DeepAI
Log In Sign Up

Bayesian Framework for Gradient Leakage

11/08/2021
by   Mislav Balunović, et al.
0

Federated learning is an established method for training machine learning models without sharing training data. However, recent work has shown that it cannot guarantee data privacy as shared gradients can still leak sensitive information. To formalize the problem of gradient leakage, we propose a theoretical framework that enables, for the first time, analysis of the Bayes optimal adversary phrased as an optimization problem. We demonstrate that existing leakage attacks can be seen as approximations of this optimal adversary with different assumptions on the probability distributions of the input data and gradients. Our experiments confirm the effectiveness of the Bayes optimal adversary when it has knowledge of the underlying distribution. Further, our experimental evaluation shows that several existing heuristic defenses are not effective against stronger attacks, especially early in the training process. Thus, our findings indicate that the construction of more effective defenses and their evaluation remains an open problem.

READ FULL TEXT

page 8

page 12

page 13

12/05/2022

Refiner: Data Refining against Gradient Leakage Attacks in Federated Learning

Federated Learning (FL) is pervasive in privacy-focused IoT environments...
10/12/2019

Quantification of the Leakage in Federated Learning

With the growing emphasis on users' privacy, federated learning has beco...
05/19/2021

User Label Leakage from Gradients in Federated Learning

Federated learning enables multiple users to build a joint model by shar...
11/19/2021

Understanding Training-Data Leakage from Gradients in Neural Networks for Image Classification

Federated learning of deep learning models for supervised tasks, e.g. im...
03/11/2021

TAG: Transformer Attack from Gradient

Although federated learning has increasingly gained attention in terms o...
08/27/2018

Data Poisoning Attacks against Online Learning

We consider data poisoning attacks, a class of adversarial attacks on ma...
07/27/2019

ActShare: Sensitive Data Sharing with Reliable Leaker Identification

Data sharing among multiple parties becomes increasingly common today, s...