A Robust Database Watermarking using Local Differential Privacy

Abstract

Nowadays, privacy protection is built on the huge sharing of big data development and rapid development of information, and copyright protection is needed for immediate resolution. However, any kind of technology is not existing which can conclude the problems of copyright and the protection of privacy continuously. Existing database watermarking systems have less privacy protection though the database watermarking is a research topic that has been widely studied for its feature of copyright tracing. In this paper, we develop an experimental technique, call local differential privacy-based database watermarking to protect the privacy as well as the ownership of the database. In the technique, watermarking embeds the database to determine the location of the database and local differential privacy protect the privacy of users. Especially our two proposed methods are the Laplace Mechanism-based Database Watermarking (LMDW) and Randomized Response-based Database Watermarking (RRDW) for two classical local differential privacy mechanisms the Laplace Mechanism (LM) and the Randomized Response (RR) respectively. Watermarking is designed with the distortion created by local differential privacy so that algorithms can reduce the data distortion caused by local differential privacy and maintain the local differential privacy nature which is proven by formal theoretical analysis and experimental evaluation. A sensitive database test suggests that the proposed algorithms have higher utility and strong consistency against various watermarking attacks.