2020

Image Obfuscation for Privacy-Preserving Machine Learning

Raynal, Mathilde, Achanta, Radhakrishna, Humbert, Mathias

Understand

Privacy becomes a crucial issue when outsourcing the training of machine learning (ML) models to cloud-based platforms offering machine-learning services.

While solutions based on cryptographic primitives have been developed, they incur a significant loss in accuracy or training efficiency, and require modifications to the backend architecture.
A key challenge we tackle in this paper is the design of image obfuscation schemes that provide enough privacy without significantly degrading the accuracy of the ML model and the efficiency of the training process.
In this endeavor, we address another challenge that has persisted so far: quantifying the degree of privacy provided by visual obfuscation mechanisms.

Built on

Two-dimensional signal and image processing
Jae S Lim · 1990
Earlier work this paper cites.
Digital Image Processing
Rafael Gonzalez and Richard Woods · 1992
Earlier work this paper cites.
Applied Logistic Regression, 2nd Ed. Chapter 5
DW Hosmer and S Lemeshow · 2000
Earlier work this paper cites.
Evaluating the effectiveness of pixelation and blurring on masking the identity of familiar faces
Karen Lander, Vicki Bruce, and Harry Hill · 2001
Earlier work this paper cites.
Preserving Privacy by De-identifying Facial Images
Elaine Newton, Latanya Sweeney, and Bradley Malin · 2003
Earlier work this paper cites.
Image quality assessment: From error visibility to structural similarity
Zhou Wang, Alan Conrad Bovik, Hamid Rahim Sheikh, and Eero P. Simoncelli · 2004
Earlier work this paper cites.
An introduction to roc analysis
Tom Fawcett · 2006
Earlier work this paper cites.
Model-based face de-identification
Ralph Gross, Latanya Sweeney, Jeffrey Cohn, Fernando de la Torre, and Simon Baker · 2006
Earlier work this paper cites.
Blur filtration fails to preserve privacy for home-based video conferencing
Carman Neustaedter, Saul Greenberg, and Michael Boyle · 2006
Earlier work this paper cites.
Image information and visual quality
H. R. Sheikh and A. C. Bovik · 2006
Earlier work this paper cites.
Understanding diagnostic tests 3: receiver operating characteristic curves
Anthony K Akobeng · 2007
Earlier work this paper cites.
Asirra: A captcha that exploits interest-aligned manual image categorization
Jeremy Elson, John (JD) Douceur, Jon Howell, and Jared Saul · 2007
Earlier work this paper cites.
An analysis of single-layer networks in unsupervised feature learning
Adam Coates, Andrew Ng, and Honglak Lee · 2011
Earlier work this paper cites.
Novel dataset for fine-grained image categorization
Aditya Khosla, Nityananda Jayadevaprakash, Bangpeng Yao, and Li Fei-Fei · 2011
Earlier work this paper cites.
Image quality assessment based on gradient similarity
Anmin Liu, Weisi Lin, and Manish Narwaria · 2011
Earlier work this paper cites.
A blur-robust descriptor with applications to face recognition
Raghuraman Gopalan, Sima Taheri, Pavan Turaga, and Rama Chellappa · 2012
Earlier work this paper cites.
Ml confidential: Machine learning on encrypted data
Thore Graepel, Kristin Lauter, and Michael Naehrig · 2012
Earlier work this paper cites.
Robust Image Obfuscation for Privacy Protection in Web 2.0 Applications
Andreas Poller, Martin Steinebach, and Huajian Liu · 2012
Earlier work this paper cites.
P3: Toward privacy-preserving photo sharing
Moo-Ryong Ra, Ramesh Govindan, and Antonio Ortega · 2013
Earlier work this paper cites.
Recovery of Images with Missing Pixels using a Gradient Compressive Sensing Algorithm
Isidora Stankovic · 2014
Earlier work this paper cites.
Vsi: A visual saliency-induced index for perceptual image quality assessment
Original
Lin Zhang, Ying Shen, and Hongyu Li · 2014
Earlier work this paper cites.
Hacking smart machines with smarter ones: How to extract meaningful data from machine learning classifiers
Giuseppe Ateniese, Luigi V Mancini, Angelo Spognardi, Antonio Villani, Domenico Vitali, and Giovanni Felici · 2015
Earlier work this paper cites.
Fast, Privacy Preserving Linear Regression over Distributed Datasets based on Pre-Distributed Data
Martine De Cock, Rafael Dowsley, Anderson C. A. Nascimento, and Stacey C Newman · 2015
Earlier work this paper cites.

Similar

Model inversion attacks that exploit confidence information and basic countermeasures
Matt Fredrikson, Somesh Jha, and Thomas Ristenpart · 2015
Cited alongside, same era.
Adam: A method for stochastic optimization
Diederik P. Kingma and Jimmy Ba · 2015
Cited alongside, same era.
Face Recognition across Non-Uniform Motion Blur, Illumination, and Pose
Abhijith Punnappurath, Ambasamudram Narayanan Rajagopalan, Sima Taheri, Rama Chellappa, and Guna Seetharaman · 2015
Cited alongside, same era.
Privacy-Preserving Deep Learning
Reza Shokri and Vitaly Shmatikov · 2015
Cited alongside, same era.
Deep Learning with Differential Privacy
Martín Abadi, Andy Chu, Ian Goodfellow, H. Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang · 2016
Cited alongside, same era.
On the (in)effectiveness of mosaicing and blurring as tools for document redaction
Steven Hill ans Zhimin Zhou, Lawrence Saul, and Hovav Shacham · 2016
Cited alongside, same era.
Scalable and secure logistic regression via homomorphic encryption
Yoshinori Aono, Takuya Hayashi, Le Trieu Phong, and Lihua Wang · 2016
Cited alongside, same era.
Scalable and secure logistic regression via homomorphic encryption
Yoshinori Aono, Takuya Hayashi, Le Trieu Phong, and Lihua Wang · 2016
Cited alongside, same era.
PUPPIES: Transformation-Supported Personalized Privacy Preserving Partial Image Sharing
Jianping He, Bin Liu, Deguang Kong, Xuan Bao, Na Wang, Hongxia Jin, and George Kesidis · 2016
Cited alongside, same era.
Deep Residual Learning for Image Recognition
K. He, X. Zhang, S. Ren, and J. Sun · 2016
Cited alongside, same era.
GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, and Bernhard Nessler · 2016
Cited alongside, same era.
Defeating Image Obfuscation with Deep Learning
Richard McPherson, Reza Shokri, and Vitaly Shmatikov · 2016
Cited alongside, same era.

Then

Property inference attacks on fully connected neural networks using permutation invariant representations
Karan Ganju, Qi Wang, Wei Yang, Carl A Gunter, and Nikita Borisov · 2018
Later among the works it cites.
Privacy-preserving ridge regression with only linearly-homomorphic encryption
Irene Giacomelli, Somesh Jha, Marc Joye, C David Page, and Kyonghwan Yoon · 2018
Later among the works it cites.
Data Augmentation by Pairing Samples for Images Classification
Hiroshi Inoue · 2018
Later among the works it cites.
Securelr: Secure logistic regression model via a hybrid cryptographic protocol
Yichen Jiang, Jenny Hamer, Chenghong Wang, Xiaoqian Jiang, Miran Kim, Yongsoo Song, Yuhou Xia, Noman Mohammed, Md Nazmus Sadat, and Shuang Wang · 2018
Later among the works it cites.
Logistic regression model training based on the approximate homomorphic encryption
Andrey Kim, Yongsoo Song, Miran Kim, Keewoo Lee, and Jung Hee Cheon · 2018
Later among the works it cites.
Secure logistic regression based on homomorphic encryption: Design and evaluation
Miran Kim, Yongsoo Song, Shuang Wang, Yuhou Xia, and Xiaoqian Jiang · 2018
Later among the works it cites.
SGD on Random Mixtures: Private Machine Learning under Data Breach Threats
Kangwook Lee, Kyungmin Lee, Hoon Kim, Changho Suh, and Kannan Ramchandran · 2018
Later among the works it cites.
A Haar Wavelet-Based Perceptual Similarity Index for Image Quality Assessment
Rafael Reisenhofer, Sebastian Bosse, Gitta Kutyniok, and Thomas Wiegand · 2018
Later among the works it cites.
Natural and Effective Obfuscation by Head Inpainting
Qianru Sun, Liqian Ma, Seong Joon Oh, Luc Van Gool, Bernt Schiele, and Mario Fritz · 2018
Later among the works it cites.
mixup: Beyond Empirical Risk Minimization
Hongyi Zhang, Moustapha Cisse, Yann N. Dauphin, and David Lopez-Paz · 2018
Later among the works it cites.
Differential privacy for image publication
Liyue Fan · 2019
Later among the works it cites.
Mixup Based Privacy Preserving Mixed Collaboration Learning
Yingwei Fu, Huaimin Wang, Kele Xu, Haibo Mi, and Yijie Wang · 2019
Later among the works it cites.
Logan: Membership inference attacks against generative models
Jamie Hayes, Luca Melis, George Danezis, and Emiliano De Cristofaro · 2019
Later among the works it cites.
Dp-admm: Admm-based distributed learning with differential privacy
Zonghao Huang, Rui Hu, Yuanxiong Guo, Eric Chan-Tin, and Yanmin Gong · 2019
Later among the works it cites.
Evaluating differentially private machine learning in practice
Bargav Jayaraman and David Evans · 2019
Later among the works it cites.
Synthesizing differentially private datasets using random mixing
Kangwook Lee, Hoon Kim, Kyungmin Lee, Changho Suh, and Kannan Ramchandran · 2019
Later among the works it cites.
Exploiting Unintended Feature leakage in Collaborative Learning
Luca Melis, Congzheng Song, Emilio De Cristofaro, and Vitaly Shmatikov · 2019
Later among the works it cites.
Towards deep neural network training on encrypted data
Karthik Nandakumar, Nalini Ratha, Sharath Pankanti, and Shai Halevi · 2019
Later among the works it cites.
Ml-leaks: Model and data independent membership inference attacks and defenses on machine learning models
Ahmed Salem, Yang Zhang, Mathias Humbert, Mario Fritz, and Michael Backes · 2019
Later among the works it cites.
A framework for evaluating image obfuscation under deep learning-assisted privacy attacks
J. Tekli, B. al Bouna, R. Couturier, G. Tekli, Z. al Zein, and M. Kamradt · 2019
Later among the works it cites.
Instahide: Instance-hiding schemes for private distributed learning
Yangsibo Huang, Zhao Song, Kai Li, and Sanjeev Arora · 2020
Closest in time.
pHash.org: Home of pHash, the open source perceptual hash library
Evan Klinger and David Starkweather · 2020
Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…

Reading the bibliography…