(CLOUD 2020) A Network Intrusion Detection Approach Based on Asymmetric Convolutional Autoencoder

Abstract: With the continuous development of Internet technology, cy- berspace security protection technology is still facing many new intrusion threats. Aiming at the complex and changeable network abnormal traffic and intrusion behavior, we apply the deep learning approach to the field of intrusion detection. In this paper, we detail our proposed asymmetric convolutional autoencoder (ACAE) for feature learning. In addition, we propose a network intrusion detection model based on the combination of asymmetric convolutional autoencoder and random forest. This ap- proach can well combine the advantages of deep learning and shallow learning. Our proposed approach is evaluated on KDD99 and NSL-KDD dataset, and compared with other approaches in the field of intrusion detection. Our model can effectively improve the classification accuracy of network abnormal traffic and it has strong robustness and scalability in current NIDSs.

Authors: Shujian Ji (University of Chinese Academy of Sciences & SIAT, China); Chengzhong Xu (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, USA); Kejiang Ye (Chinese Academy of Sciences, China)

Email: cz.xu@siat.ac.cn, yekejiang@gmail.com