The two main objectives of this paper are: (1) to better define the public-key (PK) watermarking problem -- in terms of properties, design requirements and usage, and (2) to propose one solution to the problem by using neural networks functions. Our survey of public key watermarking begins with the review of the state of the art. Different aspects of PK watermarking are then discussed, among which: basic robustness properties, usage of PK systems, attacks on the public and secret detectors, types of PK strategies, and strong vs weak PK watermarking systems. Accordingly, a PK system using multi-layers neural networks (NN) functions is proposed to match many PK system requirements. The approach is shortly presented for the linear case. Theoretical results are given, showing that it is possible to design PK systems approaching the detection performances of secret key watermarking-- a very unusual feature of PK systems. Experimental results are given on both simulated signals and image, confirming the predicted results and showing great resistance to JPEG compression. The paper ends with openings for new research directions.

Date Published: 1 August 2001
PDF: 10 pages
Proc. SPIE 4314, Security and Watermarking of Multimedia Contents III, (1 August 2001); doi: 10.1117/12.435410

Published in SPIE Proceedings Vol. 4314:
Security and Watermarking of Multimedia Contents III
Ping Wah Wong; Edward J. Delp III, Editor(s)