Research On Attack And Resistibility Of Digital Watermarking For Relational Databases

The digital watermarking technique is an important branch of information security technique, and also is one of the most efficient technical means protecting the digital works' copyrights. Along with the development of database technique and wide application of database management system, the issue of data security in databases captures more recognition, and providing copyright protection for databases has been one of the important issues in the area of database security. Using digital watermarking technique to protect databases'copyrights attracts researchers'interest, so the relational database digital watermarking technique appears. With the development of relational database digital watermarking technique, there are many researches on watermarking attacks, and how to design algorithms resistive to attacks is one of the research emphases. Only numeric relational databases are studied. Based on analysis of existing watermarking algorithms we research on current watermarking attacks and present efficiently resistive algorithm to the attacks yet not resolved.Firstly, an algorithm efficiently resistive to reversibility attack is proposed in this thesis. For the sake of resolving the reversibility attack, characteristically computing the original relational database is designed to construct characteristic codes, and the characteristic codes based on database content will be embedded into a database as a watermark. Elaborate method of characteristically computing is given and distinct watermarking embedding method different from former methods is presented in this thesis. Experimental results show that the algorithm is efficiently resistive to reversibility attack and is robust to subset attacks. In the end, the algorithm's deficiency is analyzed and improved further.Afterwards, a watermarking algorithm efficiently resistive to attribute attacks is proposed in this thesis. The attribute attacks in relational databases are sorted different types and corresponding settlements are proposed respectively. In the interest of resisting attribute reordering attack, the attributes waiting for inserting watermark will be sorted by their names before watermark embedding and extracting; for resisting attribute deletion attack, the watermark will be embedded into several attributes repeatedly; for resisting attribute addition attack and alteration attack, majority voting mechanism is adopted, and also it is required the watermark be dispersive enough. The final experimental analysis shows that the algorithm is efficiently resistive to attribute attacks, and also it is robust.