about using a string for the network's input either using a feature vector (hashing?) and the output being the probability of belonging to one of the predefined families.
Which kind of model should I use? How can I send strings as the network's input?
I found that a "sequence to sequence" model should fit my requirements or a LSTM model, but to be honest I am a bit lost.