跳转至

通过零知识证明实现图像溯源的软修订

文章背景与核心概要

数字图像溯源标准（如 C2PA）通过附加签名记录来追踪照片的来源、编辑历史和版权信息。然而，这种透明度往往与隐私保护产生冲突，因为共享元数据可能会暴露创作者或拍摄地点的敏感上下文信息。

本文引入了图像溯源的软修订（soft redaction）概念，利用零知识证明（ZKP）对隐藏数据进行处理，以替代敏感的溯源声明。作者提出了与 C2PA 标准兼容的实用距离证明框架：位置邻近性利用 ZKP 电路中的切比雪夫多项式近似来证明图像拍摄于某个公开参考点附近，而不泄露精确坐标；生物特征相似度（\(L_2\) 距离）实现了对生物特征嵌入的隐私保护相似性声明，以强制执行肖像权；防欺诈与感知哈希将距离证明结构应用于视觉指纹，从而通过水印安全地恢复被剥离的溯源元数据。

性能测试表明，这些零知识证明可以在几秒钟内生成，并在几毫秒内完成验证，使其在现实世界的应用中具有极高的实用价值。

摘要

Summary

Digital image provenance standards (such as C2PA) attach signed records of origin, editing history, and rights to photos. However, this transparency often conflicts with privacy, as sharing metadata can expose sensitive context about creators or capture locations.

This paper introduces soft redaction for image provenance, replacing sensitive provenance claims with zero-knowledge proofs (ZKPs) over hidden data. The authors propose practical distance-proof frameworks that are compatible with C2PA standards: * Location Proximity: Using Chebyshev polynomial approximations in ZKP circuits to prove an image was captured near a public reference point without revealing exact coordinates. * Biometric Likeness (\(L_2\) Distance): Enabling privacy-preserving claims of likeness over biometric embeddings to enforce personality rights. * Anti-Spoofing & Perceptual Hashing: Applying distance-proof construction to visual fingerprints to help recover stripped provenance metadata securely via watermarks.

Performance tests show that these ZKPs can be constructed in seconds and verified in milliseconds, making them practical for real-world applications.

元数据与出版详情

Metadata & Publication Details

  • 作者: Muhammad Awan, John Collomorse
  • Authors: Muhammad Awan, John Collomorse
  • 主要学科: 密码学与安全 (cs.CR)
  • Primary Subject: Cryptography and Security (cs.CR)
  • 次要学科: 人工智能 (cs.AI)
  • Secondary Subject: Artificial Intelligence (cs.AI)
  • 提交日期: 2026年8月7日
  • Submission Date: August 7, 2026
  • 会议状态: 将在 ECCV 2026 计算机视觉中的隐私、公平、问责与透明度研讨会 (PFATCV) 上发表
  • Conference Status: To appear at ECCV 2026 workshop on Privacy Fairness Accountability and Transparency in Computer Vision (PFATCV)

链接与资源