Mole Recruitment: Poisoning of Image Classifiers via Selective Batch Sampling

dc.contributor.authorWisdom, Ethan
dc.contributor.authorGokhale, Tejas
dc.contributor.authorXiao, Chaowei
dc.contributor.authorYang, Yezhou
dc.date.accessioned2024-02-27T19:24:02Z
dc.date.available2024-02-27T19:24:02Z
dc.date.issued2023-03-30
dc.description.abstractIn this work, we present a data poisoning attack that confounds machine learning models without any manipulation of the image or label. This is achieved by simply leveraging the most confounding natural samples found within the training data itself, in a new form of a targeted attack coined "Mole Recruitment." We define moles as the training samples of a class that appear most similar to samples of another class, and show that simply restructuring training batches with an optimal number of moles can lead to significant degradation in the performance of the targeted class. We show the efficacy of this novel attack in an offline setting across several standard image classification datasets, and demonstrate the real-world viability of this attack in a continual learning (CL) setting. Our analysis reveals that state-of-the-art models are susceptible to Mole Recruitment, thereby exposing a previously undetected vulnerability of image classifiers. Code can be found here: http://github.com/wisdeth14/MoleRecruitment
dc.description.sponsorshipThis work was supported by NSF grants #1750082, #2101052. EW was partially supported by ASU Fulton Engineering Schools’ Dean’s fellowship.
dc.description.urihttps://arxiv.org/abs/2303.17080
dc.format.extent15 pages
dc.genrejournal articles
dc.genrepreprints
dc.identifierdoi:10.13016/m2jbvd-akcq
dc.identifier.urihttps://doi.org/10.48550/arXiv.2303.17080
dc.identifier.urihttp://hdl.handle.net/11603/31720
dc.relation.isAvailableAtThe University of Maryland, Baltimore County (UMBC)
dc.relation.ispartofUMBC Computer Science and Electrical Engineering Department Collection
dc.rightsThis item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
dc.rightsAttribution-NonCommercial-NoDerivs 4.0 Internationalen
dc.rights.urihttps://creativecommons.org/licenses/by-nc-nd/4.0/
dc.titleMole Recruitment: Poisoning of Image Classifiers via Selective Batch Sampling
dc.typeText
dcterms.creatorhttps://orcid.org/0000-0002-5593-2804

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
2303.17080.pdf
Size:
3.68 MB
Format:
Adobe Portable Document Format

License bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
license.txt
Size:
2.56 KB
Format:
Item-specific license agreed upon to submission
Description: