School of Cyber Science and Engineering, Southeast University, Nanjing, 210096, China.
Department of Mathematics, Southeast University, Nanjing, 210096, China.
Neural Netw. 2023 Sep;166:622-633. doi: 10.1016/j.neunet.2023.07.046. Epub 2023 Aug 1.
In this paper, the fixed-time synchronization control for neural networks with discontinuous data communication is investigated. Due to the transmission blocking caused by DoS attack, it is intractable to establish a monotonically decreasing Lyapunov function like the conventional analysis of fixed-time stability. Therefore, by virtue of recursive and reduction to absurdity approaches, novel fixed-time stability criteria where the estimated upper bound of settling-time is inherently different from existing results are presented. Then, based on the developed conditions, an event-triggered control scheme that can avoid Zeno behavior is designed to achieve synchronization of master-slave neural networks under DoS attack within a prescribed time. For comparison, the established control scheme is further discussed under the case without DoS attack, and the circumstance that there is no attack or event-triggered mechanism, respectively. Simulation results are finally provided to illustrate the significant and validity of our theoretical research. Especially, in terms of encryption and decryption keys generated from the synchronization behavior of chaotic networks, we specifically discuss the application of the proposed fixed-time synchronization scheme to image and audio encryption.
本文研究了具有不连续数据通信的神经网络的固定时间同步控制。由于 DoS 攻击导致的传输阻塞，像固定时间稳定性的常规分析那样建立单调递减的 Lyapunov 函数是棘手的。因此，通过递推和反证方法，提出了新颖的固定时间稳定性准则，其中估计的 settling-time 的上界与现有结果本质上不同。然后，基于所提出的条件，设计了一种事件触发控制方案，可以避免 Zeno 行为，从而在给定时间内实现主从神经网络在 DoS 攻击下的同步。为了进行比较，在没有 DoS 攻击的情况下进一步讨论了所建立的控制方案，并分别讨论了没有攻击或事件触发机制的情况。最后提供了仿真结果以说明我们理论研究的有效性和重要性。特别是，就从混沌网络的同步行为生成的加密和解密密钥而言，我们专门讨论了所提出的固定时间同步方案在图像和音频加密中的应用。
