<title>Approaches to biometric watermarks for owner authentification</title>

Abstract

One major application domain for digital watermarks is copyright protection. Besides the design of watermarking algorithms, technologies for copyright holder identification have to be investigated. To ensure authenticity of an individual person, a wide number of biometric procedures exist. We define and describe new biometric watermarks, which denote the application of biometric reference data of individuals within digital watermarks to identify and verify ownership. Amongst the two classes of physiological and senso-motoric biometric schemes, the later appears more appropriate for biometric watermarks, as only these provide implicit expressions of intention. As such, we choose on-line handwriting as an appropriate base technology for our three new scenarios in biometric watermarking. In the first approach, embedding keys are being generated from biometric reference data, which requires stable and robust features and leads to rather complex keys. To overcome the complexity boundaries, the second approach develops a biometric reference hash, allowing key look-ups in key certifying servers. Although this proceeding leads to less complex keys, it still requires stable features. The third approach describes the embedding of biometric reference data within a watermark, allowing owner verification by more variant features, but limitations apply due to capacity of watermarking systems and also protection of the reference data is required. While most handwriting-based verification systems are limited to signature contexts, we discuss two additional context types for user authentication: passphrases and sketches.