Robust Watermarking of Audio with Blind Self-Authentication O. FAROOQ, S. DATTA, AND J. BLACKLEDGE Electronic and Electrical Engineering, Loughborough University, Loughborough Leicestershire, LE11 3TU UK Abstract—In this paper robust audio watermarking based on chirp signal is proposed which also includes capability of blind self-authentication. Watermark sequence is extracted from the original audio signal based on sub-bands energy obtained by using wavelet decomposition. The watermarking scheme is such that the watermark sequence remains unchanged for the original and watermarked audio. This enables two independent extraction processes for this sequence. A total match in the sequence extracted by the two procedures ensures the authenticity of audio data and any mismatch indicates tampering. Different simulated attacks were carried out on the watermarked audio and in all the cases tampering was detected by the proposed technique. The test audio signals were selected from the Speech Quality Assessment Material (SQAM) for embedding watermark. The scheme was found to be robust for attacks such as up-sampling, down-sampling, scaling, filtering and additive noise. Key-words— Audio Watermarking Chirps, Robust Watermarking, Self-Authentication, Wavelet Transform. 1 Introduction During the recent years, growth in digital multimedia technologies has made users to switch from analog world to digital. The use of internet and wireless application has made it easy and fast to transmit data and with the availability of powerful computing, digital copying and tampering of data can be easily performed. From the viewpoint of media producers and content providers, the possibility for unlimited copying of digital data without loss of fidelity is undesirable because it may cause considerable financial loss and also violate copyright [1]. Editing/tampering the content may give misleading/wrong information. Therefore, the protection and enforcement of intellectual property rights for digital media has become an important issue. Cryptography has been used to guarantee secure transfer of data from point-to-point. However, once the data has been decrypted by the valid recipient it can be subsequently re-distributed in its original form without the knowledge of the actual owner. One solution to ownership protection is digital watermarking, which is a technique of embedding a low energy signal (watermark) in the original multimedia data (host i.e. audio, image, and video) for its protection such that the perceptual quality does not degrade. This requires the watermark to be robust and should not easily be removed from the data/media unless there is an appreciable degradation in the perceptual quality. The removal of watermark (or attack on watermark) is achieved by using techniques such as lossy compression, scaling, signal cropping, re-sampling, re-quantization, etc of the media. To authenticate originality of data, another class of watermarking is carried out known as fragile watermarking. This is contrary to robust watermarking, and the watermark breaks as soon as any processing is applied to the watermarked signal. Thus, these two variations of watermarking technique depend on the application where it has to be used. Robust watermarks are generally used for copyright and ownership verification. In comparison, fragile watermarks are useful for purposes of authentication and integrity attestation. Developing imperceptible audio watermark is difficult as compared to image watermark because of high sensitivity of the human ear. The audio watermarking techniques exploit the fact that the human auditory system is insensitive to small amplitude changes, either in time [1, 2], frequency [3, 4, 5] domain, 7th WSEAS Int. Conf. on Electronics, Hardware, Wireless and Optical Communications, Cambridge, UK, February 20-22, 2008 ISSN: 1790-5117 Page 35 ISBN: 978-960-6766-40-4