2006 Fiscal Year Final Research Report Summary

Information Embedding Ensuring the Complete Recoverability of Cover Data

Research Project

Project/Area Number	16500007
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Single-year Grants
Section	一般
Research Field	Fundamental theory of informatics
Research Institution	Gunma University
Principal Investigator	YOKOO Hidetoshi GUNMA UNIVERSITY, Department of Computer Science, Professor, 工学部, 教授 (70134153)
Project Period (FY)	2004 – 2006
Keywords	Information Theory / Information hiding / Steganography / Watermarking / Data Compression / Reversible Embedding
Research Abstract	This study throws light on the relationship between lossless data compression and lossless data embedding. Lossless data embedding is a watermarking technique that is capable of recovering the cover data after the embedded data has been extracted. We establish an information-theoretic model for a one-dimensional formalization of lossless generalized LSB-embedding methods. Based on the model, we first consider the problem of minimizing the Length of the cover data required to embed a watermark, and design a universal embedding method that can asymptotically attain the minimum cover data length. We then develop an LZ77 variation that performs neither compression nor expansion. Instead, it embeds as much extra information as possible while keeping the input file size. Asymptotic embedding capacity is then derived on the proposed variation, which shows a duality relation between data compression and lossless data embedding.