Assessing Cyber Threat Resilience in IoT Networks through Machine Learning–Based DDoS Detection

Abstract

The fast increase in Internet of Things (IoT) networks has made them susceptible to DDoS attacks, which are capable of disrupting its communication and reducing system availability. The work explores machine learning- based DDoS attack detection and compares the performance of Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), Naive Bayes, Random Forest, and an Ensemble approach. Experimental evidence demonstrates that SVC and Random Forest performed highly in balanced classification and lower reliability was obtained with Naive Bayes though computationally inexpensive. Ensemble model was able to outperform all the individual classifiers, and was more robust with a combination of individual classifier strengths. These results underscore Ensemble learning as a better and dependable method of mitigating DDoS attacks in IoT systems.