{"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-images-for-variational","title":"Adversarial Images for Variational Autoencoders","arxiv_id":"1612.00155","date":"2016-12-01","proceeding":null,"authors":["Pedro Tabacof","Julia Tavares","Eduardo Valle"],"abstract":"We investigate adversarial attacks for autoencoders. We propose a procedure\nthat distorts the input image to mislead the autoencoder in reconstructing a\ncompletely different target image. We attack the internal latent\nrepresentations, attempting to make the adversarial input produce an internal\nrepresentation as similar as possible as the target's. We find that\nautoencoders are much more robust to the attack than classifiers: while some\nexamples have tolerably small input distortion, and reasonable similarity to\nthe target image, there is a quasi-linear trade-off between those aims. We\nreport results on MNIST and SVHN datasets, and also test regular deterministic\nautoencoders, reaching similar conclusions in all cases. Finally, we show that\nthe usual adversarial attack for classifiers, while being much easier, also\npresents a direct proportion between distortion on the input, and misdirection\non the output. That proportionality however is hidden by the normalization of\nthe output, which maps a linear layer into non-linear probabilities.","url_abs":"http://arxiv.org/abs/1612.00155v1","url_pdf":"http://arxiv.org/pdf/1612.00155v1.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-images-for-variational","repo_url":"https://github.com/tabacof/adv_vae","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"adversarial-attack","task_name":"Adversarial Attack"}],"methods":[{"method_slug":"linear-layer","method_name":"Linear Layer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1612.00155","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}