Feature Extraction and Ranking using Fuzzy-Ksi for Network Intrusion Detection using a Krill Herd Optimized Neural Network

The rising frequency of cyberattacks highlights the urgent need for reliable network intrusion detection systems (NIDS). High dimensionality data present problems for existing NIDS, increasing processing complexity and inefficiency. This research presents a novel method for optimizing a feedforward neural network (FFNN) using the Krill Herd Algorithm (KHA) and the Fuzzy Kulczynski Similarity Index (KSI) for feature ranking. The major conclusions show that, when compared to benchmarks, the proposed algorithm greatly improves NIDS's performance, obtaining an accuracy of 99.5%. This study shows that the combined feature ranking and optimization strategy is better than conventional methods, advancing the development of intelligent and adaptive NIDS that can handle developing cyber threats.