Exploring the Vulnerability of the State-of-the-Art Content Moderation Image Classifiers Against Adversarial Attacks

Simple item page
dc.contributor.advisorNajafirad, Peyman
dc.contributor.advisorVishwamitra, Nishant
dc.contributor.authorSeong, Andrew
dc.contributor.committeeMemberFernandez, Amanda
dc.date.accessioned2024-03-26T22:49:51Z
dc.date.available2024-03-26T22:49:51Z
dc.date.issued2023
dc.description.abstractThe goal of this research is to assess and describe the vulnerabilities of deep-learning based image classifiers in the context of content moderation. While similar assessments have been made on adversarial attacks involving covering up offending parts of images in order to bypass computer vision-based content moderation, no work has been done around assessing the effectiveness of the more sophisticated adversarial attacks that does not alter the context of the images. In order to achieve this, I study the effect of various adversarial attacks in different strengths and their combinations on the classification accuracy of various state-of-the-art content moderation APIs designed to classify pornographic images employed by online social media platforms. The discovered weaknesses have been shared with respective online social media platforms to alert them to their weaknesses.
dc.description.departmentComputer Science
dc.format.extent1 electronic resource (39 pages)
dc.format.mimetypeapplication/pdf
dc.identifier.isbn9798381179521
dc.identifier.urihttps://hdl.handle.net/20.500.12588/6293
dc.languageeng
dc.subjectDeep-learning
dc.subjectOnline social media
dc.subjectComputer vision
dc.subjectAdversarial attacks
dc.subject.classificationComputer science
dc.subject.classificationWeb studies
dc.titleExploring the Vulnerability of the State-of-the-Art Content Moderation Image Classifiers Against Adversarial Attacks
dc.typeThesis
dc.type.dcmiText
thesis.degree.departmentComputer Science
thesis.degree.grantorUniversity of Texas at San Antonio
thesis.degree.levelMasters
thesis.degree.nameMaster of Science

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
aseong_thesis_120823.pdf
Size:
1.55 MB
Format:
Adobe Portable Document Format
Download

Collections

Electronic Theses and Dissertations - Open Access