Intrusion detection based on image transformations and data augmentation

Abstract

The increasing growth of users and communication networks in different platforms has led to the emergence of various types of network attacks. intrusion detection systems (IDS) are one of the important solutions to cope with these problems. An IDS determines whether incoming traffic is intrusive or normal. IDSs often achieve high efficiency with methods based on deep neural networks. However, one of the shortcomings of these methods is the lack of sufficient attention to the spatial features in the data. This research presents an intrusion detection method based on image transformations and data augmentation is presented. In the proposed method, the intrusion detection process is performed by transforming the traffic vector into an image using a convolutional neural network (CNN). Also, we use data augmentation and dimension reduction techniques to increase accuracy and reduce complexity in the proposed method. Simulation results on network security laboratory - knowledge discovery and data mining (NSL-KDD) show that the proposed IDS can classify intrusion traffic with an accuracy of 97.58%.