Objective evaluation of sound quality for attacks on robust audio watermarking

Abstract

Various attacks on robust audio watermarking have been proposed. Reversible signal processing attacks, such as sampling frequency conversion, degrade sound quality of the distributed watermarked audio (stego audio) and disturb extraction of hidden data so that copyright detection systems using automated crawling are invalidated. Reversible signal processing of the attack can recover sound quality of the degraded audio data. In order to prove validity and security of audio watermarking system, analysis of the presumed attacks or reversible signal processing on stego audio, is required. However, these attacks on audio signal also degrade sound quality of commercial music where such pieces of music are considered to be not suitable for appreciation. Therefore, degradation of sound quality induced by various attacks should be taken into account to decide if the intensity of the attacks are realistic. In this study, objective audio quality measurement (PEAQ) was applied to the audio signals including typical perceptual coding, MP3, tandem MP3, MPEG4AAC, and reversible signal processing of sampling frequency conversion, noise addition, frequency shift, bandpass filtering, and echo addition. The results indicate requirements for robustness and criteria of the attacks on high quality and robust audio watermarking technology.