PSO-based Feature Selection for DDoS Detection using Machine Learning in Smart Agriculture

Distributed denial-of-service (DDoS) attacks pose a significant threat to smart agriculture systems by disrupting services through the saturation of targeted networks with traffic from multiple sources. Real-time detection of these attacks has become a critical cybersecurity task in smart agriculture. However, the analysis of traffic in DDoS attack detection is often hindered by the large number of features present, leading to prolonged processing times. Furthermore, current DDoS detection methods usually have a limited scope regarding the types of attacks they can effectively detect, overlooking the full spectrum of potential DDoS attack scenarios. This work presents a DDoS attack detection using Particle Swarm Optimization (PSO) for feature selection (FS) combined with various machine learning classifiers. The PSO-FS-DDoS-ML method is evaluated on the CICIoT-2023 dataset, which includes twelve distinct types of DDoS attacks. We conducted a comprehensive analysis before and after feature selection in both binary and multiclass classification scenarios, demonstrating improvements in classification accuracy, precision, recall, F1-score, and reductions in false positive rates and computational time. Experimental results show that PSO-FS-DDoS-ML achieves an accuracy of 99.99% for Decision Trees (DT), 99.97% for Random Forest (RF), and 99.93% for k-Nearest Neighbors (KNN) in multiclass classification after feature selection, with 24, 18, and 22 selected features, respectively. In binary classification, PSO-FS-DDoS-ML achieves 100% accuracy with 0% false positive rate (FPR).