An Adaptive Resource-Aware MK-CKKS Framework with Linear- Complexity Multi-Key Aggregation for Privacy-Preserving Federated Learning in Resource-Constrained IoT Environments
Abstract
Abstract Homomorphic encryption (HE) enables privacy-preserving federated learning (FL) in IoT environments, but standard CKKS parameters are tuned for cloud-scale workloads, and multi-key aggregation introduces key-switching cost that grows quadratically with the number of clients. This paper proposes an adaptive resource-aware MK-CKKS framework built around a reference-domain key-switching protocol that reduces multi-key aggregation complexity from O(K²) to O(K), combined with a joint computation–communication–energy optimization model under an explicit 128-bit security constraint. Unlike prior simulation-based evaluations, all results here come from real experiments: federated training on the real MNIST and CIFAR-10 datasets under non-IID partitioning (94.98% and 46.97% mean accuracy over 3 seeds, K = 20); direct execution rather than projection of the O(K²) and O(K) key-switching operation counts for K up to 200, repeated across 3 seeds, showing measured speedups of 9.4×–100.6×; a re-verified multi-party threshold-CKKS experiment (OpenFHE) confirming correctness for K up to 32; real ciphertext-packing measurements showing 3,900×–4,060× communication reduction; and a real scalability sweep showing 39–41% communication and 66–68% latency reduction over conventional CKKS-FL. We report remaining evidence gaps transparently: the exact per-client independent-key MK-CKKS construction and physical IoT hardware validation remain future work, since no public library or physical…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.