How Much Energy Does Future‑Proof Cryptography Use on Our Raspberry Pis?

Cryptography is about to undergo a radical shift. Quantum computers, using Shor's algorithm, threaten current encryption methods like ECDH that secure key exchanges online. To counter this, NIST has begun standardizing post‑quantum (PQC) algorithms and selected ML‑KEM (FIPS‑203) as a potential replacement. Yet a practical question remains: how much energy does such a complex algorithm consume, especially on tiny devices like the Raspberry Pi? The new study in JCP introduces a framework to measure energy use during PQC key generation. Researchers connected a Raspberry Pi to a current meter, installed OpenSSL with ML‑KEM support, and compared results with traditional techniques. Findings reveal that although ML‑KEM performs more mathematical operations, its power draw stays within acceptable limits for IoT applications. This marks a crucial step toward making quantum‑resistant security feasible on low‑cost hardware, paving the way for a network more resilient to the quantum future.

The proposed framework relies on CPU‑level energy profiling. The Raspberry Pi 4, equipped with a 1.5 GHz ARM Cortex‑A72 processor, was powered through a supply monitored by a high‑resolution current sensor (0.1 mA precision). The measurement software logged consumption in milliseconds, isolating the key‑generation phase. ML‑KEM, a lattice‑based cryptosystem, performs polynomial‑ring matrix multiplications, typically with parameters n=768 and q≈3329 for medium security. These operations demand more clock cycles than ECDH, which relies on elliptic‑curve scalar multiplication. Nonetheless, analysis showed an average energy use of about 12 mJ per ML‑KEM key generation, versus 5 mJ for ECDH. The gap stems mainly from increased memory accesses and modular arithmetic. To further optimize, authors suggest pre‑computation techniques, reducing polynomial coefficient precision, and implementing hardware‑accelerated versions on FPGA or ASIC. Additionally, the evaluation considered ambient temperature effects, as thermal variations affect processor supply voltage. In IoT scenarios, where devices run long on limited batteries, opting for lower security parameters (e.g., ML‑KEM‑512) can cut consumption by up to 30 % while still offering robust quantum‑resistant protection. This study provides a quantitative foundation for embedded system designers balancing security, performance, and energy autonomy.