Fuzz Testing of UAV Configurations Based on Evolutionary Algorithm

With the widespread application of Unmanned Aerial Vehicle (UAV) technology, its security issues have also attracted much attention, among which the configuration attack against the UAV flight control system is one of the current research hotspots. Attackers always upload seemingly normal configuration combinations and cause an imbalance in the UAV state by exploiting configuration item verification vulnerabilities. This paper accumulates flight data through simulation, generates configuration combinations within the security range using differential evolution-based fuzz testing, uses neural networks to guide configuration item variants, and applies these configuration combinations to the AutoTest of UAV flight control systems. The experimental results show that the configuration combinations generated by fuzz testing can guide the UAV to course deviation, spin, crash and other unstable states; the code coverage and function coverage of the position and attitude code library base in the flight control system have also reached a high level.