A Security-Enhanced Federated Learning Scheme Based on Homomorphic Encryption and Secret Sharing

Cong Shen,Wei Zhang,Tanping Zhou,Lingling Zhang
DOI: https://doi.org/10.3390/math12131993
IF: 2.4
2024-06-28
Mathematics
Abstract:Although federated learning is gaining prevalence in smart sensor networks, substantial risks to data privacy and security persist. An improper application of federated learning techniques can lead to critical privacy breaches. Practical and effective privacy-enhanced federated learning (PEPFL) is a widely used federated learning framework characterized by low communication overhead and efficient encryption and decryption processes. Initially, our analysis scrutinized security vulnerabilities within the PEPFL framework and identified an effective attack strategy. This strategy enables the server to derive private keys from content uploaded by participants, achieving a 100% success rate in extracting participants' private information. Moreover, when the number of participants does not exceed 300, the attack time does not surpass 3.72 s. Secondly, this paper proposes a federated learning model that integrates homomorphic encryption and secret sharing. By using secret sharing among participants instead of secure multi-party computation, the amount of effective information available to servers is reduced, thereby effectively preventing servers from inferring participants' private gradients. Finally, the scheme was validated through experiments, and it was found to significantly reduce the inherent collusion risks unique to the federated learning scenario. Moreover, even if some participants are unavailable, the reconstructable nature of secret sharing ensures that the decryption process can continue uninterrupted, allowing the remaining users to proceed with further training. Importantly, our proposed scheme exerts a negligible impact on the accuracy of model training.
mathematics
What problem does this paper attempt to address?
### Problems the Paper Attempts to Solve This paper aims to address the issues of data privacy and security in Federated Learning (FL) within intelligent sensor networks. Specifically: 1. **Analyzing Security Vulnerabilities of the PEPFL Framework**: - The paper first analyzes the Privacy-Enhanced Federated Learning (PEPFL) framework proposed by Chen et al. and identifies specific security flaws. - An attack method is proposed, through which the server can compute and obtain all participants' private information, validating the effectiveness of the attack. 2. **Proposing an Improved Solution**: - Based on Homomorphic Encryption (HE) and Secret Sharing (SS) technologies, a new federated learning model is proposed. - By using secret sharing instead of multi-party secure computation, the amount of effective information that the server can obtain is reduced, effectively preventing the server from inferring participants' private gradients. - Experimental validation shows that the new scheme significantly reduces the risk of collusion in federated learning scenarios, and even if some participants are unavailable, the system can continue decryption and further training with minimal impact on model training accuracy. In summary, this paper is primarily dedicated to solving the security and privacy protection issues encountered in the practical application of federated learning, particularly in intelligent sensor networks. By proposing new encryption and decryption mechanisms, the security and privacy protection performance of the system is enhanced.