Iterative compensation schemes for multimedia content authentication

摘要

Multimedia content authentication can be achieved by deriving a fragile content-based signature from a media signal and embedding it robustly back into the signal. The perturbation due to embedding needs to be compensated with mechanisms that lead to producing the original signature. Closed form solutions for compensation are not always feasible. In this paper, we propose two iterative compensation schemes to address such scenarios. The first is based on a set-theoretic approach by using projections onto convex sets (POCS), where signature generation and signal fidelity are modeled as convex constraint sets. The other scheme uses an additive iterative approach used in the cases where the POCS approach is not applicable. Examples of successful image authentication are presented to demonstrate the effectiveness of the proposed schemes over a range of watermark-embedding robustness. A comparison with the closed-form compensation scheme, when feasible, is also presented.