SSD: A State-Based Stealthy Backdoor Attack for IMU/GNSS Navigation System in UAV Route Planning

Z. X. Wang, Yang Li, Jie Zhang, Xingshuo Han, Kangbo Liu, Yang Lyu, Yuan Zhou, Tianwei Zhang, Quan PanIEEE Trans. IFS 2025 · DOI

TL;DR

This paper introduces SSD, a novel "state-triggered backdoor attack" for Unmanned Aerial Vehicles (UAVs) that exploits their motion dynamics to spoof GNSS navigation data. By cleverly timing attacks during specific maneuvers, SSD significantly degrades UAV positioning accuracy and mission success rates while remaining highly stealthy against common detection methods.

What Problem Does This Paper Address?

Unmanned Aerial Vehicles (UAVs) are increasingly relied upon for critical missions, making their navigation systems a prime target for cyberattacks. While Global Navigation Satellite System (GNSS) spoofing is a known threat, existing research often overlooks how a UAV's own movement affects the success and detectability of these attacks. Specifically, the dynamic nature of UAV maneuvers, like sharp turns or changes in speed, can create windows of vulnerability that are not well understood or exploited. This paper aims to bridge this gap by investigating how a UAV's motion state influences GNSS spoofing effectiveness and stealthiness, and by proposing a new attack that leverages these dynamics.

Methods

The core of this paper is the State-based Stealthy Backdoor (SSD) attack. Unlike traditional attacks that might be constantly active or require complex computations to evade detection, SSD acts like a "backdoor" that is triggered by specific conditions.

Here's how it works:

  • Motion State Analysis: The authors first analyze how different UAV motion states (linear vs. nonlinear) impact the effectiveness of GNSS spoofing. They find that nonlinear motion states amplify positioning errors and make speed-related attacks harder to detect.
  • State-Triggered Backdoor: SSD uses the UAV's motion state as a trigger. Specifically, it leverages periods of acceleration (indicating a change in motion) as the trigger.
  • Combined Attack: When triggered, SSD executes a combined attack on both the UAV's position and velocity measurements from GNSS. This is done in a staged manner to maximize impact.
  • Stealthiness: The attack is designed to be stealthy by carefully tuning parameters to keep detection statistics within acceptable thresholds. It avoids complex optimization, relying instead on carefully chosen parameters and straightforward calculations.
  • Navigation System Model: The paper models the UAV's navigation system using an Extended Kalman Filter (EKF), a standard approach for fusing data from Inertial Measurement Units (IMUs) and GNSS. They analyze how attacks affect the EKF's state estimation.

Datasets

The paper does not explicitly mention training datasets. For evaluation, they used three representative mission trajectories:

  • Trajectory I: Straight-line Path: Represents typical linear motion for simple tasks.
  • Trajectory II: Spiral Path: Represents nonlinear motion, involving significant changes in altitude and direction.
  • Trajectory III: U-shape Path: A hybrid trajectory combining linear and nonlinear motion, designed for high-precision tasks.

These trajectories were simulated using the PX4 autopilot's ECL EKF2 navigation algorithm.

Key Insights and Ideas

  • Motion Matters: A UAV's movement is not just a passive factor; it actively influences its vulnerability to cyberattacks. Nonlinear motion states, in particular, create opportune moments for attackers.
  • Exploiting Vulnerable Windows: Instead of a constant attack, SSD strategically targets specific motion states (like acceleration during maneuvers) to maximize damage and stealth. This "backdoor" approach is more efficient and harder to detect.
  • Velocity Attacks are Stealthier in Nonlinear Motion: While positional attacks are generally more impactful, velocity attacks become significantly stealthier when the UAV is in nonlinear motion. This is a key insight for designing more sophisticated attacks.
  • Combined Attacks are More Effective: Attacking both position and velocity simultaneously, especially when triggered by motion states, yields much higher localization errors than single-point attacks.
  • Mission Impact is Crucial: Evaluating attacks solely on navigation error is insufficient. The paper introduces Average Per-Waypoint Displacement Error (APDE) to better assess how attacks impact the UAV's ability to complete its mission objectives.

Key Results

  • Increased Positioning Errors: SSD significantly increases positioning errors, with Average Displacement Error (ADE) and Final Displacement Error (FDE) increasing by an average of 425% and 591%, respectively, across different scenarios.
  • Compromised Mission Completion: APDE increased by up to 635.96%, indicating a severe impact on the UAV's ability to reach waypoints and complete missions. In one scenario, deviations reached over 10 meters.
  • High Stealthiness: SSD maintained detection statistics below the threshold for state-of-the-art detectors (chi-square, NLC, LTW) across various trajectories, demonstrating its effectiveness in evading detection.
  • Generalizability: The attack proved effective against different EKF-based navigation frameworks (ECL EKF2 and CD-EKF), showing its potential applicability across various UAV platforms.

Limitations and Future Work

  • Dependency on a Priori Knowledge: The current SSD implementation relies on knowing the UAV's motion state (specifically acceleration) in real-time. This requires accurate object detection and tracking systems, which can be challenging.
  • Adaptive Qk and Rk: The paper notes that the current attack uses fixed values for process noise covariance (Qk) and measurement noise covariance (Rk). Future work could explore adaptive strategies for these parameters to potentially mitigate attacks or enhance their effectiveness.
  • Energy Consumption: While stealthy, sustained attacks might lead to increased energy consumption on the attacking end, which could be a practical limitation.
  • Defense Strategies: The authors suggest that future research should focus on developing robust dynamic defenses that can adapt to changing motion states and attack strategies. This could involve dynamically adjusting Qk and Rk or employing more sophisticated real-time defense mechanisms.

Summary updated

Abstract

Unmanned aerial vehicles (UAVs) are increasingly employed to perform high-risk tasks that require minimal human intervention. However, they face escalating cybersecurity threats, particularly from GNSS spoofing attacks. While previous studies have extensively investigated the impacts of GNSS spoofing on UAVs, few have focused on its effects on specific tasks. Moreover, the influence of UAV motion states on the assessment of cybersecurity risks is often overlooked. To address these gaps, we first provide a detailed evaluation of how motion states affect the effectiveness of network attacks. We demonstrate that nonlinear motion states not only enhance the effectiveness of position spoofing in GNSS spoofing attacks but also reduce the probability of detecting speed-related attacks. Building upon this, we propose a state-triggered backdoor attack method (SSD) to deceive GNSS systems and assess its risk to trajectory planning tasks. Extensive validation of SSD’s effectiveness and stealthiness is conducted. Experimental results show that, with appropriately tuned hyperparameters, SSD significantly increases positioning errors and the risk of task failure, while maintaining high stealthy rates across three state-of-the-art detectors.
BibTeX
@article{wang2025ssd,
  author = {Z. X. Wang and Yang Li and Jie Zhang and Xingshuo Han and Kangbo Liu and Yang Lyu and Yuan Zhou and Tianwei Zhang and Quan Pan},
  title = {{SSD: A State-Based Stealthy Backdoor Attack for IMU/GNSS Navigation System in UAV Route Planning}},
  journal = {IEEE Transactions on Information Forensics and Security},
  year = {2025},
  doi = {10.1109/tifs.2025.3646036},
}

Loading related content...

Figure