A Chinese Text Watermarking Based on Statistic of Phrase Frequency

Although tamper and delete attack problem is difficult to overcome, this paper proposes a novel algorithm based on the statistical characteristics of phrase frequency to solve it. The algorithm takes watermark as information entropy and Chinese characters´ code as partial probability distribution values. As a result, the problem of sequence absence brought by delete attack is converted into sorting order changing problem. To approximate watermark information entropy, two approximation algorithms are presented, the performance of which were also analyzed. Moreover after being organized in cyclic codes, the rest probability distribution value is embedded in cover text. Not only does cyclic code with redundancy guarantee the overall data safety, but error correction reduces delete attack impact.