Empirical analysis of the impact of queries on watermarked relational databases

Robust relational database watermarking techniques aim to preserve the watermark in the protected assets despite daily updates and malicious attacks. Nevertheless, most techniques do not analyze the root of operations that can degrade the embedded watermark signal, being that the case of Structured Query Language (SQL) queries for relational data. In this paper, we perform an empirical analysis of the role played by different types of SQL queries in the persistence and degradation of a watermark signal in a relational data set. According to the type of operation, we present a set of alternatives for the watermark detection process, contributing to the extraction of an ownership copyright signals threshold. Moreover, our proposal allows data owners to consider variations for the selected watermarking technique, according to the suspected malicious operation performed in the copy of available data. Finally, we give evidence of the importance of some features to be considered when selecting a watermark source, making possible the resilience against update operations that affect a high amount of protected data and the enhancement of a partially detected watermark.