Mal-D2gan: Double-Detector Based Gan for Malware Generation
Nam Hoang Thanh, Trung Pham Duy, Lam Bui Thu · 2024
In recent years, machine learning (ML) has been developed to effectively detect malware. Most researchers have focused on improving detection performance, but they have often overlooked the robustness of ML models. In addition, many machine learning algorithms are highly susceptible to intentional attacks. To address these issues, we propose generating adversarial malware samples using GANs to enhance malware detectors' robustness. However, since current GAN models suffer from limitations such as unstable training and weak adversarial samples, we propose the Mal-D2GAN model to address these problems. Specifically, the Mal-D2GAN architecture was designed with a double-detector and a least square loss function and tested on a dataset of 20,000 samples. The results show that the Mal-D2GAN model reduced the detection accuracy (true positive rate) in 8 malware detectors. The performance was then compared with the existing MalGAN and Mal-LSGAN models.