STAMP-V: Steganographic Traceability for ​ AI-Generated Images with Multimodal Verification

Xinlei Guan

Co-Presenters: Boyang Li, Xinlei Guan

College: Hennings College of Science Mathematics and Technology

Major: BS.COMPSCI/CYBERS

Faculty Research Mentor: Boyang Li

Abstract:

The rapid growth of generative AI has intensified challenges in content moderation and digital forensics, particularly when benign AI-generated images are paired with harmful or misleading text. This contextual misuse undermines traditional moderation systems and complicates attribution, as synthetic images typically lack persistent metadata or device signatures. We introduce STAMP-V, a steganography-enabled provenance framework that embeds cryptographically signed identifiers into images at creation time and verifies provenance through multimodal harmful content detection. Our system evaluates five watermarking methods across spatial, frequency, and wavelet domains, and integrates a CLIP-based fusion model that performs multimodal harmful-content detection as part of the provenance pipeline. Experiments demonstrate that spread-spectrum watermarking, especially in the wavelet domain, provides strong robustness under blur distortions, and our multimodal fusion detector achieves an AUC-ROC of 0.99, enabling reliable cross-modal provenance verification. These components form an end-to-end forensic pipeline that enables reliable tracing of harmful deployments of AI-generated imagery, supporting accountability in modern synthetic media environments.