Detecting Attacks and Optimizing Routes in Radio-frequency Networks Using Machine Learning and Graph Theory

In the context of the widespread use of Radio Frequency (RF) communication networks as in electronic warfare, ensuring security and optimization has become increasingly important. This study investigates methods for detecting attacks and determining optimal routes within RF networks. In this paper, we investigate a novel framework that can proactively detect attacks in RF-based Electronic Warfare (EW), find the signal blockage to install anti-jammer and recommend an optimal path for mission success. First, we propose a logistic regression-based machine learning (ML) mechanism to train a model to differentiate between attack signals vs normal communications. Second, we devise a state–action–reward–state–action (SARSA)-based reinforcement learning (RL) scheme to find an end-to-end path for reaching to RF-enabled mission target. Third, we have A* to restore connectivity by deploying anti-jammers in optimal places. Finally, we have used a real-world Electronic Warfare (EW) dataset to evaluate the proposed framework. Our experiment shows that the proposed logistic regression produces reasonably accurate attack detection, with 98% correct classification. Further, we have achieved higher accuracy in path planning with our RL agent, where normalized Euclidean distance error between 0.1 to 0.3 as compared to the optimal distance. These results showcase the feasibility of integrating machine learning and graph theory to enhance the security and optimization of RF networks.