Lattice based distributed threshold additive homomorphic encryption with application in federated learning

Article Computer Science, Artificial Intelligence

Privacy-preserving federated learning based on multi-key homomorphic encryption

Jing Ma et al.

Summary: With the advancement of machine learning and IoT, security and privacy have become critical concerns. This paper proposes an improved federated learning scheme, which encrypts model updates and requires collaboration among participating devices to prevent privacy leakage.

INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS (2022)

添加到收藏夹

Article Computer Science, Information Systems

Efficient Privacy-Preserving Federated Learning With Unreliable Users

Yiran Li et al.

Summary: In this study, we propose an efficient privacy-preserving federated learning (EPPFL) scheme for unreliable users. By iteratively excluding irrelevant components and using weighted aggregation, the FL model can converge rapidly and improve training efficiency. Additionally, a secure framework based on the threshold Paillier cryptosystem is established to protect user privacy information.

IEEE INTERNET OF THINGS JOURNAL (2022)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Distributed additive encryption and quantization for privacy preserving federated deep learning

Hangyu Zhu et al.

Summary: The study introduces an improved encryption-based protocol for federated learning, which collaboratively generates key pairs and utilizes threshold key sharing technique to avoid the computational burden of encrypting and decrypting the entire model. This approach significantly reduces communication costs and computational complexity without compromising performance and security.

NEUROCOMPUTING (2021)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Secure Federated Matrix Factorization

Di Chai et al.

Summary: The article introduces a secure matrix factorization framework FedMF under the federated learning setting, which enhances user privacy protection by designing a user-level distributed matrix factorization framework and using homomorphic encryption. The feasibility of FedMF is verified through testing with real movie rating dataset, while challenges for applying it in practical research are also discussed.

IEEE INTELLIGENT SYSTEMS (2021)

添加到收藏夹

Article Engineering, Electrical & Electronic