Design And Analysis Of Secure Spread Spectrum Watermarking

As the Internet and multimedia technologies develop by leaps and bounds,it is convenient for people to share multimedia data,such as images,videos,and audio.However,this also enables malicious people to pirate multimedia data to make a profit.Consequently,the owners of multimedia data have to suffer serious economic losses.As an effective method to protect the copyright of multimedia data,the digital watermarking technique has attracted great attention from the public over the past two decades.The digital watermarking technique embeds copyright information（i.e.watermark）into multimedia data in an imperceptible way.If the copyright is infringed,owners can extract the embedded watermark to protect their legal interests.Spread spectrum（SS）watermarking is the implementation of digital watermarking and has been deeply studied due to its good robustness and security.However,some challenging and open problems still need to be solved,including:（1）in the watermarkedonly attack（WOA）scenario,for robust SS watermarking methods,the secret key can be accurately estimated due to their poor security,and thus the watermark is read,tampered with or even removed;the more secure SS watermarking methods are less robust,and the watermark may be removed under the interference of signal processing operations and noise.（2）in the known-message attack（KMA）scenario,the security analysis of SS watermarking depends on secret key estimators.However,the estimation accuracies of existing estimators are not satisfactory,and thus it is difficult to accurately evaluate the security.To address these two problems,this thesis analyzes existing SS watermarking methods and secret key estimators.Then,two novel methods are proposed as follows:（1）In the WOA scenario,this thesis proposes the Hungarian-based truncated robust natural watermarking.For the proposed method,watermarked correlations in a given circle are moved onto the circle along its radius.Next,Hungarian optimization is utilized to decrease the embedding distortion.As a result,the robustness is significantly increased.Besides,watermarked correlations are circularly distributed,which hides the information about the secret key and thus preventing attackers from accurately estimating the secret key.Extensive theoretical analyzes and experimental results show that the proposed method improves the robustness while guaranteeing the security of the secret key.（2）In the KMA scenario,this thesis proposes the equivalent keys（EK）-based estimator to estimate the secret key.Equivalent keys form a hypercone,whose axis represents a real carrier.Analyzing the hypercone by Monte Carlo simulation,we find that uniformly-sampled equivalent keys can be used to estimate the real carrier.Therefore,the proposed estimator exploits observations to select equivalent keys.Then,an estimated carrier is obtained by adding up the sampled equivalent keys.Compared with existing estimators,the proposed estimator is more effective in terms of estimation accuracy,generality,and theoretical analysis.