Research On Strong Robustness Digital Watermarking Methods For Relational Databases

The rapid development of the technology industry is accompanied by an explosion in the amount of information.All industries need to collect and apply a large amount of data,so the scale of accumulated data is growing at an alarming rate.The business value of massive data cannot be underestimated,but its value also brings unprecedented challenges to the field of data security.As one of the bottom storage carriers of data,relational database stores large scale data,rich data content and high degree of data privacy.Driven by the interests behind the database,attackers copy,distribute,and tamper with the database owner without the knowledge of its owner.Data leakage of the database will cause huge losses,so it is urgent to protect the ownership of the database and confirm the data ownership.Therefore,in this paper,we explore a robust database watermarking method based on distortion and distortion-free for relational databases.For the distortion-free database watermark,this paper considers that the database is vulnerable to watermarking under the Queries scenario(i.e.,insert,delete,and update operations on data tables),and studies how to improve the ability of the database watermark to resist external attacks.So we propose a strongly robust distortion-free database watermarking method based on polyhedra abstract domain.For the Queries scenario of data,the method uses the polyhedra abstract domain to analyze the database program statements that perform Queries operations,obtains the corresponding numerical invariants,and determines the stable cells of the database according to the numerical invariants.At the same time,considering the security of multidimensional sensitive attributes,the database is grouped using clustering weighted multidimensional bucket grouping technique,which makes it difficult for an attacker to infer the association law of the data even if the data is leaked.By obtaining the highest bit,primary key and stable cell encoding values in specific attributes,we construct a strong robust and distortion-free database watermark,so that the watermark has better resistance to attacks in Queries scenario.For distorted database watermarking,this paper proposes a strongly robust distorted database watermarking method based on dynamic difference expansion,considering the inherent contradiction problem of improving watermark embedding capacity and reducing data distortion.The method performs singular value decomposition on the low-frequency part of the haar wavelet transformed image,extracts some of the feature values,and uses the residual feature values as the watermark sequence to be embedded,so that the watermark sequence of the same length contains more information and reduces the length of the embedded watermark.The method also combines the adaptive differential evolution algorithm and the minimum difference algorithm to select the best embedding attribute bits,and dynamically selects the traditional difference expansion technique or the improved difference expansion technique.It alleviates the problems of low computational efficiency,large data distortion,and poor robustness of traditional difference expansion techniques in embedding watermarks,and improves the embedding capacity of watermarks while reducing the distortion of data.This paper shows the feasibility of the above two database watermarking methods through experiments.The distortion-free database watermarking method is robust in Queries scenarios and does not change the original data,protecting the integrity of the database.The distorted database watermarking method can effectively improve the embedding capacity of watermark within a certain range of data distortion,so it shows good robustness when the database is under attack,and this method can also effectively trace the leaker,and has obvious advantages over existing algorithms.