Research

My research centers on advancing the integrity, authenticity, and trustworthiness of visual content by developing unified, generalizable frameworks for image manipulation detection and localization (IMDL) across natural and biomedical domains. I design standardized datasets, efficient state-space and attention-based architectures, and vision–language–guided generative pipelines to address long-standing challenges in reproducibility, interpretability, and cross-domain transfer.

At its core, my work reformulates IMDL as a structured reasoning problem—modeling source–target relationships, semantic consistency, and information flow between real and synthetic regions. By bridging generative modeling, forensic detection, and semantic understanding, my research enables both the creation and detection of realistic manipulations guided by language or diffusion-based synthesis. This dual perspective—combining generation and detection—lays the groundwork for standardized, explainable, and trustworthy visual forensics that generalize beyond specific datasets, manipulation types, or domains.

Research Interests

• Scientific Integrity in Biomedical Publications
• Natural Image Manipulation Detection and Localization
• AI-generated Image Forgery Detection
• Generative Model Fingerprinting
• Vision-Language Modeling
• Diffusion and State-Space Models
• Multimodal and Representation Learning
• Synthetic Dataset Generation
• Trustworthy and Explainable AI