Evaluation of Adversarial Training on Different Types of Neural Networks in Deep Learning-based IDSs
Rana Abou Khamis, Ashraf Matrawy
Abstract
Network security applications, including Intrusion Detection Systems (IDS) of deep neural networks (DNN), are increasing rapidly to make detection task of anomaly activities more accurate and robust. With the rapid increase of using DNN and the volume of data traveling through systems, different growing types of adversarial attacks to defeat DNN create a severe challenge. In this paper, we focus on investigating the effectiveness of different evasion attacks and how to train a resilience deep learning-based IDS using different Neural networks, e.g., Artificial Neural Network (ANN), Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). We use the min-max formulation to formulate the problem of training robust intrusion detection systems against adversarial samples using two benchmark datasets. Our experiments on different deep learning algorithms and different benchmark datasets demonstrate that defense using adversarial training based min-max formulation increases the robustness of the network under the assumption of our threat model and five state-of-the-art adversarial attacks.