Abstract A novel framework is proposed in this paper for detection of seal and signature entities with hierarchical recovery capability based on self-embedding watermarking. First of all, the entities of importance called authentication entities are identified in the documents. Once these entities are located, they are made secure via integrity check bits. The authentication is performed at the pixel level through three integrity check bits based on pixel value, location, and neighborhood. The recovery information from these authentication entities is extracted and embedded after encryption in multiple locations throughout the document. In case of tampering, the tampered regions are detected as well as localized and reverse mapping is performed to fetch the recovery information for reconstruction of the tampered regions. To validate the efficacy of the proposed framework, a number of experiments have been conducted with multiple documents and different attack scenarios. The imperceptibility of the watermarked and recovered documents has been evaluated using Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM) metrics.

中文翻译：

---

篡改数字文档中基于水印自嵌入的分层恢复的印章和签名实体检测

摘要 本文提出了一种基于自嵌入水印的具有分层恢复能力的印章和签名实体检测框架。首先，在文档中标识称为认证实体的重要实体。一旦找到这些实体，就通过完整性检查位使它们变得安全。基于像素值、位置和邻域，通过三个完整性校验位在像素级别执行身份验证。来自这些认证实体的恢复信息在加密后被提取并嵌入到整个文档的多个位置。在篡改的情况下，篡改区域被检测以及局部化并执行反向映射以获取用于重构篡改区域的恢复信息。为了验证所提出框架的有效性，已经对多个文档和不同的攻击场景进行了大量实验。已使用峰值信噪比 (PSNR) 和结构相似性指数 (SSIM) 指标评估了带水印和恢复文档的不可察觉性。