Verify a Valid Message in Single Tuple: A Watermarking Technique for Relational Database

Abstract

The leakage of sensitive digital assets is a major problem which causes huge legal risk and economic loss. Service providers need to ensure data security for owners. Robust watermarking techniques play a critical role in ownership protection of relational databases. In this paper, we proposed a new Double-layer Ellipse Model called DEM that embeds a valid message in each candidates tuples. Each watermark can independently prove ownership. The main idea of DEM is to use watermarks itself to locate and make the most of contextual information to verify validity of watermarks. Under the framework of DEM, we propose a robust and semi-blind reversible watermarking scheme. Our scheme handles non-significant data for locating and embedding. The scheme generates watermark by exchanging data groups extracted from scattered attributes using Computation and Sort-Exchange step. Key information such as primary key, most significant bit (MSB) become a assistant feature for verifying the validity of embedded watermark. Our scheme can be applied on all type of numerical attributes (e.g. Integer, float, double, boolean). In robust experiments, the scheme is proved to be extremely resilient to insertion/detection/alteration attacks in both normal and hard situations. From a practical point of view, our solution is easy to implement and has good performance in statistics, incremental updates, adaptation.