An Enhancing Comprehensive Machine Learning Framework for DDoS Defense Through Leveraging Multiple Algorithms
Bhasha Pydala, Dinasekhar Govardhan, Chapalamadugu Venkatesh, Kornepalli Lokeshwar Goud, Kavadi Dinesh, V. Jyothsna · 2024
Addressing Distributed Denial of Service (DDoS) attacks involves leveraging organizational assets, particularly the website framework. Departing from prior studies relying on an outdated KDD dataset, the paper emphasizes the importance of using the latest data for a comprehensive understanding of the current DDoS landscape. Utilizing machine learning techniques, including Random Forest, k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Gaussian, and Naive Bayes classification algorithms, the study introduces an amalgamated model for classifying and predicting DDoS attack types. The DDoS attack network logs dataset from GitHub and Python as a simulator were utilized. Model performance evaluation involved generating confusion matrices. In the first classification using Random Forest, Precision (PR) and Recall (RE) both reached high values, yielding an average Accuracy (AC) in a satisfactory range (81% to 90%). In the second classification, PR and RE both attained notable values, resulting in an average AC within an acceptable range. The innovation lies in the construction of an integrated model that combines Random Forest, KNN, SVM, Gaussian, and Naive Bayes to enhance overall predictive performance. Notably, a significant improvement in defect determination accuracy was observed when compared to existing research. This underscores the effectiveness of the integrated model in addressing the dynamic landscape of DDoS attacks.