Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

QUADFormer: Learning-based Detection of Cyber Attacks in Quadrotor UAVs

Pengyu Wang, Zhaohua Yang, Nachuan Yang, Zikai Wang, Jialu Li, Fan Zhang, Chaoqun Wang, Jiankun Wang, Max Q. -H. Meng, Ling Shi

TL;DR

The paper addresses cyber-attack resilience in quadrotor UAVs by combining EKF-derived residuals with a transformer-based detector, QUADFormer, to capture high-order dynamics and non-Gaussian noise. The method introduces Temporal Proximity Concentration (TPC) and Dynamic Context Mapping (DCM) within a semi-supervised transformer framework and optimizes detection with a Temporal-Contextual Disparity (TCD) loss alongside reconstruction and classification objectives. A resilient state estimation module ensures safe operation by adaptively fusing reliable sensors when attacks are detected. Through extensive simulations and real-world GPS spoofing experiments, QUADFormer demonstrates superior detection performance compared with traditional statistics-based detectors and other learning-based approaches, enabling robust UAV mission execution. The work provides practical implications for secure UAV operation and offers publicly releasable code to foster further research.

Abstract

Safety-critical intelligent cyber-physical systems, such as quadrotor unmanned aerial vehicles (UAVs), are vulnerable to different types of cyber attacks, and the absence of timely and accurate attack detection can lead to severe consequences. When UAVs are engaged in large outdoor maneuvering flights, their system constitutes highly nonlinear dynamics that include non-Gaussian noises. Therefore, the commonly employed traditional statistics-based and emerging learning-based attack detection methods do not yield satisfactory results. In response to the above challenges, we propose QUADFormer, a novel Quadrotor UAV Attack Detection framework with transFormer-based architecture. This framework includes a residue generator designed to generate a residue sequence sensitive to anomalies. Subsequently, this sequence is fed into a transformer structure with disparity in correlation to specifically learn its statistical characteristics for the purpose of classification and attack detection. Finally, we design an alert module to ensure the safe execution of tasks by UAVs under attack conditions. We conduct extensive simulations and real-world experiments, and the results show that our method has achieved superior detection performance compared with many state-of-the-art methods.

QUADFormer: Learning-based Detection of Cyber Attacks in Quadrotor UAVs

TL;DR

The paper addresses cyber-attack resilience in quadrotor UAVs by combining EKF-derived residuals with a transformer-based detector, QUADFormer, to capture high-order dynamics and non-Gaussian noise. The method introduces Temporal Proximity Concentration (TPC) and Dynamic Context Mapping (DCM) within a semi-supervised transformer framework and optimizes detection with a Temporal-Contextual Disparity (TCD) loss alongside reconstruction and classification objectives. A resilient state estimation module ensures safe operation by adaptively fusing reliable sensors when attacks are detected. Through extensive simulations and real-world GPS spoofing experiments, QUADFormer demonstrates superior detection performance compared with traditional statistics-based detectors and other learning-based approaches, enabling robust UAV mission execution. The work provides practical implications for secure UAV operation and offers publicly releasable code to foster further research.

Abstract

Safety-critical intelligent cyber-physical systems, such as quadrotor unmanned aerial vehicles (UAVs), are vulnerable to different types of cyber attacks, and the absence of timely and accurate attack detection can lead to severe consequences. When UAVs are engaged in large outdoor maneuvering flights, their system constitutes highly nonlinear dynamics that include non-Gaussian noises. Therefore, the commonly employed traditional statistics-based and emerging learning-based attack detection methods do not yield satisfactory results. In response to the above challenges, we propose QUADFormer, a novel Quadrotor UAV Attack Detection framework with transFormer-based architecture. This framework includes a residue generator designed to generate a residue sequence sensitive to anomalies. Subsequently, this sequence is fed into a transformer structure with disparity in correlation to specifically learn its statistical characteristics for the purpose of classification and attack detection. Finally, we design an alert module to ensure the safe execution of tasks by UAVs under attack conditions. We conduct extensive simulations and real-world experiments, and the results show that our method has achieved superior detection performance compared with many state-of-the-art methods.
Paper Structure (23 sections, 2 theorems, 34 equations, 11 figures, 2 tables, 3 algorithms)

This paper contains 23 sections, 2 theorems, 34 equations, 11 figures, 2 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Residue-based Attack Detection
  4. Learning-based Attack Detection
  5. Preliminaries
  6. UAV Model
  7. Extended Kalman Filter
  8. False Data Injection Attack
  9. Performance Metric and Problem Formulation
  10. Methodology
  11. UAV Residue Generator
  12. UAV Attack Detector
  13. Overall Architecture
  14. UAV Attack Attention
  15. Loss Function
  16. ...and 8 more sections

Key Result

Lemma 1

reif1999stochastic Consider the UAV system given by (eq:system_model) and the extended Kalman filter described in (eq:ekf_predict) - (eq:ekf_update). If Then the estimation error is bounded in mean square, and moreover it is almost sure bounded, i.e.,

Figures (11)

  • Figure 1: Illustration of UAV operation under cyber attacks.
  • Figure 2: Overview of QUADFormer framework. The framework starts with an effective residue generator (right, Sec. \ref{['sec:residue_generator']}). The residue sequences are then fed into a novel attack detector (middle, Sec. \ref{['sec:attack_detector']}). The UAV achieves resilient state estimation based on the detection results (left, Sec. \ref{['sec:state_estimation']}).
  • Figure 3: Real-world experiment setup.
  • Figure 4: UAV autonomous "Eight-shape" flight in three different cases.
  • Figure 5: UAV autonomous "UST-shape" flight in three different cases.
  • ...and 6 more figures

Theorems & Definitions (2)

  • Lemma 1
  • Theorem 1