Research On Fragile Digital Watermarking In Relational Databases

With the fast development of information technology, an increasing number ofdatabases (e.g., parametric specifications, surveys, and life sciences data) aredistributed through the internet. The piracy problem has become one of the mostdevastating threats to networking systems and electronic business. Besides, onlinedistributed data can be easily tampered in a malicious way that may invalidate theirlegitimate usage as evidence on the court in case of dispute. Thus the authenticity andintegrity of database content is a critical issue that has to be taken care seriously.In this study, this thesis focus on several major problems that identified fromprevious relational databases watermarking techniques. First, existing watermarkingtechnigues depend on a specific order of database attributes, marking them vulnerableto attribute related attacks. Second, most of watermarking techniques aredistortion-based and only designed for specific data types, such as numeric data,categorical data, non-numeric multiword attributes, etc. Third, in addition to thetamper detection and localization, tamper testoration has not been resolved on mostintegrity verification scheme. That makes some important (sensitive) data lost theirvalues forever after being modified.To solve these problems, the paper mainly discussed the application of fragilewatermarking in terms of integrity verification in relational databases. The maincontribution of this thesis includes:a) This thesis proposes a method based on zero-watermarking approach forcontent integrity checking of database relations. The method is called“zero-watermark” because the watermark bits are not actually encoded in the data.Instead, this thesis use a trusted third party called Certification Authority (CA) withwhom the watermark is registered and notarized along the original data forcertification purpose. This method resolves the inherent conflict between security andimperceptibility. Furthermore, there is no constraint on the type of attributes beingwatermarked. Security analysis and experimental results showed that our proposedmethod can detect and localized malicious data modifications with high probability.b) In order to improve the availability of data in relational databases，this thesispropose a novel tamper recovery fragile watermarking techinique. The proposedmethod based Reed-Solomon coding technique can restore partial alterations of critical data’s key information in watermarked databases. Reed-Solomon codes areerror detecting and correcting non binary cyclic codes. RS codes are widely used indigital communication and storage systems due to their excellent random and bursterror-correcting capability. By using the embedded watermark based onReed-Solomon coding techinique, the proposed method can not only detect andlocalize the modifications at group level, but also recover the key information ofsensitive attribute values from the database mo dified locations. Theoretical analysisand experimental results demonstrate that this scheme is effective. Moreover,database owners can choose a suitable Reed-Solomon code based on databasescapacity, and decide the effect of detection or correction accor ding to theirrequirements.