Defeating Image Obfuscation with Deep Learning

Abstract

We demonstrate that modern image recognition methods based on artificial neural networks can recover hidden information from images protected by various forms of obfuscation. The obfuscation techniques considered in this paper are mosaicing (also known as pixelation), blurring (as used by YouTube), and P3, a recently proposed system for privacy-preserving photo sharing that encrypts the significant JPEG coefficients to make images unrecognizable by humans. We empirically show how to train artificial neural networks to successfully identify faces and recognize objects and handwritten digits even if the images are protected using any of the above obfuscation techniques.

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• PlaNet - Photo Geolocation with Convolutional Neural Networks
• Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not?
• Faceless Person Recognition; Privacy Implications in Social Media

Cited by in corpus (27)

• Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning
• On Low-Resolution Face Recognition in the Wild: Comparisons and New Techniques
• GRNN: Generative Regression Neural Network -- A Data Leakage Attack for Federated Learning
• When Machine Learning Meets Privacy: A Survey and Outlook
• A Study of Face Obfuscation in ImageNet
• Privacy Intelligence: A Survey on Image Privacy in Online Social Networks
• Face Recognition in Low Quality Images: A Survey
• On Generating Identifiable Virtual Faces
• DP-Image: Differential Privacy for Image Data in Feature Space
• Adversarial Image Perturbation for Privacy Protection -- A Game Theory Perspective
• Towards Privacy-Preserving Visual Recognition via Adversarial Training: A Pilot Study
• A Utility-Preserving GAN for Face Obscuration
• AnonymousNet: Natural Face De-Identification with Measurable Privacy
• Image Obfuscation for Privacy-Preserving Machine Learning
• Differentially Private Imaging via Latent Space Manipulation
• DeepBlur: A Simple and Effective Method for Natural Image Obfuscation
• Obfuscation via Information Density Estimation
• Temporally coherent video anonymization through GAN inpainting
• VideoDP: A Universal Platform for Video Analytics with Differential Privacy
• Linguistic Characteristics of Censorable Language on SinaWeibo
• Natural and Effective Obfuscation by Head Inpainting
• "Healthy surveillance": Designing a concept for privacy-preserving mask recognition AI in the age of pandemics
• Distribution Discrepancy Maximization for Image Privacy Preserving
• Deep Poisoning: Towards Robust Image Data Sharing against Visual Disclosure
• Multitask Identity-Aware Image Steganography via Minimax Optimization
• Linguistic Fingerprints of Internet Censorship: the Case of SinaWeibo
• Robustness Analysis of Face Obscuration