ackdav / fypj

Software Vulnerability Analysis and Discovery using Deep Neural Network

FYPJ

Software Vulnerability Analysis and Discovery using Deep Neural Network

Information

• Augmented afl-fuzz - contains the afl-fuzz incorporated with the neural network and additional original sources
• Dataset collection - Dataset gathering scripts, compressed .npz datasets used, seed files used to fuzz
• Jupyter Notebooks - training of neural network
• Models - models trained in .h5 format

Dependencies

• Keras
• Tensorflow
• numpy

Command used to fuzz and run afl-fuzz

afl-fuzz -i <seed file input dir> -o <output dir> <readelf program> -a @@

Standard commands to run afl-fuzz with addition of -a parameter used for the readelf which displays all information. Note: @@ is a placeholder and has to be included when running the augmented afl-fuzz segfault occurs.