Optimising DDoS Attack Detection Using Feature Engineering and Machine Learning Techniques in High-Volume Network Traffic Analysis

Background: DoS attacks, including DDoS, are common cyber threats that disrupt network services; Methods: This study focuses on optimizing DoS attack detection in highvolume traffic using the BoT-IoT dataset. A preprocessing pipeline was implemented involving missing value handling, categorical data encoding, and Z-score normalization for feature standardization. Key features, such as pkSeqID, time, and lime, were selected for model training; four classification models-MLP, NB, SVM, and LR-were evaluated based on F1-score, recall, accuracy, and precision. Results: The MLP model fared better than the others, obtaining 96.42% F1-score, 95.30% recall, 98.83% accuracy, and 97.62% precision; Concluding: The findings highlight MLP's effectiveness as a scalable and efficient solution for DoS attack detection in highvolume networks, providing a reliable method to secure modern network infrastructures from evolving cyber threats.