Integrated Feature-Based Network Intrusion Detection System Using Incremental Feature Generation

Machine learning (ML)-based network intrusion detection systems (NIDSs) depend entirely on the performance of machine models. Therefore, many studies have been conducted to improve ML Nevertheless, relatively few focused feature set, which significantly affects In addition, features are generated by analyzing data collected after session ends, requires a significant amount memory and long processing time. To solve this problem, study presents new set existing NIDSs. Current session-feature-based NIDSs largely classified into using single-host multi-host set. This research merges two different sets an integrated is used train model for NIDS. incremental generation approach proposed eliminate delay between end time creation The improved NIDS was confirmed through experiments. Compared based models sets, with improves rate 4.15% 5.9% average, respectively.