Publication by Kanta MATSUURA

K. Matsuura. ``Effects of Data Hiding on Remote Data Analysis''. IPSJ Journal, Vol.42, No.10, pp.2526-2532, Oct. 2001.: (Abstract) When digital data are distributed over an open network, they do not always keep their original bit-strings: for example, issues related to copyright protection may necessitate the use of digitally watermarked images. When we want to use a watermarking system, there is a requirement that the data be kept acceptable. Usually, the difference between the original data and the watermarked data is evaluated by human recognition; for instance, if users cannot notice the difference visually, the watermark is considered acceptable.; We here suggest another scenario: if the user is not a human being but a remote computer which analyzes the digital data, the acceptance criteria might be different. This paper studies how data hiding affects remote data analysis. Specifically, a design competition among remote computers is simulated. The result suggests the importance of how the design problem is represented; smaller condition numbers of the coefficient matrix provide better robustness of the competition when the signal-to-noise ratio (SNR) is sufficiently high. This effect is demonstrated by means of a phantom-experiment study.; (Keywords) digital watermarking, watermark evaluation, entrusted inverse analysis.