$H_{\infty }$ Pinning Synchronization of Switched Coupled Neural Networks With Attack Decomposition Technique: A Strictly Monotonic Decreasing LKF Approach
Qihan Qi, Xinsong Yang, Shuang Yuan, Peng Shi, Guanghui Wen · IEEE Transactions on Control of Network Systems · 2025
This paper investigates the$H_{\infty }$pinning global exponential synchronization almost surely (GES a.s.) for switched coupled neural networks (SCNNs) with bounded time-varying delay under output injection attacks. By proposing attack decomposition technique (ADT), the output signal under injection attacks is split into attack-eliminated signal and attack-related signal. For reducing control costs and saving communication resources, the quantized pinning control (QPC) is designed to synchronize the SCNNs by using attack-eliminated signal. Moreover, the attack-related signal is also used to accurately estimate unknown cyber-attacks. A new discretized Lyapunov-Krasovskii functional (LKF) approach is designed to guarantee the value of LKF strictly monotonic decreasing no matter whether switching happened or not, which reduces the conservativeness of obtained results. To guarantee the$H_{\infty }$pinning GES a.s. with minimal non-weighted$\mathcal {L}_{2}$-gain, sufficient conditions formulated by linear matrix inequalities (LMIs) are established. Finally, numerical example is provided to verify the effectiveness of theoretical analysis.