Data-Driven Secure Control for UMVs Against False Data Injection Attacks Using GLU–GRU

doi:10.1109/tr.2026.3702051

Data-Driven Secure Control for UMVs Against False Data Injection Attacks Using GLU–GRU

2026-06-10

0

PRE

AI

DOI：10.1109/tr.2026.3702051

Original

Original request for help

Save

Abstract

En 中文

Since traditional data-driven prediction methods rely on static equivalent data models, their forecasting accuracy cannot be guaranteed when dealing with long-term attack impacts. To address this limitation, this article proposes an attack detection and compensation mechanism based on a hybrid neural network. By incorporating a gated linear unit (GLU) layer into the architecture followed by a gated recurrent unit (GRU) layer, we propose a deep GLU–GRU network, which enhances feature extraction and achieves higher prediction accuracy. Moreover, a dynamic event-triggered mechanism is designed, with adjustable triggering parameters that enable flexible transformation into fixed-threshold or time-triggering modes. This mechanism simultaneously reduces the computational burden of the controller and conserves communication resources. Furthermore, to address the challenge of obtaining continuous system states under event-triggered conditions, we construct an extended state observer using successfully transmitted outputs and the predictive outputs from the GLU–GRU, thereby enabling continuous disturbance estimation. The proposed data-driven control strategy employs an attack compensation mechanism to ensure robust trajectory tracking performance against false data injection attacks. The entire design process relies on input/output data only, and simulation comparisons validate the effectiveness of the proposed methods.

Keywords:

Data-driven control (DDC)

event -triggered (ET) control

false data injection (FDI) attacks

unmanned marine vehicles (UMVs)

Data-Driven Secure Control for UMVs Against False Data Injection Attacks Using GLU–GRU

Abstract

Journal

Organization