Skip to Main Content
The recognition of digital audio data manipulation is a challenge addressed by various fragile and content- fragile watermarking algorithms. But so far none of the approaches provides satisfying results with respect to manipulation detection. Especially distinguishing malicious attacks from allowed post production operations is still an open issue. We introduce an integration of our watermarking-based approach for audio data falsification recognition into a content-based watermark authentication method. This combines the capabilities of known feature-embedding algorithms with our forensic approach using watermarking as a hint for falsification estimation. We create a feature that can be extracted out of the same domain as for the watermark embedding. Our method is verified using several audio data sets of speech.