Optimized CNN-Based Intrusion Detection System Using Genetic Algorithm and Particle Swarm Optimization

Ensuring network security and preventing malicious attacks are crucial aspects of modern network management. In this context, various algorithms are employed to differentiate unauthorized access from authorized access, enabling monitoring of network traffic and detecting different types of attacks. However, traditional techniques have faced difficulty maintaining accuracy when predicting malicious activities. This study proposes an effective Intrusion Detection System mechanism designed to detect and classify attacks using the NSL-KDD and UNSW-NB15 datasets to address these challenges. The architecture of the proposed Convolutional Neural Network is optimized through hyperparameter tuning via Genetic Algorithm and Particle Swarm Optimization. As a result, the proposed approach achieves high-performance outcomes, attaining an accuracy of 99.80% with the NSL-KDD dataset and 99.02% with the UNSW-NB15 dataset. This innovative approach demonstrates significant improvement in identifying network intrusions compared to traditional methods. Consequently, it offers a reliable and efficient solution for enhancing network security, providing accurate detection of various attack types while minimizing false positives. The proposed model’s robust performance suggests that it can serve as an effective solution for real-time intrusion detection in modern network settings.