Key Protected Classification for Collaborative Learning

TL;DR

This paper introduces a novel key-based classification model designed for collaborative learning that inherently resists GAN-based attacks, enhancing privacy in sensitive data domains.

Contribution

It proposes a new classification approach using class-specific private keys and a training scheme that safeguards class scores against adversarial GAN attacks.

Findings

The model effectively resists GAN attacks in experiments.

High-dimensional keys improve robustness without added complexity.

Source code is publicly available for reproducibility.

Abstract

Large-scale datasets play a fundamental role in training deep learning models. However, dataset collection is difficult in domains that involve sensitive information. Collaborative learning techniques provide a privacy-preserving solution, by enabling training over a number of private datasets that are not shared by their owners. However, recently, it has been shown that the existing collaborative learning frameworks are vulnerable to an active adversary that runs a generative adversarial network (GAN) attack. In this work, we propose a novel classification model that is resilient against such attacks by design. More specifically, we introduce a key-based classification model and a principled training scheme that protects class scores by using class-specific private keys, which effectively hide the information necessary for a GAN attack. We additionally show how to utilize high dimensional keys to improve the robustness against attacks without increasing the model complexity. Our detailed experiments demonstrate the effectiveness of the proposed technique. Source code is available at https://github.com/mbsariyildiz/key-protected-classification.