Locally Differentially Private Distributed Deep Learning via Knowledge Distillation
This addresses privacy issues for data owners in domains like healthcare who are hesitant to share sensitive data for distributed learning, though it appears incremental as it builds on existing privacy techniques.
The paper tackles the problem of privacy concerns in distributed deep learning by proposing LDP-DL, a framework that uses local differential privacy and knowledge distillation, and it shows that LDP-DL outperforms competitors like DP-SGD, PATE, and DP-FL in privacy budget and model accuracy on datasets such as CIFAR10, MNIST, and FashionMNIST.
Deep learning often requires a large amount of data. In real-world applications, e.g., healthcare applications, the data collected by a single organization (e.g., hospital) is often limited, and the majority of massive and diverse data is often segregated across multiple organizations. As such, it motivates the researchers to conduct distributed deep learning, where the data user would like to build DL models using the data segregated across multiple different data owners. However, this could lead to severe privacy concerns due to the sensitive nature of the data, thus the data owners would be hesitant and reluctant to participate. We propose LDP-DL, a privacy-preserving distributed deep learning framework via local differential privacy and knowledge distillation, where each data owner learns a teacher model using its own (local) private dataset, and the data user learns a student model to mimic the output of the ensemble of the teacher models. In the experimental evaluation, a comprehensive comparison has been made among our proposed approach (i.e., LDP-DL), DP-SGD, PATE and DP-FL, using three popular deep learning benchmark datasets (i.e., CIFAR10, MNIST and FashionMNIST). The experimental results show that LDP-DL consistently outperforms the other competitors in terms of privacy budget and model accuracy.