Tiny ML-based Secure and Energy Efficient Unmanned Aerial Vehicles Surveillance Framework for Smart Cities

Recent advances in smart cities require a more effective, secure, and energy-efficient surveillance system to protect people. This paper proposes a multi-layered architecture for a UAV-based traffic monitoring system, designed to enhance situational awareness and system resilience in real-time traffic environments. The architecture consists of four key layers: the Traffic Layer, where the UAV continuously monitors and gathers data from diverse traffic entities; the UAV Layer, responsible for data collection and packet transmission; the Security Layer, which employs quantized TinyML models, including decision trees, artificial neural networks (ANN), and support vector machines (SVM), to detect and mitigate Denial-Of-Service (DoS) attacks; and the Controller Layer, which manages verified data and orchestrates system actions based on traffic conditions and cyber security alerts. Implementing machine learning algorithms optimized for resource efficiency using quantization techniques ensures secure communication, real-time data processing, and energy conservation within the UAV. Our results demonstrate the system’s ability to maintain robust performance under complex traffic conditions while effectively mitigating potential cyber security threats in an energy efficient manner.