Detecting Malicious Replay Attacks on Drones Using Machine Learning

As Unmanned Aerial Vehicles (UAVs) or drones are increasingly adopted across various sectors, ensuring their safety and security has become a paramount concern. Cyber-attacks, such as replay attacks, can severely disrupt drone operations, leading to real-world consequences like operational failures or deviations from intended flight paths. Therefore, the ability to detect and mitigate replay attacks is crucial for the safety and security of UAVs. In this paper, we propose an intrusion detection system that employs diverse network-based features and machine learning to detect replay attacks. The system utilizes supervised learning, training on features extracted from network packets captured during drone operations to distinguish attack traffic from normal traffic. We evaluate our system using a comprehensive replay attack dataset, which includes 21,096 instances of normal and attack scenarios. Our experimental results demonstrate that machine learning classifiers, including Logistic Regression, Random Forest, Random Tree, SVM, and AdaBoost, effectively detect replay attacks, achieving accuracy rates ranging from 90% to 93.4% by Adaboost. These findings highlight the efficacy of using machine learning classifiers, trained on network packet features, to detect wireless replay attacks, thereby enhancing drone cybersecurity.