Artificial Complex-Valued Neural Processing Using Memristive Hyperchaotic Synchronization

This study intends to boost the communication security in Industrial Internet of Things (IIoT) by addressing the alteration challenges of secure key exchange over public network. The objectives are to propose a cognitive Artificial Neural Network (ANN) synchronization combined with memristive hyperchaotic systems towards obtaining better synchronization performance, accuracy, random behaviour and overall security. The proposed method makes use of complex-valued ANN synchronization assisted with memristive hyperchaotic system that has dimensions 4D, 5D and 6D to produce high-entropy cryptographic keys. In this paper, Pseudo-Random Number Generation (PRNG) based on hyperchaotic is used to generate an input vector for ANN synchronization. A bi-directional learning approach is applied in the process of synchronization to improve the distribution of neuronal keys in a public channel between ANNs. A series of experiments were conducted using synchronized speed rate, entropy test as well as accuracy dominant test and performance key extraction phase using precision and recall and also F1-score. The proposed approach outperforms the existing traditional methods. It improves synchronization speed by 15%, synchronization accuracy by 93.8%, precision by 92.0% and F1 score by 93.3%. Therefore, and it can produce higher reliable key generation without any errors because the proposed system is generating high-entropy cryptographic keys with more than 7.5 bits that are able to strengthen IIoT communication against attacks by providing more randomness and security. Besides that, this faster and secured key exchange is possible due to memristive hyperchaotic structures that provide an excellent PRNG support, whereby they produced a remarkable reduction in error rate generated in key synchronization as well as low latency was achieved during key exchange process. An ANN-synchronized key exchange scheme is proposed in this research as the solution for secure IIoT communication problem. The proposed synchronization method uses hyperchaotic systems and coordination of two ANNs, which provides low-latency synchronization with high precision and high F1 score in comparison to the existing methods published in literature. The proposed key generation grows linearly with the number of nodes utilized in a network. Thus, it can also be used for IIoT real-time applications without privacy concerns.