Machine Learning-based Threat Detection for DDoS Prevention in SDN-Controlled IoT Networks
Anshika Sharma, Himanshi Babbar · 2024
The combination of machine learning (ML) methods with Software-Defined Networking (SDN) is examined to detect Distributed Denial of Service (DDoS) attacks. The growing size and sophistication of DDoS attacks is a challenge to conventional defence measures as they continue to pose serious dangers to network infrastructures. SDN-based DDoS detection in practice is explained, including how SDN controllers are deployed to monitor entire networks and how ML models are integrated to continuously monitor and analyse network traffic. The adoption of SDN-based DDoS detection is not without its challenges; these include compatibility with current infrastructure, the choice of suitable ML algorithms, including Support Vector Machine (SVM), Logistic Regression (LR), Random Forest (RF) and K-Nearest Neighbour (KNN) and the requirement for ongoing model training to adjust to changing threats. Beyond just mitigating immediate threats, this integrated approach has the potential to improve network infrastructure resilience generally and lessen the impact of DDoS attacks on vital services. Performance metrics such as recall, accuracy, precision, and F1-score have been applied to these models. The results show that the LR model, with an accuracy rate of 86%, performs better than other models such as SVM, RF and KNN, which have accuracy rates of 71%, 60%, and 65%, respectively, when applied to the UNB CIC-IOT 2023 dataset.