A Two-Layer Deception Attack Detection Framework for UAV

doi:10.1109/JIOT.2026.3679816

A Two-Layer Deception Attack Detection Framework for UAV

2026-04-01

0

PRE

AI

DOI：10.1109/JIOT.2026.3679816

Original

Original request for help

Save

Abstract

En 中文

This article investigates the problem of real-time deception attack detection for unmanned aerial vehicle (UAV) under resource-constrained conditions. First, deception attacks are classified into trajectory hijacking attacks and constant-bias attacks according to their dynamic characteristics. Then, we develop a physically interpretable two-level XGBoost feature construction that decouples transient anomaly detection from long-term bias identification. Furthermore, an event-triggered activation mechanism is developed to effectively reduce computational burden while maintaining the reliable detection performance. The verification results of the testing set show that for trajectory hijacking attacks, the detection accuracy of the proposed framework reaches 95.8%, and for constant deviation attacks, the detection accuracy of the proposed framework is 98.9%. Moreover, through analysis of complexity and cost, it is demonstrated that the proposed method is lightweight and highly efficient. Real-world UAV flight experiments further verify the effectiveness and practicality of the proposed method under real deception attacks, highlighting its suitability for real-time onboard deployment.

Keywords:

Deception attack detection

event-triggered mechanism

two-level feature construction

unmanned aerial vehicle (UAV)

XGBoost

A Two-Layer Deception Attack Detection Framework for UAV

Abstract

Journal

Organization