非對稱式資訊隱藏技術之研究

• Main Author: 宋嘉宏
• Other Authors: 婁德權
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/fpfy7h

博士 === 國防大學中正理工學院 === 國防科學研究所 === 91 === 　　Data hiding is a mechanism that is different from traditional cryptography techniques in the field of the information security. Data hiding systems embed information inside a host signal, which is typically an image, an audio, or a video signal. One can recover the embedded information even if the composite host signal undergoes a variety of attacks as long as the resulting corruption does not unacceptably degrade the host signal. These types of systems play an important role in embedding the secret information into meaningful and insignificant data so that people cannot directly perceive the embedded information through human senses such as sight and hearing. The advantage of data hiding, compared to traditional encryption techniques, is the innocent nature of the host signal that camouflages the protected information, and it does not arouse enemy interest in interception or attacks on the signal. 　　In this dissertation, some schemes are proposed for applications of imagery data hiding, which include covert communication, intellectual property protection (IPR), and verification of data integrity. Current research in the field of imagery data hiding shows that the trade-off between maintaining image quality and increasing robustness of resisting attacks to destroy the protected information presents a big challenge. 　　For covert communication, we introduce a framework in which to design an asymmetric data hiding technique to protect the communication between two parties based on the concept of chaos and asymmetric cryptography. In this framework, performance is characterized by a computable orbit and a chaotic mapping that provides confusion and diffusion for mixing and de-mixing information perceptually. 　　An important application of data hiding is watermarking that is used for IPR. Two major criteria to evaluate its effectiveness of watermarking are the quality of the resultant image and the robustness of the embedded information. We use chaotic mapping and discrete cosine transform to propose an asymmetric watermarking scheme, where the key that is used for embedding information is different from the one for detecting. We also propose a blind watermarking scheme, where watermark detection will not require the original image. In this proposed scheme, discrete wavelet transform is employed to extract significant coefficients of images, and the de-correlating principle is then applied to a set of twin watermarks, which are drawn from the same pseudo-random sequence, to embed them into different sub-bands of the image. Another robust watermarking scheme is also proposed that can utilize a invariant property of vector quantization to resist various image manipulations even when compound attacks are applied to the watermarked image. 　　Additionally, we propose an authentication scheme for the integrity of an image. The invariant points of an image is used here to generate the error-correction codes (ECC) that are embedded into an image so that the extracted ECC codes can be used to evaluate the degree of attack on the image and even provide the proof of malicious tampering. 　　Some experimental results will demonstrate the performance of proposed schemes towards solving the problems that result from the widespread use of digital communication network to disseminate multimedia content, including unauthorized generation of perfect copies of digital multimedia, authentication of multimedia content, tamper-detection, and secret communication.