Memory-Dependent Event-Trigger Scheme for Secure Control of Memristive Neural Networks: Dealing With Deception Attacks
Yingjie Fan, Huang Xia, Zhen Wang
Abstract
This article is concerned with the secure control problem of memristive neural networks (MNNs) subject to deception attacks. To deal with the influence of deception attacks, a secure control scheme with a memory-dependent event-trigger (MDET) scheme is developed for MNNs while the desired performance can be guaranteed. Here, the MDET scheme is designed to remember the memory characteristic of dynamical process. On this basis, an interval-scheduled looped function (ISLF) is constructed. The feature of ISLF lies in that the positivity of Lyapunov functions can be dropped within the scheduled intervals and the rest are saved. Combined with discrete Lyapunov theory, inequality techniques, and continuous Lyapunov theory, some sufficient conditions are presented to ensure that the closed-loop MNNs are mean-square globally asymptotically stable in the presence of deception attacks. In contrast with the previous works, the improvements of the established trigger scheme and ISLF are well discussed. Lastly, simulation results are carried out to verify the effectiveness of the control scheme.