Content-based audio authentication watermarking

The protection of media data integrity is a challenge addressed by various cryptographic techniques. But none so far only few approaches provide satisfying results with respect to the manipulation detection. Especially distinguishing between allowed audio editing and malicious tampering of the media content or operations with annoying impairment of the perceptual audio data quality is still an open issue. We introduce a content-based authentication watermarking technique. The experimental evaluation of our algorithm shows that the content feature is robust against operation with imperceptible impairment of the audio data content and provides detection of malicious tampering of the perceivable audio data quality and manipulation of the data content. The problem of high watermark payload we solve with a content feature transparent watermark embedding. Our watermark information is embedded in the same domain as used for the content feature extraction without impairment of the content feature. Our method is verified using audio speech data with an overall playback time of about five hours.