A Wasserstein GAN-based Framework for Adversarial Attacks Against Intrusion Detection Systems

Intrusion detection system (IDS) has become an essential component of modern communication networks. The major responsibility of an IDS is to monitor communication networks for malicious attacks or policy violations. Over the past years, machine learning (ML) and deep learning (DL) have been employed to construct effective IDS. However, recent studies have shown that the reliability of ML/DL-based IDS is questionable under adversarial attacks. In this paper, we propose a framework based on Wasserstein generative adversarial networks (WGANs) to generate adversarial traffic to evade ML/DL-based IDS. Compared with the existing adversarial attack generation schemes, the proposed framework only involves highly restricted modification operations and the output of the framework is carefully regulated, ultimately preserving the type of the intended malicious traffic. In our research, we validated the effectiveness of the proposed framework by launching adversarial attacks of varied types against multiple ML/DL-based IDS. Our experimental results in terms of detection rate and evasion increase rate indicate that the proposed framework can completely deceive the IDS based on Naive Bayes (NB), Logistic Regression (LR), Random Forest (RF), and Recurrent Neural Network (RNN). In addition, the framework can partially evade the IDS based on Decision Tree (DT), Gradient Boosting (GB), and Multilayer Perceptrons (MLP).