Akademik Veri Yönetim Sistemi
Cognitive Models and Artificial Intelligence Conference Proceedings, İstanbul, Türkiye, 25 - 26 Mayıs 2024, ss.7-12, (Tam Metin Bildiri)
Network security is becoming more and more important as a result of the growing use of the internet and technological developments. Intrusion Detection Systems (IDS) are utilized to detect attacks, preferably by monitoring real-time events. Therefore, they hold a significant position in network security. By detecting network attacks based on anomalies, IDSs can ensure network security. Deep learning (DL) can be used to create efficient intrusion detection systems. In this study, three different intrusion detection systems are proposed using Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and their hybrid combination, CNN-LSTM. Deep learning models have complex structures and require adjustment of various hyperparameters, which can be time-consuming when done manually. To address this issue, particle swarm optimization (PSO) is employed in this study for hyperparameter optimization. PSO automatically selects hyperparameters and enhances performance. The CIC-IDS2017 dataset is used to validate the approaches. Models are evaluated with different performance metrics including accuracy, recall, precision, false alarm rate, F1-score, and error rate. The hybrid CNN-LSTM model performs better than other models due to its ability to capture both temporal and spatial features simultaneously. Consequently, the results clearly illustrate the significance and impacts of hyperparameter optimization in intrusion detection systems.