Hybrid BiLSTM-SVM Intrusion Detection with Decision-Based Flow Ranking
Asraa A. Abd Al-Ameer, Alaa Akram Huby · International Journal of Safety and Security Engineering · 2025
Intrusion Detection Systems (IDS) are important in facing the development of cyber threats such as Distributed Denial of Service (DDoS), phishing, and malware attacks, so, their promulgating is important.Bidirectional Long Short-Term Memory (BiLSTM) with Support Vector Machines (SVM) has been integrated and proposed as a hybrid model in this paper to improve detection accuracy and threat response.The proposed system steps include comprehensive data preprocessing, feature extraction using BiLSTM, and classification with SVM, and all these using and leveraging the UNSWNB15 dataset.Deep learning takes advantage of BiLSTM's ability and traditional machine learning leverage from SVM's efficiency so this integration captures temporal patterns through BiLSTM and manages high-dimensional data through SVM.Experimental findings showed that the proposed system is accurately proficient in distinguishing between normal and attack traffic, achieving high levels of accuracy values, such as precision 95%, recall 94%, and F1-scores 95%, where the accuracy value reaches 99%.Besides, to improve the efficiency of threat management, SVM's decision function scores have been used to employ a ranking technique by the proposed system.Therefore, this research highlights the hybrid model value in enhancing IDS performance.