A 28 nm 75.6 KOPS 13 nJ Computing-in-Memory Pipeline Number Theoretic Transform Accelerator for PQC

Lattice-based cryptography (LBC) exploits the learning with errors (LWE) problem and is the main algorithm standardized for Post-Quantum Cryptography (PQC). Number theoretic transforms (NTT) account for most of the latency and energy in the computation of the LWE problem. This brief presents a Compute-in-Memory (CIM) configurable-pipeline NTT accelerator for PQC. The accelerator incorporates a bidirectional pipeline array to minimize data latency, CIM processing elements to reduce memory access, and a parallel PQC circuit for LBC protocol deployment. A 28 nm chip of the accelerator consumes only 13 nJ per 256-point NTT, while achieving a throughput of 75.6 KOPS that achieves a remarkable reduction of up to 78% in clock cycles and a 45% reduction in energy consumption than state-of-the-art designs.