InnoGuard: secure authorship verification using ECC-encoded watermarks and randomized embedding for AIGC-resilient images

The rise of AI-generated content (AIGC) presents urgent challenges in image authenticity, copyright protection, and the detection of manipulation. To address these challenges, we propose InnoGuard, a robust Computer Vision framework built on EditGuard, which imperceptibly embeds copyright data, perceptual hash, and flexible, high-capacity metadata into images. It leverages SHA-256, pHash, error correction codes, and a pseudo-random walk embedding strategy to ensure authenticity, integrity, and resilience against AIGC-driven tampering and degradation. Extensive experiments demonstrate high recovery accuracy under common distortions (Gaussian noise, JPEG compression, Poisson noise) and AI regeneration watermarking attacks, while preserving high fidelity, making InnoGuard a practical solution for securing visual media in the era of generative AI.