A High-Throughput Parasitic TRNG in Self-Rectifying Memristor Based CIM for Edge Secure Computing

Edge computing systems are resource-constrained and security-critical, necessitating on-chip TRNG with high area efficiency. For the first time, we demonstrated a parasitic TRNG in a self-rectifying memristor (SRM) based CIM array by sharing the same array and peripheral circuit with CIM. The resistance fluctuations at low read voltage are utilized to generate entropy. Thanks to the high parallelism, the proposed TRNG achieves a high throughput of 2.56 Gb/s, meeting the demands for frequent cryptographic key generation of edge computing system. The parasitic-in-CIM method presented in this work can be extended to other devices exhibiting resistance fluctuations. Our research offers an area-efficient and high-throughput solution for future CIM-based edge secure computing systems.