Low-Cost Neural Network Spoof Detection for Unmanned Aerial Systems (UAS)​

This project develops a real-time, hardware-in-the-loop (HIL) test platform for simulating and detecting spoofed UAV sensor data. The system combines a virtual sensing environment, a real-time middleware bridge, and an onboard neural network to evaluate how LiDAR, GPS, and multi-sensor data can be manipulated and reliably identified in adversarial conditions.

Project Goals

• Simulate realistic and adversarial UAV sensor environments
• Generate spoofed and authentic LiDAR/GPS data streams
• Detect spoofing using embedded neural networks
• Support SIL and HIL validation workflows
• Provide scalable, low-cost research platform

System Overview

The platform integrates:

• Unreal Engine 5 for virtual LiDAR generation and spoofing
• Simulink Bridge for real-time data synchronization
• UAV Edge Device (Jetson) for neural inference
• Ground Control Station for configuration and monitoring

Use this site to explore the architecture, schedule, testing strategy, and project resources.