Quasi-synchronization of stochastic memristive neural networks subject to deception attacks

In this paper, the quasi-synchronization problem of stochastic memristive neural networks (MNNs) subject to deception attacks is investigated via hybrid impulsive control. Deception attacks in the MNN synchronization model, which involve the attacker attempting to inject some false data into sensor-to-controller channels to destroy the control signal, are investigated from the perspective of network communication security. The attack conditions are described using stochastic variables that obey the Bernoulli distribution. Inspired by existing impulsive differential inequalities, a new inequality is proposed, which is useful for dealing with quasi-synchronization in impulsive systems. Thereafter, sufficient conditions and the error bound are obtained for validating the quasi-synchronization of stochastic MNNs subject to deception attacks based on the proposed inequality and Lyapunov stability theory. In the absence of an attack, the globally complete synchronization problem for stochastic MNNs is investigated. Additionally, the attack effects and their mitigation through control parameter design are discussed. Finally, the simulation results are presented to validate the theoretical analysis.

Automated summary

In this paper, the quasi-synchronization problem of stochastic memristive neural networks subject to deception attacks is investigated using hybrid impulsive control. Attack conditions are described using stochastic variables, and a new inequality is proposed for dealing with quasi-synchronization in impulsive systems. Sufficient conditions and error bounds are obtained for validating the quasi-synchronization, and the effects of attacks and their mitigation are discussed.