Two key estimation techniques for the broken arrows watermarking scheme

Authors:
Patrick Bas;Andreas Westfeld
Affiliations:
CNRS, Grenoble, France;University of Applied Sciences, Dresden, Germany
Venue:
Proceedings of the 11th ACM workshop on Multimedia and security
Year:
2009

Citing 9
Cited 2

Principal Angles between Subspaces in an A-Based Scalar Product: Algorithms and Perturbation Estimates

SIAM Journal on Scientific Computing
A regression-based restoration technique for automated watermark removal

Proceedings of the 10th ACM workshop on Multimedia and security
Evaluation of an optimal watermark tampering attack against dirty paper trellis schemes

Proceedings of the 10th ACM workshop on Multimedia and security
Broken arrows

EURASIP Journal on Information Security
Spread-spectrum watermarking security

IEEE Transactions on Information Forensics and Security
Watermarking security: theory and practice

IEEE Transactions on Signal Processing - Part II
Kerckhoffs-Based Embedding Security Classes for WOA Data Hiding

IEEE Transactions on Information Forensics and Security
Security of Lattice-Based Data Hiding Against the Known Message Attack

IEEE Transactions on Information Forensics and Security
Applying informed coding and embedding to design a robust high-capacity watermark

IEEE Transactions on Image Processing

Towards robust and secure watermarking

Proceedings of the 12th ACM workshop on Multimedia and security
Soft-SCS: improving the security and robustness of the scalar-costa-scheme by optimal distribution matching

IH'11 Proceedings of the 13th international conference on Information hiding

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents two different key estimation attacks targeted for the image watermarking system proposed for the BOWS-2 contest. Ten thousand images are used in order to estimate the secret key and remove the watermark while minimizing the distortion. Two different techniques with distinct strategies are proposed. The first one combines a regression-based denoising process to filter out the component of the original images and a clustering algorithm to compute the different components of the key. The second attack is based on an inline subspace estimation algorithm, which estimates the subspace associated with the secret key without computing eigen decomposition. The key components are then estimated using Independent Component Analysis and a strategy designed to leave efficiently the detection region is presented. On six test images, the two attacks are able to remove the mark with very small distortions (between 41.8 dB and 49 dB).