{"id":{"repo_id":"msu","oai_identifier":"oai:d.lib.msu.edu:etd_52970"},"canonical_url":"https://search.ndltd.org/etd/msu/oai:d.lib.msu.edu:etd_52970","repository":{"repo_id":"msu","name":"Michigan State University","base_url":"https://d.lib.msu.edu/oai"},"display":{"title":"ESTABLISHING SECURE AND INTERPRETABLE DEFENSE SYSTEMS TOWARDS DIGITAL FORGERY AND FACE PRESENTATION ATTACKS","abstract":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_html":"Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","abstract_has_math":false,"creators":["Guo, Xiao"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Liu, Xiaoming","Liu, Sijia","Morris, Daniel","Ross, Arun"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T18:59:34Z","subjects":["Computer science"],"languages":["eng"],"rights":["Attribution 4.0 International"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52970","local:Guo_grad.msu_0128D_21919"],"render_values":[{"text":"etd:52970","href":null,"code":true},{"text":"local:Guo_grad.msu_0128D_21919","href":null,"code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/doi:10.25335/atgt-4f84","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/doi:10.25335/atgt-4f84","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_52970","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Liu, Xiaoming","Liu, Sijia","Morris, Daniel","Ross, Arun"]},{"key":"dc:creator","label":"Author","values":["Guo, Xiao"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:relation","label":"Dc Relation","values":["Electronic Theses & Dissertations"]},{"key":"dc:type","label":"Dc Type","values":["text","Electronic dissertations"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer science"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["eng"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution 4.0 International"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["etd:52970","local:Guo_grad.msu_0128D_21919","https://doi.org/doi:10.25335/atgt-4f84"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","With the rapid growth of generative models such as GANs and diffusion models, AI-generated or edited images have become easy to create and hard to identify, and they can spread misleading or biased content. At the same time, face recognition systems are increasingly threatened by physical spoofing, including printed photos, screen replays, and 3D masks. These risks make reliable and interpretable detection methods essential for protecting digital media integrity and biometric authentication. To this end, this thesis presents a series of works that improve robustness, interpretability, and generalization in detection systems. First, we propose HiFi-Net, a unified framework for image forgery detection and localization, which generalizes across diverse manipulation types while precisely identifying manipulated regions. Second, we introduce a multi-modal approach (M2F2-Det) that combines CLIP and large language models to produce clear textual explanations, improving transparency in detecting subtle facial forgeries. Third, we develop HuForDet, a holistic framework that bridges facial artifact analysis and full-body semantic reasoning to detect modern human image forgeries. Fourth, we propose Deepfake-Agent, a unified agentic framework that adaptively coordinates multiple MLLM-based forensic experts to comprehensively fuse low-level visual artifacts with high-level semantic reasoning, which dynamically resolves conflicting evidence through reinforcement learning-based aggregation. Fifth, we generalize model parsing beyond GAN-based synthesis. We propose the LGPN method to infer generative model hyperparameters from synthetic images, enabling robust attribution across both GAN and diffusion model families. Finally, we address domain shift in face anti-spoofing by proposing a source-free domain adaptation method (FAS-wrapper). This approach updates a pre-trained model using only unlabeled target-domain data, maintaining high performance in both the source and target domains. Together, these contributions advance practical forensic and anti-spoofing defenses against the evolving spectrum of digital and physical attacks.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf","1 online resource (129 p.) : ill."]},{"key":"dc:title","label":"Title","values":["ESTABLISHING SECURE AND INTERPRETABLE DEFENSE SYSTEMS TOWARDS DIGITAL FORGERY AND FACE PRESENTATION ATTACKS"]}]}],"canonical_facts":{"dc:contributor":["Liu, Xiaoming","Liu, Sijia","Morris, Daniel","Ross, Arun"],"dc:creator":["Guo, Xiao"],"dc:date":["2026"],"dc:description":["Thesis (Ph.D.)--Michigan State University. Computer Science - Doctor of Philosophy, 2026","With the rapid growth of generative models such as GANs and diffusion models, AI-generated or edited images have become easy to create and hard to identify, and they can spread misleading or biased content. At the same time, face recognition systems are increasingly threatened by physical spoofing, including printed photos, screen replays, and 3D masks. These risks make reliable and interpretable detection methods essential for protecting digital media integrity and biometric authentication. To this end, this thesis presents a series of works that improve robustness, interpretability, and generalization in detection systems. First, we propose HiFi-Net, a unified framework for image forgery detection and localization, which generalizes across diverse manipulation types while precisely identifying manipulated regions. Second, we introduce a multi-modal approach (M2F2-Det) that combines CLIP and large language models to produce clear textual explanations, improving transparency in detecting subtle facial forgeries. Third, we develop HuForDet, a holistic framework that bridges facial artifact analysis and full-body semantic reasoning to detect modern human image forgeries. Fourth, we propose Deepfake-Agent, a unified agentic framework that adaptively coordinates multiple MLLM-based forensic experts to comprehensively fuse low-level visual artifacts with high-level semantic reasoning, which dynamically resolves conflicting evidence through reinforcement learning-based aggregation. Fifth, we generalize model parsing beyond GAN-based synthesis. We propose the LGPN method to infer generative model hyperparameters from synthetic images, enabling robust attribution across both GAN and diffusion model families. Finally, we address domain shift in face anti-spoofing by proposing a source-free domain adaptation method (FAS-wrapper). This approach updates a pre-trained model using only unlabeled target-domain data, maintaining high performance in both the source and target domains. Together, these contributions advance practical forensic and anti-spoofing defenses against the evolving spectrum of digital and physical attacks.","Description based on online resource. Title from PDF t.p. (Michigan State University Fedora Repository, viewed ).","Includes bibliographical references."],"dc:format":["application/pdf","1 online resource (129 p.) : ill."],"dc:identifier":["etd:52970","local:Guo_grad.msu_0128D_21919","https://doi.org/doi:10.25335/atgt-4f84"],"dc:language":["eng"],"dc:relation":["Electronic Theses & Dissertations"],"dc:rights":["Attribution 4.0 International"],"dc:subject":["Computer science"],"dc:title":["ESTABLISHING SECURE AND INTERPRETABLE DEFENSE SYSTEMS TOWARDS DIGITAL FORGERY AND FACE PRESENTATION ATTACKS"],"dc:type":["text","Electronic dissertations"]},"updated_at":"2026-11-18T18:59:34Z"}