Database Watermarking Based On Text Format

The easy production and large transmission of digital data in today's internet-based application environments caused the problem of piracy of data owners. Proof of ownership rights of outsourced relational databases is a crucial issue that has to be taken into much attention. Data service providers have provided an environment that has allowed users to search and access databases remotely. This environment exposes the data service providers to the threat of data theft and intellectual right is violated. Therefore, there is a need to develop capabilities to identify and prevent data piracy.Digital watermarking was proposed as a new alternative method to enforce intellectual property rights and protect databases from such piracy.Each watermarking method consists of an embedding and extracting algorithm. In the embedding algorithm, a water mark is inserted into the data and while in the extracting algorithm, the watermark information is extracted. The watermarks embedded into the original data should not significantly affect the quality of the original data and should not be easily destroyed. In these algorithms, there might be little or no perceptible differences between the original data content and the watermarked one.In this thesis, a new relational database watermarking method for non numeric data based on text format is proposed. In this method, a watermark is embedded by either shifting or not shifting the horizontal location of a word within an attribute to the right side depending on the watermarking bits. The main idea is to embed secret information by modifying the structural format of the position of the word, so that the resulting watermarked tuples or attributes are similar to the normal unmarked tuples or attributes. In this way, the secret information cannot be obtained by anyone other than the computer program with the secret key known only to the owner of the database. The words are shifted in the way that it doesn't change their meanings. We used the Levenshtein distance between words within attributes of selected tuples to locate the position where to hide the watermark.The relation database to be watermarked doesn't need to have only non numerical attributes, other types of data also may be exist but only non numerical multi word attributes will carry the marks. The proposed method can be applied to any relation with at least a primary key and an attribute. Extraction of the watermark bits follows the same process as embedding and the watermark bits are extracted in the location determined by the minimum Levenshtein Distance between every two successive words within the attributes of selected tuples.The algorithms' performance was validated based on the standard watermarking attacks. It is more effective and robust since it is immune to the different forms of malicious attacks and does not require the original database in order to extract the embedded watermark.The robustness of the proposed algorithm was verified against a number of database attacks such as subset deletion, subset addition and subset alteration attack. The algorithm's performance when tested against different common attacks shows high resilience as proven by a series of experiments that we conducted. When compared with other existing methods, our algorithm outperforms some of them. Our new algorithm also contributes to the understanding of information hiding and addresses both theoretical and practical aspects of watermarking techniques.