| Summary: | 博士 === 輔仁大學 === 應用科學與工程研究所博士班 === 105 === There is a great demand data processing directly in encrypted domains for privacy protection in cloud computing, multiparty secure computation, and outsourcing paradigms. Furthermore, in many multimedia applications, data must be compressed for efficient transmission and storage. Therefore, if the task of compression will be performed in an untrusted environment, data encryption should be conducted before data compression. It is nontrivial to realize such an encryption-then-compression (ETC) scheme since compression exploits data redundancy, whereas encryption generally removes data correlation. In this dissertation, we propose two ETC systems for general images and electrocardiogram (ECG) signals, respectively.
In the proposed image ETC system, an original image is transformed to a prediction error map. The prediction errors are divided into distinct clusters and those having higher occurring probabilities are hidden. The residual errors are randomly permuted. Furthermore, a cluster-index map is created to retrieve the predication error for each pixel. The cluster-index map can be encrypted by using various encryption schemes, including advanced encryption standard (AES) block ciphers and chaotic permutation-diffusion encryption approach. The experimental results show that the proposed image ETC system provides high-level security and achieves comparable compression efficiency. Besides, it can be used to explore the design trade-off between the security and compression ratio for image ETC applications.
Electrocardiogram (ECG) monitoring systems are widely used in the healthcare applications. ECG signals are generally collected over long periods of time and at high resolutions so that the data storage can easily become extremely large. In the proposed ETC system for ECG data, an ECG signal is segmented based on the quasi-periodicity property and rearranged into a two-dimensional matrix. Singular value decomposition (SVD) is further applied to the modeling of a quasi-periodic process and to the compression of the ECG data. Before the compression, the matrix of ECG data is multiplied by an orthogonal key matrix, and the matrix product constitutes the encrypted ECG data. Then, the SVD technique is used to compress the encrypted data. From the experimental result, the SVD-based ETC approach does not reduce the quality of the reconstructed signals relative to unencrypted compressions. Consequently, the proposed system is demonstrated to be an effective technique for assuring data security as well as compression performance for ECG data.
|
|---|
