Malware Detection Using Machine Learning
Predicting whether an executable is malware or benign.
Malware remains a large problem, as attackers use it to disrupt systems with costly after-effects. Detection is mainly carried out using heuristic and signature-based methods, which fail to keep up with the continuous evolution of malware families.
This project explores detecting malware by extracting features from binaries and using them to train a deep neural network. We experimented with AutoEncoders, LSTMs, and CNNs, and also with raw byte sequences as features. The results show that the network can learn and extract meaningful information even from the raw bytes.
Built at the SGCSC Cybersecurity hackathon during SG-CRC 2017 at the National University of Singapore, where it won 3rd prize.