Enabling information recovery with ownership using robust multiple watermarks

Abstract

With the increasing use of databases, there is an abundant opportunity to investigate new watermarking techniques that cater to the requirements for emerging applications. A major challenge that needs to be tackled is to recover crucial information that may be lost accidentally or due to malicious attacks on a database that represents asset and needs protection. In this paper, we elucidate a scheme for robust watermarking with multiple watermarks that resolve the twin issues of ownership and recovery of information in case of data loss. To resolve ownership conflicts watermark is prepared securely and then embedded into the secretly selected positions of the database. Other watermark encapsulates granular information on user-specified crucial attributes in a manner such that the perturbed or lost data can be regenerated conveniently later. Theoretical analysis proves that the probability of identifying target locations, False hit rate and False miss rate is negligible. We have experimentally verified that the proposed technique is robust enough to extract the watermark accurately even after 100% tuple addition or alteration and after 98% tuple deletion. Experiments on information recovery reveal that successful regeneration of tampered/lost data improves with the increase in the number of candidate attributes for embedding the watermark.