Core Concepts
데이터 워터마킹을 위한 AMUSE의 적응형 다중 세그먼트 인코딩 방법은 원본 워터마크를 더 짧은 하위 메시지 세트로 매핑하고 반대로 원래 메시지를 재구성합니다.
Stats
"Applying AMUSE improved the quality and the extraction accuracy (even with attacks such as subset attack) of the watermarked dataset."
"The length of the obtained sub-messages are 16-bit and 9-bit for ˆτ = 60%, and ˆτ = 80%, respectively."
"AMUSE improves both the extraction accuracy and the quality of the watermarked dataset."
Quotes
"Our encoder adaptively adjusts the length of the sub-messages to meet the target dataset protection requirements."
"AMUSE is applicable to other modalities such as text and video."
"AMUSE improves both the extraction accuracy and the quality of the watermarked dataset."