AI-Driven DDoS Detection for Network Security: A Performance Analysis of Machine-Deep Learning Methods on Network Traffic Data

The spread of the use of internet-enabled devices has significantly increased the risk of DDoS attacks, which can flood systems with malicious information and damage network security. Traditional rule-based detection methods have been considered to lag behind in adapting to the pattern of attack, and this leads to slow or inaccurate detection of threats. This work applies ML and DL models to a real-time DDoS traffic dataset in an attempt to create an AI-based system of an efficient DDoS detection. The methodology can be evaluated by real-time performance analysis and involves extensive preprocessing, feature engineering, and model training based on classifiers RF (Random Forest), KNN (K-Nearest Neighbors), ND (Naive Bayes), Gradient Boosting (GB), SVM (Support Vector Machine), and DNN (Deep Neural Network) and LSTM (Long Short-Term Memory) architectures. Findings indicated that DNN had the best accuracy of 93.50, beating the traditional models and other current models, including Decision Tree, BiLSTM, BERT, and conventional Naive Bayes models, with the best computational performance in real-time implementation. Interpretability methods such as permutation importance, SHAP, and LIME were also used in the study to guarantee model transparency and actionable insights. This proposed approach was demonstrated to be more accurate in detection and balanced in performance over existing approaches, which will offer a powerful, scalable, and interpretable solution to dynamic network environments. The work is additive in nature of an effective and feasible AI-based platform of DDoS detection, and the future implications are the expansion of the dataset, adding flexibility to the system in real-time, and offering hybrid learning capabilities to further advance the degree of safety.