A Novel Network Intrusion Detection System Based on CNN
Lin Chen, Xiaoyun Kuang, Aidong Xu, Siliang Suo, Yiwei Yang
Abstract
Network intrusion detection system (NIDS) plays an important role in network security. It can detect the malicious traffic and prevent the network intrusion. Traditional methods used machine learning techniques such as support vector machine, Bayesian classification, decision tree and k-means. The traditional machine learning methods first need to manually select features and has obvious limitations. In this paper, we propose a novel NIDS system based on convolutional neural network. We train deep-learning based detection models using both extracted features and original network traffic. We conduct comprehensive experiments using well-known benchmark datasets. The results verify the effectiveness of our system and also demonstrate the model trained through raw traffic has better accuracy than the model trained using extracted features.