Stealthy Attacks against Robotic Vehicles Protected by Control-based Intrusion Detection Techniques

Abstract

Robotic vehicles (RV) are increasing in adoption in many industrial sectors. RVs use auto-pilot software for perception and navigation and rely on sensors and actuators for operating autonomously in the physical world. Control algorithms have been used in RVs to minimize the effects of noisy sensors, prevent faulty actuator output, and, recently, to detect attacks against RVs. In this article, we demonstrate the vulnerabilities in control-based intrusion detection techniques and propose three kinds of stealthy attacks that evade detection and disrupt RV missions. We also propose automated algorithms for performing the attacks without requiring the attacker to expend significant effort or to know specific details of the RV, thus making the attacks applicable to a wide range of RVs. We demonstrate the attacks on eight RV systems including three real vehicles in the presence of an Intrusion Detection System using control-based techniques to monitor RV’s runtime behavior and detect attacks. We find that the control-based techniques are incapable of detecting our stealthy attacks and that the attacks can have significant adverse impact on the RV’s mission (e.g., deviate it significantly from its target, or cause it to crash).

Security Analysis against Spoofing Attacks for Distributed UAVs

“…The ground station is a computer that runs ground control software that provides a Motors, Propellers, Sensors (GPS, IMU etc.,) user interface such as a map or coordinates to operate the UAV. Examples include Q Ground Control, 3 APM Planner 2.0 4 and Mission Planner. 5 In addition, the ground station often uses libraries such as dronekit 6 that enables the use of scripts to create and send messages to the UAV.…”

“…We assume that the adversary has the capability to manipulate the sensor values (such as GPS, accelerometer and gyroscope) and these manipulated values are reflected in the audit by other drones. The attack vector we assume is similar to that of [3]: a malicious flight control library is on the system. However, we assume that the adversary can perform whitebox attacks on the model (i.e., the adversary is aware of the parameters of the anomaly detector model).…”

Demo Paper: Impact of Stealthy Attacks on Autonomous Robotic Vehicle Missions

“…Control Invariants (CI) [3] and Extended Kalman Filter (EKF) [1] are techniques that use control-based estimations to derive invariants and monitor the RV's runtime behaviour to detect attacks. In this paper, we demonstrate three stealthy attacks, namely false data injection (FDI), artificial delay (AD), and switch mode (SM) attacks, based on the findings of our previous work [4], [5]. The stealthy attacks evade the CI and EKF detection techniques and result in significant adverse impact on the RV's mission (e.g., significant deviations from the target or result in a crash).…”

DPM: Towards Accurate Drone Position Manipulation

“…The best solution for this case is to guide the drone away to a designated area for safe handling. While several projects [4], [5], [6], [7], [8], [9], [10], [11] have demonstrated the feasibility of such attacks, none of them is able to provide quantitative control on practical systems. Different from all these existing methods, this paper focuses on accurate quantitative control by systematically investigating…”

“…An ideal attack is to gain the complete control of an invading drone, e.g., if attackers can take over its control channel (by cracking its encryption) or hack into its control software. While several methods have been developed to exploit specific drone settings, they require directly compromising drone software, sensors, or communication channels [8], [12], [13], [14], [15], [16], which are often difficult to achieve in practice. While these methods may work well on weak systems with known vulnerabilities, it is impractical to solely rely on such methods because the vulnerabilities can be easily patched.…”