A Comparative Assessment of Deep Learning for Adaptable DDoS Threat Detection in Cloud Computing Systems

A Comparative Assessment of Deep Learning for Adaptable DDoS Threat Detection in Cloud Computing Systems

연구 분야: Networking

학회: SN Computer Science

As cyber threats like Distributed Denial of Service (DDoS) attacks escalate, developing intelligent detection systems is imperative for securing cloud computing. This research proposes a Deep Recurrent Neural Network (Deep RNN) model with Long Short-Term Memory (LSTM) layers to analyze temporal relationships within network traffic for accurate and generalized DDoS attack detection.The Deep RNN model is evaluated on five distinct datasets representing diverse normal and attack conditions including binary classification, real-world attacks, synthesized behaviors, Internet of Things (IoT) attacks, and EV charging infrastructure attacks. For comparative analysis of our proposed model, a Deep Belief Network and Random Forest model are tested on the same datasets to benchmark performance against traditional deep learning and machine learning approaches. Experiments reveal that while the comparative deep learning and machine learning models achieve over 90% accuracy on individual datasets, the proposed Deep RNN model demonstrates more consistent high performance across all evaluation scenarios. Deep RNN maintains over 99% accuracy on real-world and IoT attack data with blended background noise. Overall, outcomes showcase promising capabilities of specialized deep learning architectures similar to our proposed Deep RNN model to harness traffic patterns over time for generalized DDoS detection under shifting network environments in the cloud. This research provides valuable insights and performance benchmarks to guide future efforts in evolving intelligent security systems against increasingly sophisticated threats in the era of cloud computing.