{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/adversarial-examples-for-generative-models","title":"Adversarial examples for generative models","arxiv_id":"1702.06832","date":"2017-02-22","proceeding":null,"authors":["Jernej Kos","Ian Fischer","Dawn Song"],"abstract":"We explore methods of producing adversarial examples on deep generative\nmodels such as the variational autoencoder (VAE) and the VAE-GAN. Deep learning\narchitectures are known to be vulnerable to adversarial examples, but previous\nwork has focused on the application of adversarial examples to classification\ntasks. Deep generative models have recently become popular due to their ability\nto model input data distributions and generate realistic examples from those\ndistributions. We present three classes of attacks on the VAE and VAE-GAN\narchitectures and demonstrate them against networks trained on MNIST, SVHN and\nCelebA. Our first attack leverages classification-based adversaries by\nattaching a classifier to the trained encoder of the target generative model,\nwhich can then be used to indirectly manipulate the latent representation. Our\nsecond attack directly uses the VAE loss function to generate a target\nreconstruction image from the adversarial example. Our third attack moves\nbeyond relying on classification or the standard loss for the gradient and\ndirectly optimizes against differences in source and target latent\nrepresentations. We also motivate why an attacker might be interested in\ndeploying such techniques against a target generative network.","url_abs":"http://arxiv.org/abs/1702.06832v1","url_pdf":"http://arxiv.org/pdf/1702.06832v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"adversarial-examples-for-generative-models","repo_url":"https://github.com/rohban-lab/Salehi_submitted_2020","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1702.06832","atlas_url":"https://app.syntology.ai/?focus=1702.06832","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}