Self-rectifying memristors with high rectification ratio for attack-resilient autonomous driving systems
Guobin Zhang, Xuemeng Fan, Jie Wang, Zijian Wang, Zhejia Zhang, Pengtao Li, Yitao Ma, Kejie Huang, Bin Yu, Qing Wan, Xiangshui Miao, Yishu Zhang
Abstract
With the rise of big data and the Internet of Things, smart devices, especially autonomous driving systems, have become prime targets for information leakage and cyberattacks. This study presents the design and fabrication of a self-rectifying memristor utilizing a TiN/HfOx/Pt structure to enhance the security and reliability of autopilot systems. Following rapid thermal annealing treatment, the self-rectifying memristor demonstrates a recorded rectification ratio exceeding 108 and a nonlinearity of over 105, coupled with minimal device-to-device (3.32%) and cycle-to-cycle variations (1.55%). We further extend the application of self-rectifying memristors into crossbar arrays for the real-time classification of autonomous driving datasets, showcasing their capability to execute artificial neural networks at the hardware level. The proposed crossbar arrays exhibit robust attack resilience, achieving classification accuracy (84.25%) comparable to those of software models (84.34%), particularly under complex attack scenarios. This work not only highlights the potential of self-rectifying memristors in bolstering the security of autonomous driving systems but also offers innovative strategies for safeguarding future intelligent transportation systems. This study presents a self-rectifying memristor (SRM) with high rectification ratio based on a TiN/HfOx/Pt structure. It demonstrates minimal device-to-device and cycle-to-cycle variations, enabling scalability and suitability for in-memory computing. SRM-based crossbar arrays achieve software-comparable accuracy and show promise to enhance autonomous driving security and reliability.