Understanding Compressive Adversarial Privacy
arXiv:1809.08911 · doi:10.1109/CDC.2018.8619455
Abstract
Designing a data sharing mechanism without sacrificing too much privacy can be considered as a game between data holders and malicious attackers. This paper describes a compressive adversarial privacy framework that captures the trade-off between the data privacy and utility. We characterize the optimal data releasing mechanism through convex optimization when assuming that both the data holder and attacker can only modify the data using linear transformations. We then build a more realistic data releasing mechanism that can rely on a nonlinear compression model while the attacker uses a neural network. We demonstrate in a series of empirical applications that this framework, consisting of compressive adversarial privacy, can preserve sensitive information.
References in corpus (9)
- Sequence to Sequence Learning with Neural Networks
- Deep Learning with Differential Privacy
- Estimating Mutual Information
- Wasserstein GAN
- Context-Aware Generative Adversarial Privacy
- On surrogate loss functions and -divergences
- Infrastructure Quality Assessment in Africa using Satellite Imagery and Deep Learning
- Minimax Filter: Learning to Preserve Privacy from Inference Attacks
- Controller Synthesis for Linear Time-varying Systems with Adversaries
Cited by in corpus (6)
- AutoGAN-based Dimension Reduction for Privacy Preservation
- Constructing dynamic residential energy lifestyles using Latent Dirichlet Allocation
- Generative Adversarial Networks: A Survey Towards Private and Secure Applications
- Distributed generation of privacy preserving data with user customization
- Energy Resource Control via Privacy Preserving Data
- Generating private data with user customization