Modeling and analysis of content identification

Content fingerprinting provides a compact content-based representation of a multimedia document. An important application of fingerprinting is the identification of modified copies of the original media content. These modifications may be incidental changes that occur during the usage of multimedia, or intentional modifications made by an adversary to avoid detection. Currently, the effectiveness of content identification techniques is often assessed through benchmark databases. To complement these experimental performance evaluations, this paper develops a theoretical framework for analyzing content identification techniques. Beneficial aspects from decision theory and game theory are exploited to gain insights toward optimal system design and parameter selection.