1. Privacy-Preserving Adversarial Networks
- Author
-
Tripathy, Ardhendu, Wang, Ye, and Ishwar, Prakash
- Subjects
Computer Science - Information Theory ,Computer Science - Cryptography and Security ,Computer Science - Computer Science and Game Theory ,Computer Science - Machine Learning ,Statistics - Machine Learning ,94A15, 68T05, 62B10 - Abstract
We propose a data-driven framework for optimizing privacy-preserving data release mechanisms to attain the information-theoretically optimal tradeoff between minimizing distortion of useful data and concealing specific sensitive information. Our approach employs adversarially-trained neural networks to implement randomized mechanisms and to perform a variational approximation of mutual information privacy. We validate our Privacy-Preserving Adversarial Networks (PPAN) framework via proof-of-concept experiments on discrete and continuous synthetic data, as well as the MNIST handwritten digits dataset. For synthetic data, our model-agnostic PPAN approach achieves tradeoff points very close to the optimal tradeoffs that are analytically-derived from model knowledge. In experiments with the MNIST data, we visually demonstrate a learned tradeoff between minimizing the pixel-level distortion versus concealing the written digit., Comment: 16 pages
- Published
- 2017