DETECTION OF DDOS ATTACK IN A CLOUD COMPUTING ENVIRONMENT USING DEEP LEARNING TECHNIQUE

The security and reliability of cloud computing environments face significant threats from the escalating frequency and sophistication of Distributed Denial of Service (DDOS) attacks, which cause substantial financial losses and service disruptions while often serving as entry points for further system compromise. This research addresses this critical challenge by developing and evaluating deep learning-based detection models using two contemporary datasets: CICDDOS2019 (254,797 normal and 51,404 attack instances with 78 features) and IDS_ISCX_2012. To mitigate class imbalance, a balanced subset of 50,000 instances per class was created through random under-sampling, with optimal feature selection performed using the K-best method. Two advanced recurrent neural network architectures were implemented and compared: Bidirectional Long Short-Term Memory (BI-LSTM) and Gated Recurrent Unit (GRU), both enhanced with temporal attention mechanisms to focus on critical attack patterns within sequential network traffic. Experimental results demonstrated that GRU outperformed BI-LSTM across both datasets, achieving accuracies of 0.93 and 0.65 on IDS_ISCX_2012 and CICDDOS2019 respectively, compared to BI-LSTM's 0.91 and 0.61. The GRU model's simplified architecture proved more computationally efficient while effectively addressing the vanishing gradient problem common in recurrent networks. This study successfully establishes a robust framework for DDOS attack detection in cloud environments, contributing to enhanced network security through improved accuracy, reduced false positives, and practical implement ability for real-time threat mitigation