Diagnosing Vulnerability of Variational Auto-Encoders to Adversarial Attacks
arXiv:2103.06701
Abstract
In this work, we explore adversarial attacks on the Variational Autoencoders (VAE). We show how to modify data point to obtain a prescribed latent code (supervised attack) or just get a drastically different code (unsupervised attack). We examine the influence of model modifications (-VAE, NVAE) on the robustness of VAEs and suggest metrics to quantify it.
References in corpus (7)
- Auto-Encoding Variational Bayes
- Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
- NVAE: A Deep Hierarchical Variational Autoencoder
- BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling
- Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images
- Adversarial Attacks on Variational Autoencoders
- Improving VAEs' Robustness to Adversarial Attack