Blockchain-Based Practical and Privacy-Preserving Federated Learning with Verifiable Fairness

Article Computer Science, Information Systems

Differential Privacy for Deep and Federated Learning: A Survey

Ahmed El Ouadrhiri et al.

Summary: This article introduces the main ideas of differential privacy in protecting user privacy during the deep learning process, including the use of noise to protect the original data and learning parameters. The study also discusses various types of probability distributions that satisfy the differential privacy mechanism and provides an overview of different variants of differential privacy.

IEEE ACCESS (2022)

Add to Collection

Article Automation & Control Systems

Blockchain-Enabled Federated Learning Data Protection Aggregation Scheme With Differential Privacy and Homomorphic Encryption in IIoT

Bin Jia et al.

Summary: This article presents a blockchain-enabled federated learning application model and data protection methods based on differential privacy and homomorphic encryption. Experimental results demonstrate that the method has better performance in data sharing and model sharing.

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS (2022)

Add to Collection

Proceedings Paper Computer Science, Information Systems

Advancing Blockchain-based Federated Learning through Verifiable Off-chain Computations

Jonathan Heiss et al.

Summary: This paper introduces the use of zero-knowledge proofs as an alternative to incentive and penalty mechanisms in blockchain-based federated learning. By utilizing off-chain computations and verification on the blockchain, the correctness of the learning processes can be ensured.

2022 IEEE INTERNATIONAL CONFERENCE ON BLOCKCHAIN (BLOCKCHAIN 2022) (2022)

Add to Collection

Article Engineering, Electrical & Electronic

Pain-FL: Personalized Privacy-Preserving Incentive for Federated Learning

Peng Sun et al.

Summary: This paper proposes a personalized privacy-preserving incentive mechanism called Pain-FL for FL, which provides customized payments for workers with different privacy preferences to compensate for privacy leakage costs while ensuring satisfactory convergence performance of FL models. Each worker agrees on a customized contract with the server, specifying a privacy-preserving level and payment, and perturbs gradients to be uploaded with that level in exchange for payment. Analytically derived optimal contracts under complete and incomplete information models optimize convergence performance of FL models while maintaining economic properties. Experimentally evaluated, Pain-FL demonstrates practicality and effectiveness.

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS (2021)

Add to Collection

Article Computer Science, Information Systems

PFLM: Privacy-preserving federated learning with membership proof

Changsong Jiang et al.

Summary: Privacy-preserving federated learning achieves privacy protection for user data through membership proof and the PFLM scheme, which eliminates the stringent requirements for thresholds in the original schemes. The new scheme utilizes cryptographic accumulators and a result verification algorithm based on ElGamal encryption to enhance security and correctness verification.

INFORMATION SCIENCES (2021)

Add to Collection

Article Computer Science, Information Systems

Privacy-Preserving Federated Learning Framework Based on Chained Secure Multiparty Computing

Yong Li et al.

Summary: The proposed privacy-preserving FL framework chain-PPFL utilizes single-masking and chained-communication mechanisms to achieve practical privacy preservation with some communication cost, while maintaining the accuracy and convergence speed of the training model. Extensive experiments demonstrate the effectiveness of the scheme in protecting sensitive information of participants in FL.

IEEE INTERNET OF THINGS JOURNAL (2021)

Add to Collection

Article Engineering, Electrical & Electronic