Abstract: In modern network environments, Distributed Denial-of-Service (DDoS) attacks represent a critical security threat, often rendering services unavailable to legitimate users. Conventional intrusion detection techniques fail to adapt to evolving attack strategies and high-dimensional network traffic. This paper proposes a PSO-optimized CNN-LSTM hybrid deep learning model for accurate DDoS detection and classification. The CNN component extracts spatial features from traffic flows, while the LSTM component captures temporal dependencies in sequential data. Particle Swarm Optimization (PSO) is employed to optimize hyperparameters, enabling faster convergence and improved performance. The model is trained and evaluated on imbalanced real-world network datasets. Experimental results indicate significant improvements, achieving over 98% accuracy and enhanced recall and F1-score compared to traditional models. The proposed method demonstrates strong potential for deployment in real-time cybersecurity systems, offering robustness, automation, and adaptability in detecting DDoS attacks.

Keywords: DDoS Detection, CNN-LSTM, Deep Learning, Particle Swarm Optimization (PSO), Cybersecurity, Network Intrusion Detection, Real-Time Security, Imbalanced Dataset, Hyperparameter Optimization, Spatial-Temporal Learning

Downloads: | DOI: 10.17148/IJIREEICE.2025.131044