AnchorMark: Real-World Anchor-Based Watermarking for Digital Content Authentication and Manipulation Detection

Document Type

Conference Proceeding

Publication Date

1-1-2026

Abstract

With the rapid advancement in artificial intelligence (AI) and deepfake techniques, determining the authenticity and provenance of digital content in cyberspace has become increasingly complex. However, existing forensics solutions are limited to identifying artifacts left by deepfake manipulations or tracing the source of digital content generated by specific AI models. Thus, they struggle to ensure the authenticity and traceability of digital content without a ground-truth reference. In this work, we propose AnchorMark, a real-world anchor-based watermarking framework for digital content authentication and manipulation detection. A binary AnchorMark meta-structure (AMM) is designed to record verifiable locators that link to physical Electrical Network Frequency (ENF) signals as the ground truth reference. In addition, we introduce GaWRI, a Generative adversarial network (GAN) based Watermarking framework with the Robustness and Imperceptibility for embedding AMM as multi-bit watermarks in multimedia (image or video). Therefore, investigators can extract AMM from watermarked content and use it to verify authenticity and provenance using a trusted physical anchor fabric. The experimental results demonstrate AnchorMark’s superior performance in terms of effectiveness and robustness in ensuring image authentication and manipulation detection under various attacks and baselines.

Publication Title

Lecture Notes of the Institute for Computer Sciences Social Informatics and Telecommunications Engineering Lnicst

ISBN

[9783032225443]

Link to Full Text

Share

COinS