Novel Memristive Neural Networks And Applications In Information Cryptography

With the rapid development of internet, multimedia, modern science and technology, the digital information can be applied in various fields. Because it is easy to get information, researchers pay more attention to the information security. However, the amount of information is increasing and the complexity of information is deepening, the development of modern electronic circuit has been limited. In the research of neuromorphic system, one of the important problems that we have faced is how to make the neural synapses satisfy the requirement of dealing complex and real information. It is a remarkable fact that memristor is a kind of nanoscale device with the memory and learning functions, so it has natural advantages to be neural synapses.Firstly, the thesis introduces the related background information. Next, relevant theoretic analysis are conducted and two kinds of synapses based on memristor are further designed. The presented synapses replace the traditional synapses to realize the weight update in chaotic neural networks. Next, applications in information cryptography are studied. The specific content of this thesis is as follows:1. In this thesis, the nonlinear memristor model, flux-controlled memristor model and memristor model with forgetting effect are respectively listed with mathematical derivation and theoretical analysis.2. Based on the nonlinear memristor model and flux-controlled memristor model, two memristive electronic synapses are designed to replace the weigh update of neural synapses by the basic theory of electronic circuits and relevant mathematic theory.3. According to the first kind of memristive electronic synapse, a novel Hermite chaotic neural network is proposed to realize the encryption of text information. Specifically, we introduce the whole process of encryption and decryption, and analyze the eventual cryptographic results. In addition, the memristor with forgetting effect is designed to be the electronic synapse in the Hermite chaotic neural network to encrypt the text. We draw a comparison in terms of the two encryption systems’ results.4. According to the second kind of memristive electronic synapse, a novel Laguerre neural network is designed to realize the encryption of picture information. The quality of the cryptographic pictures is analyzed by various methods.In this paper, two kinds of memristive neural networks with novel memristive electronic synapses are investigated to implement encryption of information. The entire research scheme will promote the development of neuromorphic system and improve the security of information encryption. In addition, it can deemed as a potential candidate to realize extra-high integration and bionic intelligent neuromorphic systems in the future.