Comparative Analysis of Machine Learning Models for Enhancing Cybersecurity on Cyber-physical Systems in Smart Grids Against DDoS Attacks

Abstract

Detecting Distributed Denial of Service (DDoS) attacks in cyber-physical systems, particularly smart grids, requires highly accurate and efficient solutions. This study evaluates the performance of several machine learning algorithms, including Logistic Regression, Naive Bayes, K-Nearest Neighbors, Decision Trees, Support Vector Machine, Random Forest, Gradient Boosting Machines, XGBoost, Artificial Neural Networks, and Recurrent Neural Networks for detecting DDoS attacks. The CICIDS2017 dataset, which includes real-world attack scenarios, was used for training and testing. The evaluation metrics, such as precision, recall, accuracy, and F1-score, demonstrate exceptional performance across most algorithms, with XGBoost achieving perfect scores on all metrics. Other models, such as RF, DT, and GBM, also show near-perfect performance, while simpler models like Naive Bayes, though slightly lower, still provide viable detection capabilities. These results emphasized the importance of advanced machine learning algorithms in ensuring the security and stability of critical infrastructure like smart grids.