Semi-blind Fingerprinting Utilizing Ordinary Existing Watermarking Techniques

Abstract

We propose a concept of feasible blind fingerprinting called “semi-blind fingerprinting” which provides as secure content trading as conventional blind fingerprinting methods. In a basic model of conventional fingerprinting for content trading, user’s security could be guaranteed only under the premise that a content provider is trustworthy. Such premise makes a scheme unpractical. To overcome this defect, various blind methods have been proposed in which cryptography is used in order to protect the information on a user. However, computation cost and robustness have not been well considered as a scheme should be feasible, though it provides high-level of protection. A semi-blind scheme fulfills a need for both feasibility and robustness. Under this scheme, feasibility can be achieved by applying newly proposed functions called “Image Division” instead of using cryptography. In order to gain robustness, a protocol for watermarking has been so designed that any well-developed existing watermarking technique can be applied with no restriction. We will not only propose the idea but also implement and evaluate several image division functions as well as an optimized embedding program in order to prove validity of the proposed method.