With the rapid adoption of wireless technologies and interconnected devices in wireless sensor networks, there has been a huge increase in security issues being brought forward to the Intrusion Detection System (IDS). Traditional IDS face difficulties with high-dimensional wireless traffic, unstable feature relevance, and weak exploration-exploitation balance which leads to higher false positives and limited adaptability. In order to address these drawbacks, we introduced a hybrid optimization method combining Aquila Synergistic Swarm Optimization and Sand Cat Swarm with Support Vector Machine to stabilize feature selection, promote global search, and strengthen SVM’s decision boundary formation leading to more dependable intrusion detection across datasets. This approach simplifies the data complexity, allows only the most significant features to be used, increases detection accuracy, reduces false detection and improves efficiency. The method was used on common benchmark datasets, such as UNSW-NB15, CICIDS-2017 and KDD'99, which showed better performance than other state-of-the-art approaches. In particular, on the UNSW-NB15 dataset, it achieved an accuracy of 98.2%, F1-score of 97.8%, and a feature reduction rate of 85.7%. The proposed approach outperforms baseline models on UNSW-NB15 with a runtime of 125s and shows consistent improvements across other datasets, making it a robust IDS solution for wireless sensor networks.
Yogaraja et al. (Tue,) studied this question.