The recent advancements in federated learning (FL) and differential privacy (DP) have significantly shaped the direction of privacy-preserving machine learning. FL continues to evolve with innovations aimed at addressing the challenges of data heterogeneity, computational efficiency, and fairness. Notably, the integration of low-rank adaptation (LoRA) with FL has shown promise in reducing computational burdens on resource-limited devices while maintaining privacy. Additionally, novel frameworks like Wasserstein Fair Federated Learning (WassFFed) and EPIC have advanced the field by ensuring fairness and privacy in distributed settings. These developments underscore the shift towards more decentralized and privacy-aware optimization frameworks, which are crucial for sensitive applications such as healthcare and finance. Furthermore, the exploration of DP in online learning and class imbalance scenarios has provided deeper theoretical insights and practical solutions, enhancing the robustness of privacy-preserving algorithms. The noteworthy papers in this area include EPIC, which demonstrates a universal decentralized optimization framework, and WassFFed, which effectively balances accuracy and fairness in federated settings.
Decentralized Privacy-Preserving Machine Learning Innovations
Sources
EPIC: Enhancing Privacy through Iterative Collaboration
DWFL: Enhancing Federated Learning through Dynamic Weighted Averaging
The Limits of Differential Privacy in Online Learning
Differential Privacy Under Class Imbalance: Methods and Empirical Insights
Client Contribution Normalization for Enhanced Federated Learning
Federated Split Learning for Human Activity Recognition with Differential Privacy
Federated LLMs Fine-tuned with Adaptive Importance-Aware LoRA
WassFFed: Wasserstein Fair Federated Learning
Differentially-Private Collaborative Online Personalized Mean Estimation
Collaborative and Federated Black-box Optimization: A Bayesian Optimization Perspective
Federated Low-Rank Adaptation with Differential Privacy over Wireless Networks
Dual-Criterion Model Aggregation in Federated Learning: Balancing Data Quantity and Quality
A Stochastic Optimization Framework for Private and Fair Learning From Decentralized Data
Federated Learning for Discrete Optimal Transport with Large Population under Incomplete Information
Efficient Federated Finetuning of Tiny Transformers with Resource-Constrained Devices
FedSub: Introducing class-aware Subnetworks Fusion to Enhance Personalized Federated Learning in Ubiquitous Systems
Locally Private Sampling with Public Data
SAFELOC: Overcoming Data Poisoning Attacks in Heterogeneous Federated Machine Learning for Indoor Localization
Laplace Transform Interpretation of Differential Privacy
SAFES: Sequential Privacy and Fairness Enhancing Data Synthesis for Responsible AI
Faster Differentially Private Top-$k$ Selection: A Joint Exponential Mechanism with Pruning