Application-oriented audio watermark benchmark service

Methodologies and tools for watermark evaluation and benchmarking facilitate the development of improved watermarking techniques. In this paper, we want to introduce and discuss the integration of audio watermark evaluation methods into the well-known web service Watermark Evaluation Testbed (WET) [1]. WET is enhanced by using. A special set of audio files with characterized content and a collection of single attacks as well as attack profiles [3] will help to select special audio files and attacks with their attack parameters. Motivation Digital watermarking has been proposed for a variety of applications, including content protection, authentication, and digital rights management. Many watermarking techniques have made claims regarding performance, such as invisibility, robustness, and capacity but currently it is not easy to objectively evaluate their performance claims. Watermark evaluation methodologies and tools allow an objective comparison of performance claims and facilitate the development of improved watermarking techniques. The performance of watermarking techniques may also be compared with the specific requirements of applications. It is clear that the development of digital watermarks is connected with their evaluation. The performance claimed by watermarking algorithms or their implementations should be verified using a fair and objective process. This process can be very complex and therefore the idea is to describe the evaluation process with special attacks and their specified or selected attack parameters or with a set of signal processing functions, which could be introduced to the content (for example, lossy compression). To improve the evaluation of digital watermarking algorithms the Watermark Evaluation Testbed (WET) system was developed at Purdue University [1][4] featuring a web-based user interface for the evaluation of still-image watermarks. In this paper, we will discuss and describe the procedure to integrate audio evaluation or benchmarking function into the existing WET service and the definition of complex attack scenarios, which are called profile attacks. The paper is structured as followed: Section 1 is a review of the existing WET benchmark for image watermarking. Section 2 is a review of the StirMark for audio and a description of audio benchmark profiles. The major portion of this section is a detailed description of the profiles, their parameters, and organization. Section 3 describes a proposed method for integrating the profiles into WET for Audio using an XML framework. Lastly, Section 4 has the conclusions and directions for future work. 1 The WET evaluation service In this section we introduce the Watermark Evaluation Testbed (WET) [4], which is a web-based watermark benchmark platform. The online service provides the user the possibility to embed, attack and retrieve information by using selected watermark and attack algorithms for still images. 1 The system is located at http://www.datahiding.org for images and http://audio-wet.ecn.purdue.edu or http://audio-wet.cs.unimagdeburg.de for audio Security, Steganography, and Watermarking of Multimedia