Online Network Intrusion Detection Based on Self-Encoder Integration

With the development of information technology, network security problems have intensified, and the performance enhancement of network intrusion detection systems has attracted much attention In this paper, we propose an online system based on self-encoder integration that utilizes deep learning to improve detection accuracy and robustness Neural networks are popular for their ability to learn complex patterns, but training resources are in high demand and existing supervised schemes require expert labeling and regular updates . Therefore, this paper introduces Kitsune, which uses unsupervised, efficient online self-encoder integration to distinguish between normal and abnormal traffic and is optimized to improve detection efficiency In this paper, we experimentally validate that the optimized Kitsune achieves improvements in accuracy, recall, false-positive rate, false-negative rate.