Integrating artificial intelligence for improved security of IoT-drones through cyber-physical attack detection

Providing cyber-resilient IoE systems has become the need of modern times. In particular, IoT drones are prone to several cyber attacks while navigating in the air. Deliberate transmission of deceptive GPS signals targeted at commercial applications can misdirect global positioning system (GPS)-guided drones, causing them to deviate from their intended paths. Thus, efficient anti-spoofing technology is required to guarantee the safety measures of drone operations. Many techniques for identifying GPS spoofing are available, but most of them need extra hardware, which may not be feasible for tiny or resource-constrained drones. In this regard, this study introduces a specialized method to identify GPS signal spoofing in these drones, called MobileNet. The MobileNet is a convolutional neural network-based transfer learning model that is adopted in this study for drone security along with Chi-square-selected features. The initial phase involves a series of steps to acquire and prepare the GPS signal dataset. Afterward, the dataset is prepared for modeling through preprocessing, data cleaning, and feature extraction. Extensive comparison analysis is performed to evaluate deep learning and transfer learning models. The experimental findings demonstrate the remarkable accuracy of 98.49% by the MobileNet model using Chi-square feature selection. This demonstrates the suitability and capability of the model to perform well in preventing GPS signal spoofing in the context of tiny drone operations.