Enhanced Detection of Distributed Denial of Service (DDoS) Attacks in Cloud Computing Using Transfer Learning, DBSCAN, and Entropy Analysis

The present research explores an innovative approach to improve DDoS detection in a cloud environment by evaluating the performance of various algorithms and methodologies. The paper integrates ML and DL approaches to develop a comprehensive framework for the recent developments in information security and the implementation of IDS. It provides an in-depth analysis of DDoS attack types and explores existing detection strategies, emphasizing the role of preprocessing sub-systems and feature selection in enhancing detection accuracy. Leveraging transfer learning and the SDN Dataset that consists of 23 significant features for the detection of intrusion in a network, a transfer learning-based Intrusion Detection System (IDS) is presented in this paper. Using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, this research categorizes attack patterns and noise giving it the flexibility of the anomaly detection method. Entropy is used as a metric to measure the level of randomness in network traffic, with any deviations from expected patterns potentially indicating an ongoing attack. The preprocessing pipeline includes rebalancing data, cleaning, splitting, and applying min–max normalization. A hypermodel combining classifiers such as Support Vector Machines (SVMs), K-Nearest Neighbors (KNN), XGBoost, and other ML models is employed for performance evaluation. The results demonstrate that the proposed deep learning and statistical framework, augmented by DBSCAN and entropy analysis, achieves superior accuracy and robustness in detecting and preventing DDoS attacks compared to traditional methods.