Using GANs to Improve the Accuracy of Machine Learning Models for Malware Detection

The increase of cyber-attacks and new malware in the last decade led to the usage of various machine learning techniques in security products. While these techniques are designed to improve accuracy, some practical constraints (such as lowering the false positive rate) often influence the selected model. This paper focuses on how various generative adversarial networks can be used to improve the average detection rate and reduce the false positives for a given neural network, by altering the training set. The result of this paper is a technique that can be used to reduce the number of false positives while preserving or in some cases increasing the detection rate.