@article{article_1658662, title={Advancing Multi-Class Intrusion Detection: A Comparative Evaluation of LSTM and Bi-LSTM on Class-Imbalanced CIC-IDS-2017}, journal={Turkish Journal of Engineering}, volume={9}, pages={578–590}, year={2025}, DOI={10.31127/tuje.1658662}, author={S.p., Senthilkumar and Balasubramanian, Suresh Kumar}, keywords={Multi-ClassIntrusion Detection, LSTM, Bi-LSTM, Class-Imbalanced, CIC-IDS-2017}, abstract={In view of continuously evolving cyber-attacks, intrusion detection systems play a crucial role in modern network infrastructures. Traditional methods conventionally rely on rule-based systems, which cannot scale well with the increasing complexity and diversity in network threats. This paper presents the application of Long Short-Term Memory and Bidirectional Long Short-Term Memory on multiclass intrusion detection using the CIC IDS 2017 dataset containing benign and malicious network traffic data. A combined preprocessing strategy of random undersampling and SMOTE was used to address the challenge of class imbalance. Both LSTM and Bi-LSTM architectures were studied for accurate classification of network behaviors. The various metrics adopted for the performance evaluation included accuracy, precision, recall, F1-score, and confusion matrix analysis. It has shown that the Bi-LSTM network is better compared with the LSTM model due to considering the contextual information in both directions, which is pretty helpful for those attack types with complicated temporal relationships. This leads to the thought that deep learning methods may boost the robustness and accuracy of an IDS significantly and, in this respect, one shall investigate the technique of Bi-LSTM.}, number={3}, publisher={Murat YAKAR}