Enhanced intrusion detection using PSO feature selection with LSTM gated RNN based on inactive packet bit classification in active routing wireless transmission

Intrusion Detection Systems (IDS) are essential for safeguarding wireless networks, particularly with the increasing frequency of sophisticated cyber threats. Traditional IDS approaches face challenges with high false positive rates and the complexity of identifying intrusions in dynamic network environments. This research introduces a novel IDS framework that combines Particle Swarm Optimization (PSO) for optimal feature selection with a Gated Long Short-Term Memory (LSTM) network for accurate intrusion classification. PSO reduces feature dimensionality, enhancing computational efficiency by selecting only the most relevant attributes from network traffic data. The Gated LSTM network is applied to capture temporal dependencies, allowing for precise classification of active and inactive packet bits to detect real-time anomalies. This combined PSO-LSTM approach effectively minimizes false positives, reduces computational complexity, and ensures robust intrusion detection, proving highly efficient for security in active wireless network environments. This framework's enhanced detection rates and reduced time complexity highlight its potential to set a new benchmark in IDS performance for wireless networks. The performance metrics, such as precision, accuracy, recall, and F1-score, improve intrusion detection based on inactive packet bit classification in active routing wireless transmission. Experimental results evaluated with the KDDcup99 dataset show that our proposed model achieves a classification accuracy of 93.61%, with a precision rate of 94.26%, recall rate of 94.01%, and F1-score of 94.03%, outperforming the conventional deep learning methods such as Deep Neural Networks (DNN) and Fuzzy Neural Networks (FNN).