Artificial Neural Networks or Neural Networks (NN) is a machine
learning technique composed of nodes called Artificial Neurons, just
like the brain possesses. Such systems use Machine Learning to
approximate highly dimensional functions and progressively learn
through examples of training set data, or in our case a large password
dump. They have shown initial promise to be effective at generating
original yet representative password candidates. Advantages to NN's
for password cracking are the low overhead for storing the final NN
model, approximately 500kb, and the ability to continually learn over
time through retraining or transfer learning.
Resources
Fast, Lean, and Accurate: Modeling Password Guessability Using Neural
Networks (USENIX '16)
https://www.usenix.org/system/files/conference/usenixsecurity16/sec16_paper_melicher.pdf
https://github.com/cupslab/neural_network_cracking
[[Home]]
#advanced
#concepts