Network Intrusion Detection System using Autoencoders

Network Intrusion Detection Systems (NIDS) play a critical role in identifying malicious activity within network traffic. However, developing a flexible and efficient NIDS capable of handling unforeseen attacks presents several challenges. In this paper, we explore the application of autoencoders for anomaly-based NIDS, focusing on various types of regularization techniques to enhance detection performance. We compare the performance of three types of autoencoders—vanilla, sparse and denoising—using the benchmark NSL_KDD dataset by evaluating different performance metrics such as accuracy, precision, recall and f1-score. We find that the vanilla and sparse autoencoders are suitable for the dataset.