XNN: Paradigm Shift in Mitigating Identity Leakage within Cloud-Enabled Deep Learning

TL;DR

XNN and XNN-d are introduced, pioneering methodologies that infuse neural network features with randomized perturbations, striking a harmonious balance between utility and privacy in the domain of cloud-based deep learning.

Abstract

In the domain of cloud-based deep learning, the imperative for external computational resources coexists with acute privacy concerns, particularly identity leakage. To address this challenge, we introduce XNN and XNN-d, pioneering methodologies that infuse neural network features with randomized perturbations, striking a harmonious balance between utility and privacy. XNN, designed for the training phase, ingeniously blends random permutation with matrix multiplication techniques to obfuscate feature maps, effectively shielding private data from potential breaches without compromising training integrity. Concurrently, XNN-d, devised for the inference phase, employs adversarial training to integrate generative adversarial noise. This technique effectively counters black-box access attacks aimed at identity extraction, while a distilled face recognition network adeptly processes the perturbed features, ensuring accurate identification. Our evaluation demonstrates XNN's effectiveness, significantly outperforming existing methods in reducing identity leakage while maintaining a high model accuracy.

Figures (8)

• Figure 1: The framework of the training stage.
• Figure 2: The framework of inference stage.
• Figure 3: The pipeline of XNN.
• Figure 4: The pipeline of XNN-d
• Figure 5: The diagram of the expectation recognition attack.