The Invisible Threat: Simulating and Analyzing the Coordinated Sensor Manipulation Attack (CSMA) on UAVs

Unmanned Aerial Vehicles (UAVs) are increasingly central to a range of applications, yet their vulnerability to covert cyberattacks via sensor data manipulation remains critically underexplored. This deficiency largely stems from the absence of high-quality datasets that simulate realistic attack patterns to train and refine detection algorithms effectively. Addressing this gap, we introduce the Coordinated Sensor Manipulation Attack (CSMA), a sophisticated threat model designed to exploit sensor fusion vulnerabilities in UAVs. Unlike conventional single-sensor spoofing attacks that produce abrupt data modifications easily detected by standard anomaly detection systems, the CSMA method involves the subtle, coordinated alteration of multiple sensor inputs. This strategy induces gradual, undetectable navigation drift, effectively evading current detection mechanisms. To validate our model, we integrate these simulated attacks into the PX4 flight log dataset, applying CSMA to demonstrate scenarios including velocity vector manipulation, multi-sensor coordination, and high-frequency noise injection. Our findings reveal the stealth and severity of CSMA, offering significant advancements in UAV security by providing a dataset that enhances the training of learning detection models with realistic, complex attack scenarios.