RRW – A Robust and Reversible Watermarking Technique for Relational Data

RRW – A Robust and Reversible Watermarking Technique for Relational Data

Advancement in information technology is playing an increasing role in the use of information systems comprising relational databases. These databases are used effectively in collaborative environments for information extraction; consequently, they are vulnerable to security threats concerning ownership rights and data tampering. Watermarking is advocated to enforce ownership rights over shared relational data and for providing a means for tackling data tampering. When ownership rights are enforced using watermarking, the underlying data undergoes certain modifications; as a result of which, the data quality gets compromised. Reversible watermarking is employed to ensure data quality along-with data recovery. However, such techniques are usually not robust against malicious attacks and do not provide any mechanism to selectively watermark a particular attribute by taking into account its role in knowledge discovery. Therefore, reversible watermarking is required that ensures; (i) watermark encoding and decoding by accounting for the role of all the features in knowledge discovery; and, (ii) original data recovery in the presence of active malicious attacks. A robust and semi-blind reversible watermarking (RRW) technique for numerical relational data has been proposed that addresses the above objectives.

Project Demo

Existing System

            The first reversible watermarking scheme for relational databases was proposed by Zhang and Yang. In this technique, histogram expansion is used for reversible watermarking of relational database.

            Histogram expansion technique is used to reversibly watermark the selected nonzero initial digits of errors. This technique is keeps track of overhead information to authenticate data quality. However, this technique is not robust against heavy attacks.

Difference expansion watermarking techniques (DEW), exploit methods of arithmetic operations on numeric features and perform transformations. The watermark information is normally embedded in the LSB of features of relational databases to minimize distortions.

Gupta and Pieprzyks’ proposed reversible watermarking technique introduces distortions as a result of the embedding process. Changes in the data are controlled by placing certain bounds on LSB. On the contrary, to limit the distortions, the data outside the limited bounds is left unwatermarked. As a result, the watermark robustness gets compromised.

Sonnleitner proposed a robust, blind, resilient and reversible, image based watermarking scheme for large scale databases. The bit string of an image is used as a watermark where one bit from the bit string is embedded in all tuples of a single partition and the same process is repeated for the rest of the partitions. This technique demonstrates a remarkable decrease in watermark detection rate during various types of heavy attacks, and the database tuples get highly distorted.

Proposed System

            This paper proposed a robust and semi-blind reversible watermarking (RRW) technique for numerical relational data.     RRW mainly comprises a data preprocessing phase, watermark encoding phase, attacker channel, watermark decoding phase and data recovery phase.

In data preprocessing phase, secret parameters are defined and strategies are used to analyze and rank features to watermark. An optimum watermark string is created in this phase by employing GA—an optimization scheme—that ensures reversibility without data quality loss.

In the watermark encoding phase, the watermark information is embedded in the selected feature(s). Two parameters, b the optimized value from the GA and hr a change matrix are used in the watermark encoding and decoding phases.

Finally, the watermarked data for intended recipients is generated. The attacker channel comprises subset alteration, subset deletion and subset insertion attacks generated by the adversary. These malicious attacks modify the original data and try to degrade its quality.

In the watermark decoding phase the embedded watermark is decoded from the suspicious data. In order to achieve this preprocessing step is performed again, and decoding strategies (feature selection on the basis of MI, b the optimized value from the GA and hr the change matrix) are used to recover the watermark.

Semi-blind nature of RRW is used mainly for data reversibility in case of heavy attacks (attacks that may target large number of tuples). Original data is recovered in data recovery phase, through post processing steps for error correction and recovery.


  RRW is robust

 The data quality remains intact after watermarking.

Leave a Reply