The Research On The Relational Database Watermarking Algorithm Based On Numeric Attributes

As the development of the computer technology and network technology, there is more and more application of the relational database in the environment of network. However, such problems like those that the data are stolen, copied illegally, etc also increase. Therefore, we should pay more attention to the intellectual property right and information safety. In addition, it is necessary to protect the copyright of the relational data by adopting the effective mechanism.Digital watermarking technology is one of secure technologies used extensively, which solve the copyright protection problem of the digital work through embedding the information including the owner marks etc into the digital media. The traditional digital watermarking technology is hard to employ directly for the characters of relational database such as vast data, little redundancy and frequently updating etc. So in order to solve the copyright protection problem of relational database, we should concern about the characters of relational database and digital watermarking technology and study the digital watermarking technology which suits to the relational data.Basis on the research about existing database watermarking technology, two relational database watermarking algorithms which based on the second-LSB and the wavelet transformation are proposed for satisfying the characteristic and application demand of the relational database in this paper. The key idea of the second-LSB watermarking algorithm is that groups the data by constructing a complete residual system, which is based on the hash values of the primary keys, and then embeds one bit into each group by judging the second-LSB of the data. In addition, the mark is retrieved in groups by comparing the threshold, which is set according to the distribution probabilities of data, with the maximum occurrence frequency of the data in the corresponding position in extraction. The idea of relational database watermarking algorithm based on wavelet transformation is that groups the data by constructing a complete residual system firstly. Next we applies wavelet transformation to the data of each group and embeds one bit in each group by setting parity similarity or difference in a pair of high and low frequency data. When the watermark is extracted, the mark is retrieved from groups by counting the frequencies of similarities and differences of parity pairs at corresponding positions and comparing with the threshold. The results of experiments have shown that our algorithms have a relatively high accuracy, a preferable imperceptibility and a good robustness. In addition, the watermark can be extracted completely, even if the number of loss data arrives to 55%-60% of the data.