Audio Watermark Robustness to Desynchronization via Beat Detection

  • Authors:
  • Darko Kirovski;Hagai Attias

  • Affiliations:
  • -;-

  • Venue:
  • IH '02 Revised Papers from the 5th International Workshop on Information Hiding
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

Watermarks are hidden, imperceptible, and robust marks augmented into a host signal such as audio or video. Recent studies show that in the presence of an adversary, "blind" watermark detection within an attacked clip is an exceptionally difficult task. In this paper, we explore two technologies, beat detection and block redundant coding, to combat de-synchronization and watermark estimation as two attacks that have demonstrated superior effectiveness in preventing watermark detectors from reliably accomplishing their goal. As a result, we have achieved robustness of spread-spectrum watermarks augmented in audio clips to almost arbitrary constant time-warp, pitch-bending, and wow-and-flutter of up to 1%. The adversary can remove the watermark by subtracting an estimate of the watermark from the signal with an amplitude in excess of 6dB with respect to the host. Such an attack vector typically affects substantially the fidelity of the "pirated" recording.