MalGAN – creating evasive malware

Using Generative Adversarial Networks (GANs), we can create adversarial malware samples to train and improve our detection methodology, as well as to identify gaps before an adversary does. The code here is based on j40903272/MalConv-keras. The adversarial malware samples are malware samples that have been modified by padding them with a small, but carefully calculated, sequence of bytes, selected so as to fool the neural network (in this case, MalConv) being used to classify the samples.

Getting ready

Preparation for this recipe involves installing the pandas, keras, tensorflow, and scikit-learn packages in pip. The command is as follows:

pip install pandas keras tensorflow sklearn...