Adversarial Examples Against the Deep Learning Based Network Intrusion Detection Systems
Kaichen Yang, Jianqing Liu, Chi Zhang, Yuguang Michael Fang · 2018
Deep learning begins to be widely applied in security applications, but the vulnerability of deep learning in front of adversarial examples raises people's concern. In this paper, we study the practicality of adversarial example in the domain of network intrusion detection systems (NIDS). Specifically, we investigate how adversarial examples affect the performance of deep neural network (DNN) trained to detect abnormal behaviors in the black-box model. We demonstrate that adversary can generate effective adversarial examples against DNN classifier trained for NIDS even when the internal information of the target model is isolated from the adversary. In our experiment we first train a DNN model for NIDS system using NSL-KDD database and achieve a performance matching the state-of-art literature, then we show how can an adversary generate adversary examples to mislead the model without knowing the internal information.