Abstract:Trustworthy federated learning aims to achieve optimal performance while ensuring clients' privacy. Existing privacy-preserving federated learning approaches are mostly tailored for image data, lacking applications for time series data, which have many important applications, like machine health monitoring, human activity recognition, etc. Furthermore, protective noising on a time series data analytics model can significantly interfere with temporal-dependent learning, leading to a greater decline in accuracy. To address these issues, we develop a privacy-preserving federated learning algorithm for time series data. Specifically, we employ local differential privacy to extend the privacy protection trust boundary to the clients. We also incorporate shuffle techniques to achieve a privacy amplification, mitigating the accuracy decline caused by leveraging local differential privacy. Extensive experiments were conducted on five time series datasets. The evaluation results reveal that our algorithm experienced minimal accuracy loss compared to non-private federated learning in both small and large client scenarios. Under the same level of privacy protection, our algorithm demonstrated improved accuracy compared to the centralized differentially private federated learning in both scenarios.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve privacy protection while maintaining high model performance in time - series data analysis. Specifically, most of the existing privacy - protected federated learning methods are designed for image data and lack applications for time - series data. In addition, adding protective noise to time - series data analysis models will significantly interfere with the time - dependent data learning process, leading to a significant drop in accuracy. Therefore, the paper proposes a new privacy - protected federated learning algorithm (DP - TimeFL), aiming to solve these problems through local differential privacy (LDP) and data obfuscation techniques, thereby reducing the impact on model performance while protecting privacy. The main contributions of the paper include: 1. Proposing a privacy - protected federated learning framework for time - series data based on LDP, DP - TimeFL, which ensures strong privacy protection. 2. By implementing the model obfuscation technique, the amplification effect of privacy protection is achieved, and at the same time, the practicality of the proposed federated learning framework is enhanced. 3. Conducting comprehensive experiments on five time - series data sets, demonstrating the security and accuracy of this method. Through the introduction of LDP and the obfuscation mechanism, this research not only extends the trust boundary of privacy protection to the client side but also realizes privacy amplification through the obfuscation technique, reducing the performance degradation caused by the use of LDP. The experimental results show that compared with non - privacy - protected federated learning, the accuracy loss of this algorithm is extremely small; at the same level of privacy protection, compared with centralized differential privacy federated learning, the accuracy of this algorithm is improved.

Shuffled Differentially Private Federated Learning for Time Series Data Analytics

Projected Federated Averaging with Heterogeneous Differential Privacy.

Federated Transfer Learning with Differential Privacy

Efficient federated learning privacy preservation method with heterogeneous differential privacy

The Impact of Differential Privacy on Model Fairness in Federated Learning.

Privacy-Preserving Federated Learning with Differentially Private Hyperdimensional Computing

Federated f-Differential Privacy

Differentially Private Federated Learning with an Adaptive Noise Mechanism

Differentially Private Federated Learning: Servers Trustworthiness, Estimation, and Statistical Inference

Enhancing Federated Learning with Adaptive Differential Privacy and Priority-Based Aggregation

Federated learning with differential privacy via fast Fourier transform for tighter-efficient combining

Performance Analysis and Optimization in Privacy-Preserving Federated Learning

Federated Learning with Differential Privacy

Federated Extra-Trees with Privacy Preserving

Differentially Private Federated Learning: A Systematic Review

Differentially private secure multi-party computation for federated learning in financial applications

Noise-Aware Algorithm for Heterogeneous Differentially Private Federated Learning

Practical Differentially Private and Byzantine-resilient Federated Learning

Privacy, accuracy, and model fairness trade-offs in federated learning

ALI-DPFL: Differentially Private Federated Learning with Adaptive Local Iterations

Differentially Private Federated Learning with Laplacian Smoothing