Secure Safety Filter: Towards Safe Flight Control under Sensor Attacks

TL;DR

Differing from existing work focusing on linear, noise-free systems, the proposed secure safety filter handles bounded measurement noise and, by leveraging reduced-order model techniques, is applicable to the nonlinear dynamics of drones.

Abstract

Modern autopilot systems are prone to sensor attacks that can jeopardize flight safety. To mitigate this risk, we proposed a modular solution: the secure safety filter, which extends the well-established control barrier function (CBF)-based safety filter to account for, and mitigate, sensor attacks. This module consists of a secure state reconstructor (which generates plausible states) and a safety filter (which computes the safe control input that is closest to the nominal one). Differing from existing work focusing on linear, noise-free systems, the proposed secure safety filter handles bounded measurement noise and, by leveraging reduced-order model techniques, is applicable to the nonlinear dynamics of drones. Software-in-the-loop simulations and drone hardware experiments demonstrate the effectiveness of the secure safety filter in rendering the system safe in the presence of sensor attacks.

Figures (7)

• Figure 1: Diagram showing the architecture of the secure safety filter operating in the presence of sensor attacks (middle). An attacker tries to induce unsafe behaviors on the drone by injecting attack signals into sensor measurement. The secure safety filter takes past (potentially corrupted) sensor measurements and input sequences, yielding a safe control signal with the least deviation from the nominal control signal. Hardware experiments on a drone verify the approach, with the unsafe behavior when it is under sensor attack (top) and safe behavior with the secure safety filter even when experiencing a sensor attack (bottom).
• Figure 2: Diagram of Software-in-the-Loop (SITL) experiment set-up
• Figure 3: Diagram of hardware experiment set-up
• Figure 4: Phase 1 - Drone flying with a nominal controller together with a safety filter; Phase 2 - Attack initiated on the data entering the Sensor Aggregator; Phase 3 - Secure safety filter active and safe-guarding the nominal controller.
• Figure 5: Quadrotor trajectories in different phases of the SITL simulation under four different attacks. For easier illustration, all the attacks are performed on $y_1$ or $y_5$ measuring $\bm{x}$-axis position. Case (a) Constant value attack - the measurement is set to be constant $2$; Case (b) Noise attack - A Gaussian noise with zero mean and standard deviation $0.5$ is added to the measurement; Case (c) Scale attack - the measurement is set to be $0.2x_1$; Case (d) Shift attack - the measurement is set to be $x_1 + 0.5$.

Theorems & Definitions (3)

• Lemma 1
• proof
• Theorem 1