A Novel Adaptive Ensemble Intelligence Framework for Strengthening Cloud Security Against DDoS Threats for Accurate Detection and Mitigation

Abstract

The high adoption of cloud computing has thoroughly increased the need to have good security controls in order to ensure the flexibility and uniformity of the services. One of the most disruptive categories of emerging cyber threats is the DDoS attacks that are capable of flooding system resources with very large volumes of traffic and advanced obfuscation methods. Traditional machine learning algorithms like K-Nearest Neighbors, Support Vector Machines, Decision Trees, and Logistic Regression will not perform well when exposed to high-dimensional, noisy, as well as imbalanced network data. To overcome these deficiencies, a novel ensemble deep learning structure along with a superior Quadratic Discriminant Analysis (QDA) is introduced in this paper to apply in the decision-level fusion. Deep neural networks are used to extract the hierarchical and discriminative aspects of traffic, and the classification under the overlapping feature distribution is optimized with QDA. Moreover, an adaptive ensemble stacking algorithm combines the results of several base learners, which optimizes predictions that are based on confidence-based weighting. The results indicate that traditional models achieved accuracies between 92% and 97%, while deep neural networks improved to 99.18% accuracy with 0.87% FPR. In contrast, the proposed Adaptive Stacking + QDA framework performed very well when compared with all baselines, reaching 99.43% accuracy, 99.36% precision, 99.25% recall, 99.30% F1-score, and the lowest false positive rate of 0.62%. These findings confirm that the proposed framework has the robustness and reliability for real-time cloud security. Experimental assessments of benchmark cloud-based DDoS datasets show that there are some effective enhancements in detection accuracy, precision, and false-positive reduction, which thoroughly highlight the framework's efficiency in real-time cloud security applications.