Adaptive observer-based FDI attack detection for quantized output feedback systems

Xin Zhang

doi:10.1016/j.jfranklin.2025.108299

2026-01-15

PRE

Xin Zhang

Yuan Liu

Qiang Ling *

DOI：10.1016/j.jfranklin.2025.108299

原文链接

原文求助

摘要

En 中文

This paper considers the attack detection problem of discrete-time dynamic quantized output feedback systems subject to process noise and false data injection (FDI) attacks. The feedback signals are transmitted over digital communication networks. Attacks occur during the transmission procedure of quantized feedback signals. Due to the limited feedback network bandwidth, quantization saturation may occur so that the quantization error bound is unknown and attack detection is significantly complicated. To resolve this issue, this paper proposes an improved dynamic quantization scheme, which is composed of zooming-out and zooming-in stages, and attack detection schemes for different quantization stages. Specifically, an improved scaling strategy is proposed to ensure the accomplishment of the zooming-out stage in one step, enabling the detection of consecutive zooming-out anomalies caused by attacks. At the zooming-in stage, the scaling parameters gradually decrease so that it becomes harder and harder to detect attacks. To address this challenge, H infinity and H-indices are employed to characterize the robustness of the residual generator against noise and the sensitivity against attacks, respectively. In this way, the impact of attacks is amplified while the influence of system noise and quantization errors is attenuated. Finally, a robust observer-based attack detector with adaptive thresholds is proposed, in which the critical detection gain of the attack detector is obtained by solving a convex optimization problem. Simulations are provided to illustrate the effectiveness of the obtained theoretical results.

Keyword:

Networked control system

False data injection attack

Attack detection

Quantized output feedback