Review of security techniques for memristor computing systems

Neural network (NN) algorithms have become the dominant tool in visual object recognition, natural language processing, and robotics. To enhance computational efficiency of these algorithms, comparison to traditional von Neuman computing architectures, researchers been focusing on memristor systems. A major drawback when using systems today is that, artificial intelligence (AI) era, well-trained NN models are intellectual property and, loaded systems, face theft threats, especially running edge devices. An adversary may steal through advanced attacks such as learning side-channel analysis. In this paper, we review different security techniques for protecting Two threat described based their assumptions regarding adversary’s capabilities: a black-box (BB) model white-box (WB) model. We categorize existing into five classes context models: thwarting (BB), encryption (WB), weight transformation fingerprint embedding (WB). also present cross-comparison limitations techniques. This paper could serve an aid designing secure