A Hybrid Optimized Perplexed Bayes Framework with Automated Response for Multi-Vector DDoS Detection in Cloud Environments

Abstract

Cloud computing security remains a critical concern as adoption accelerates globally. Among the most formidable challenges is the detection of distributed denial-of-service (DDoS) attacks, whose nonlinear nature, atypical traffic behavior, and high-dimensional feature space complicate traditional detection approaches. This research presents a novel Hybrid Optimized Perplexed Bayes (HOPB) classifier with automated response capabilities that extends the perplexed Bayes framework by integrating correlation-based feature selection with Genetic Algorithm (GA) optimization. Unlike the standard perplexed Bayes approach that uses correlation alone [1], our framework employs a two-stage feature selection mechanism combining correlation analysis with nature-inspired optimization to identify the most discriminative feature subsets. Additionally, we introduce an automated response module that initiates predefined security actions—alerting administrators, isolating affected resources, and blocking suspicious activities—upon anomaly detection, as described in the automated anomaly detection and response patent [2]. The proposed system was evaluated on the NSL-KDD and CIC-DDoS2019 datasets, achieving an overall accuracy of 99.37%, with sensitivity of 99.21% and specificity of 99.48%. Comparative analysis demonstrates improvements of 3.8% over standard perplexed Bayes without hybrid optimization and 5.2% over random forest-based detection. Furthermore, the framework incorporates an adaptive threshold mechanism for dynamic traffic pattern recognition, significantly reducing false positives in real-time cloud deployments. These findings establish the HOPB classifier with automated response as a robust, scalable solution for next-generation cloud security architectures.