DDoS Attack Detection Using Smart IDS

Abstract

Among the major threats to network security are the Distributed Denial of Service (DDoS) attacks, leading to significant disruptions and financial losses across various industries. Traditional Intrusion Detection Systems (IDS) often fall short in effectively identifying and countering these attacks due to their increasing complexity and scale. To address this challenge, this paper presents a novel approach to DDoS detection by incorporating Suricata, an open-source IDS, with Machine Learning (ML) techniques, thereby creating a Smart IDS. This integration enhances Suricata's robust rule-based detection capabilities with the predictive power of ML models, resulting in more precise and timely identification of DDoS threats. The system's architecture was evaluated using the extensive CIC2023 dataset, supplied by researchers from the Canadian Institute for Cybersecurity (CIC), which contained both benign and malicious traffic. The analysis focused on features like packet rate, flow duration, and entropy. Various ML models, including Decision Trees, Random Forest (RF), and Deep Neural Networks (DNN), were evaluated to identify the best fit for integration with Suricata. The experimental results demonstrate that the Smart IDS significantly enhances DDoS attack detection compared to traditional approaches, especially in minimizing false positives. Suricata alone reached an accuracy of 87 %, but integrating it with the ML models raised detection accuracy to 99 %. This research contributes to the advancement of network security by providing a scalable and adaptable solution for real-time DDoS detection, with promising applications in both academic research and enterprise settings.