一种基于后门水印与感知加密的问答数据集版权保护方法及装置
By introducing natural language adverb watermarking and semantic vector encryption mechanisms into the question-and-answer dataset, and combining watermark-triggered detection, the usability and traceability problems of copyright protection in existing technologies are solved, and efficient data copyright protection and secure distribution are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INST OF GEOGRAPHICAL SCI & NATURAL RESOURCE RES CAS
- Filing Date
- 2025-12-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing copyright protection schemes for question-and-answer datasets struggle to strike a balance between usability, imperceptibility, and traceability. Explicit watermarks are easily identified and removed, and traditional encryption methods cannot effectively track and assign responsibility after data decryption.
A backdoor watermarking and perceptual encryption method is adopted. Natural language adverb watermarks are generated through a large language model and high-dimensional vector encryption is performed by combining a semantic vectorization model. Copyright verification is performed by combining watermark triggering and discrimination mechanisms.
It achieves hidden copyright identification of question-and-answer datasets without changing the original semantics, and has machine-detectable and black-box forensic capabilities, improving the security and controllability of data use. It is suitable for closed-source APIs and managed inference environments.
Smart Images
Figure CN121765697B_ABST