Defeating Image Obfuscation with Deep Learning

2016/09/01 by Richard McPherson, Reza Shokri, McPherson, Richard +3 · 2 voices · 83 citations
Computer Science · #Advanced Steganography and Watermarking Techniques #Artificial intelligence #Artificial neural network #Computer science #Computer security #Computer vision #Deep learning #Digital Media Forensic Detection #Generative Adversarial Networks and Image Synthesis #Image (mathematics) #JPEG #Obfuscation #Pattern recognition (psychology) #cs.CR #cs.CV

paper · pdf · doi:10.48550/arxiv.1609.00408

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2016/09/01 · arxiv created 2016/09/06 · arxiv updated 2016/09/08 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

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.

Cited by

• A Common Pool of Privacy Problems: Legal and Technical Lessons from a Large-Scale Web-Scraped Machine Learning Dataset
• Privacy Blur: Quantifying Privacy and Utility for Image Data Release
• A Distributed Framework for Privacy-Enhanced Vision Transformers on the Edge
• Deep Models Under the GAN: Information Leakage from Collaborative Deep Learning
• Temporally coherent video anonymization through GAN inpainting
• A Study of Face Obfuscation in ImageNet
• Adversarial Image Perturbation for Privacy Protection -- A Game Theory Perspective
• Linguistic Fingerprints of Internet Censorship: the Case of SinaWeibo
• DeepBlur: A Simple and Effective Method for Natural Image Obfuscation
• Natural and Effective Obfuscation by Head Inpainting
• A Utility-Preserving GAN for Face Obscuration
• Distribution Discrepancy Maximization for Image Privacy Preserving
• When Machine Learning Meets Privacy: A Survey and Outlook
• Deep Poisoning: Towards Robust Image Data Sharing against Visual Disclosure
• DP-Image: Differential Privacy for Image Data in Feature Space
• Multitask Identity-Aware Image Steganography via Minimax Optimization
• AnonymousNet: Natural Face De-Identification with Measurable Privacy
• Obfuscation via Information Density Estimation
• Beyond Anonymization: Object Scrubbing for Privacy-Preserving 2D and 3D Vision Tasks
• On the Importance of Conditioning for Privacy-Preserving Data Augmentation

Discussions

• Defeating Image Obfuscation with Deep Learning [hn, 78 points, 24 comments]
• Defeating Image Obfuscation with Deep Learning [lobsters, 4 points, 0 comments]

Related

• DeepBlur: A Simple and Effective Method for Natural Image Obfuscation
• A Study of Face Obfuscation in ImageNet
• HiDDeN: Hiding Data With Deep Networks
• FaceSigns: Semi-Fragile Neural Watermarks for Media Authentication and Countering Deepfakes
• DeepObfuscation: Securing the Structure of Convolutional Neural Networks via Knowledge Distillation