Towards Privacy-Preserving Visual Recognition via Adversarial Training: A Pilot Study

Abstract

This paper aims to improve privacy-preserving visual recognition, an increasingly demanded feature in smart camera applications, by formulating a unique adversarial training framework. The proposed framework explicitly learns a degradation transform for the original video inputs, in order to optimize the trade-off between target task performance and the associated privacy budgets on the degraded video. A notable challenge is that the privacy budget, often defined and measured in task-driven contexts, cannot be reliably indicated using any single model performance, because a strong protection of privacy has to sustain against any possible model that tries to hack privacy information. Such an uncommon situation has motivated us to propose two strategies, i.e., budget model restarting and ensemble, to enhance the generalization of the learned degradation on protecting privacy against unseen hacker models. Novel training strategies, evaluation protocols, and result visualization methods have been designed accordingly. Two experiments on privacy-preserving action recognition, with privacy budgets defined in various ways, manifest the compelling effectiveness of the proposed framework in simultaneously maintaining high target task (action recognition) performance while suppressing the privacy breach risk.

arXiv commentA significant extension of this paper is accepted by TPAMI-20. A conference version of this paper is accepted by ECCV-18. A shorter version of this paper is accepted by ICML-18 PiMLAI workshop

References in corpus (16)

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• Deep Learning with Differential Privacy
• Two-Stream Convolutional Networks for Action Recognition in Videos
• UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild
• De-anonymizing Social Networks
• Visualizing Deep Convolutional Neural Networks Using Natural Pre-Images
• Action Recognition using Visual Attention
• DARLA: Improving Zero-Shot Transfer in Reinforcement Learning
• Defeating Image Obfuscation with Deep Learning
• Disentangling Factors of Variation via Generative Entangling
• Minimax Filter: Learning to Preserve Privacy from Inference Attacks
• Individual Privacy vs Population Privacy: Learning to Attack Anonymization
• Privacy-Preserving Visual Learning Using Doubly Permuted Homomorphic Encryption
• Learning to Succeed while Teaching to Fail: Privacy in Closed Machine Learning Systems
• Semi-Coupled Two-Stream Fusion ConvNets for Action Recognition at Extremely Low Resolutions
• Linear Disentangled Representation Learning for Facial Actions

Cited by in corpus (5)

• Privacy in Deep Learning: A Survey
• AutoGAN-based Dimension Reduction for Privacy Preservation
• PCAL: A Privacy-preserving Intelligent Credit Risk Modeling Framework Based on Adversarial Learning
• MM-Hand: 3D-Aware Multi-Modal Guided Hand Generative Network for 3D Hand Pose Synthesis
• Delving into Robust Object Detection from Unmanned Aerial Vehicles: A Deep Nuisance Disentanglement Approach