Self-rectifying memristors with high rectification ratio for attack-resilient autonomous driving systems

Area of Science:

• Materials Science
• Electrical Engineering
• Computer Science

Background:

• Smart devices, including autonomous driving systems, are vulnerable to cyberattacks and data breaches due to increasing big data and IoT adoption.
• Enhancing the security and reliability of autonomous driving systems is critical for the widespread adoption of intelligent transportation.

Purpose of the Study:

• To design and fabricate a self-rectifying memristor for improved security in autonomous driving systems.
• To evaluate the performance and attack resilience of memristor-based crossbar arrays for real-time data classification.

Main Methods:

• Fabrication of TiN/HfOₓ/Pt self-rectifying memristors using rapid thermal annealing.
• Characterization of device performance, including rectification ratio, nonlinearity, and device variations.
• Implementation of memristor crossbar arrays for artificial neural network execution and autonomous driving dataset classification.

Main Results:

• The self-rectifying memristor achieved a rectification ratio >10⁸ and nonlinearity >10⁵ with low device variations (3.32% device-to-device, 1.55% cycle-to-cycle).
• Memristor crossbar arrays demonstrated robust attack resilience, achieving classification accuracy of 84.25% on autonomous driving datasets.
• Performance was comparable to software models (84.34%) even under complex attack scenarios.

Conclusions:

• Self-rectifying memristors offer a promising solution for enhancing the security and reliability of autonomous driving systems.
• Memristor-based crossbar arrays can perform hardware-level artificial neural network computations, crucial for real-time data processing in intelligent vehicles.
• This research provides innovative strategies for securing future intelligent transportation systems against cyber threats.