Robust Reversible Watermarking for relational database

The large structure of dataset for sharing for particular target or authenticate target but there are some other parties that can attack on data easily and use that data illegally and can claim that data ownership. Attacker can gain ownership on that sharing data. Due to this original data get modified and quality of data also reduced so this original data is not useful for any extraction information system, it gives irrelevant data or reduced data which is not useful for further processing. To avoid this we used a system Reversible Watermarking which protects data from attack of middle parties while sharing data and also preserve ownership of the data. It avoids data tampering and reduction of data. Quality of data also gets preserved. Feature selection in RRW uses all combinations of features to calculate importance (MI) of the features. Also RRW does not support non-numeric data. We introduced technique which works on nominal data and uses less features for calculation which enhance the speed and accuracy and performance of RRW. In supervised learning, feature's importance depends on co-relation between Feature and class variable, there is no need to consider all combination.