Answers are grounded in this paper and cite their sources.
Demonstrates efficient encryption scheduling in IoT devices, suggesting improved performance and precision.
Homomorphic encryption (HE) lets edge and Internet-of-Things (IoT) devices compute on encrypted data, yet practitioners typically commit to a single scheme at design time. Partially homomorphic Paillier is exact but addition-only and slow on vectors; fully homomorphic CKKS is fast on packed vectors, but its context is memory- and setup-heavy and its smallest parameter sets are numerically unsafe for multiplication. We argue that edge scheme choice is a constrained runtime-scheduling problem rather than one-shot classification, and present a measurement-grounded, resource-aware hybrid HE scheduler that routes each workload across Paillier and CKKS paths under hard operation, precision, memory, and security constraints, while managing the CKKS context lifecycle—build, reuse, evict—within a device memory budget. Validated on a Raspberry Pi 4, two hardware findings drive the design. First, the cheapest CKKS context is numerically unsafe: multiplication error at \(N{=}4096\) reaches \(≈ \) 10 at 256 elements while \(N{≥ }8192\) stays below \(10⁻⁴,\) so a context must be rejected on measured error. Second, because context setup is expensive, building once and reusing is 4.3× faster at five workloads and 14.3× at fifty. Run end-to-end on the board, the scheduler completes a full encrypted task on a constrained device where neither static scheme can. A gradient-boosting model ranks the feasible paths, reaching 95.7% accuracy on unseen workloads at negligible overhead. Where CKKS fits it is the fastest single scheme; the framework’s value is feasibility, precision-safety, and reuse, not selection-driven speed. All paths meet a uniform 112-bit security floor; energy is model-estimated, not board-measured.
No takes yet. Share an insight, caveat, or question.
Bouabidi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: