RECENT PROGRESS OF DIFFERENTIALLY PRIVATE FEDERATED LEARNING WITH THE SHUFFLE MODEL

As an improved distributedmachine learning, federated learning has achieved significant success various domains. However, to prevent data leakage and improve the security of FL, there are an increasing amount of studies on exploring how to integrate FL with other techniques. One bottleneck challenge behind it is that how to efficiently balance the privacy and the efficient of communication to achieve the optimal solution. In this paper, we conduct a survey on existing studies on differentially private FL with shuffle model, which seems the efficient way to solve the above problem. We start the survey by providing several key notation to achieve efficient exploration. Then we conduct the survey according to the role of shuffle model for solving the problem between privacy and accuracy. Furthermore, we present two types of shuffle, single shuffle and m shuffles with the statistical analysis for each one in boosting the privacy amplification of users with the same level of accuracy by reasoning the practical results of recent papers. Meanwhile, the research on exploration in shuffle model is at an early stage at present. Finally, we conclude the paper by pointing out a few future directions