Privacy-preserving Federated Learning based on Multi-key Homomorphic Encryption

TL;DR

The paper addresses privacy in federated learning for IoT by introducing xMK-CKKS, a multi-key homomorphic encryption scheme with an aggregated public key and decryption shares that enables secure aggregation of locally trained updates. The approach delivers collaboration-based decryption to reveal only the aggregated model, providing resilience to collusion among up to $k<N-1$ devices and the server, while remaining practical for resource-constrained devices. The authors formalize the threat model, provide security analysis, and demonstrate through experiments on the UP-FALL dataset with 10 Jetson Nano devices that xMK-CKKS maintains accuracy close to standard FL while reducing energy consumption to about $2.4$ W and offering favorable ciphertext and computation overheads compared to Paillier-based schemes. Overall, the work enables scalable, privacy-preserving FL on IoT devices with strong collusion resistance and practical performance, highlighting future work on Byzantine robustness.

Abstract

With the advance of machine learning and the internet of things (IoT), security and privacy have become key concerns in mobile services and networks. Transferring data to a central unit violates privacy as well as protection of sensitive data while increasing bandwidth demands.Federated learning mitigates this need to transfer local data by sharing model updates only. However, data leakage still remains an issue. In this paper, we propose xMK-CKKS, a multi-key homomorphic encryption protocol to design a novel privacy-preserving federated learning scheme. In this scheme, model updates are encrypted via an aggregated public key before sharing with a server for aggregation. For decryption, collaboration between all participating devices is required. This scheme prevents privacy leakage from publicly shared information in federated learning, and is robust to collusion between $k<N-1$ participating devices and the server. Our experimental evaluation demonstrates that the scheme preserves model accuracy against traditional federated learning as well as secure federated learning with homomorphic encryption (MK-CKKS, Paillier) and reduces computational cost compared to Paillier based federated learning. The average energy consumption is 2.4 Watts, so that it is suited to IoT scenarios.

Figures (5)

• Figure 1: Different federated learning schemes with various types of encryption. We propose multi-key homomorphic encryption in which parameters are not able to decrypt other parameters encrypted weights.
• Figure 2: Privacy-preserving federated learning based on xMK-CKKS multi-key homomorphic encryption
• Figure 3: Illustration of the experimental setting and distribution strategy utilized in our case study. 10 Jetson Nano IoT devices are utilized as remote devices storing the data while a server was computing the model aggregates.
• Figure 4: Classification accuracy comparison with different number of training epochs executed locally($L$) by each device in one aggregation round
• Figure 5: Computational cost for Paillier based federated learning, MK-CKKS based federated learning, and xMK-CKKS based federated learning in different stages