Improving VAEs' Robustness to Adversarial Attack
arXiv:1906.00230
Abstract
Variational autoencoders (VAEs) have recently been shown to be vulnerable to adversarial attacks, wherein they are fooled into reconstructing a chosen target image. However, how to defend against such attacks remains an open problem. We make significant advances in addressing this issue by introducing methods for producing adversarially robust VAEs. Namely, we first demonstrate that methods proposed to obtain disentangled latent representations produce VAEs that are more robust to these attacks. However, this robustness comes at the cost of reducing the quality of the reconstructions. We ameliorate this by applying disentangling methods to hierarchical VAEs. The resulting models produce high-fidelity autoencoders that are also adversarially robust. We confirm their capabilities on several different datasets and with current state-of-the-art VAE adversarial attacks, and also show that they increase the robustness of downstream tasks to attack.
arXiv commentMain paper of 9 pages, followed by appendix
References in corpus (15)
- Adam: A Method for Stochastic Optimization
- Auto-Encoding Variational Bayes
- Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
- Deep Convolutional Inverse Graphics Network
- Disentangling by Factorising
- Adversarial Autoencoders
- Improving Variational Inference with Inverse Autoregressive Flow
- Motivating the Rules of the Game for Adversarial Example Research
- BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling
- Structured Disentangled Representations
- Adversarial Images for Variational Autoencoders
- Practical Lossless Compression with Latent Variables using Bits Back Coding
- On Nesting Monte Carlo Estimators
- Adversarial Attacks on Variational Autoencoders
- On Implicit Regularization in -VAEs