| Nowadays,big data technology has become a vital part of Internet technology.How to use data analysis methods to dig out the value behind the data is the core issue of big data technology.At the same time,the issue of secure data distribution has become more and more important.As an important data storage method,relational databases are also facing various challenges in distribution.Relational databases are easy to leaks during distribution.How to trace the leakers after the leakage of the relational databases has turned a research focus.In view of the possible leakage problems of the relational database after the distribution.For the numerical and non-numerical relational databases,this paper proposes two traceability methods respectively to ensure the safe distribution of the relational.This paper has carried out some work as follows.According to the characteristics of the large redundant space of the numerical relational database data,a database fingerprint traceability method based on multi-layer coding and random forest algorithm was proposed.For the problem of most database fingerprints have poor collusion resistance and less user capacity.This paper uses multi-layer coding method to combine Optimal Focused Fingerprints from Orthogonality codes(OFFO)and Cover Free Family codes(CFF).While improving fingerprint resistance to collusion,this method expands the number of users that can be accommodated.In addition,the database fingerprint embedding algorithm also faces the problem of poor database availability and poor robustness of fingerprint embedding.This method scores the importance of attributes to select attributes by the random fores algorithm,and filter tuples according to the second MSB(Most Significant Bit)of the attribute value.This method can improve the robustness of the database fingerprint embedding algorithm while ensuring the availability of the database.According to the small redundancy space of non-numerical relational databases,this paper proposes a zero-watermark traceability method based on keeping reference relationship and attribute value classification.Since any degree of data distortion will have a greater impact on the usability of non-numerical relational databases,this paper uses a zero-watermark method of attribute value classification to embed it without distortion.Then,in view of the problem that the zero-watermark can not be used to trace the source of a relational database,the method of splitting the database by keeping the reference relationship makes the zero-watermark method have the ability of leaking the source.Finally,this paper verifies that the fingerprint generated by the database fingerprint traceability method has good anti-collusion ability and can accommodate more users through multiple sets of comparative experiments,and the traceability method has strong fingerprint robustness while ensuring the availability of the database.this paper also verifies the feasibility of the zero-watermark traceability method through multiple sets of comparative experiments.The experimental results show that the data of this method is not distorted and has good robustness under different attacks. |
