(An) analysis of acoustic injection attacks on MEMS IMU of drones and its mitigation

Abstract

Drones equipped with microelectromechanical system (MEMS) inertial measurement unit (IMU) sensors are susceptible to acoustic injection attacks, which compromise the sensors’ output and can lead to drone crashes. Despite various proposed mitigation strategies, their practical limitations prevent them from ensuring the safe arrival of the drone at its intended destination in the presence of an attack. The objective of this research is to address the challenge of acoustic injection attacks by recovering compromised sensor values, enabling effective mitigation. To achieve this, a realistic testbed was constructed, followed by an extensive study for investigating the impact of resonant MEMS sensors on drones. This investigation revealed the crucial role of sampling jitter, which refers to inconsistent timing delays in retrieving sensor values, in drone crashes during an attack. While sampling jitter is typically considered negligible for real-time requirements, it was found to be a critical factor when drones are subjected to attacks. This is because sampling jitter causes resonant sensor signals to spread into the drones’ control logic’s in-band range, bypassing safety mechanisms like low-pass filters.