Study On Microaggregation Algorithm For Sensitive Attributes Diversely

The purpose of privacy protection technologies is protecting the relationshipbetween sensitive attributes and unique identifiers of individuals when publishing data.Deleting the identifiers just cannot avoid exposing privacy where we release data. Theattacker can still get the individual’s privacy by linking. To solve this problem,k-anonymity is proposed to protect privacy information. It requires that there exist atleast k records which cannot be distinguished on quasi-identifier in the published data,making the attacker cannot identify the privacy belongs to which individual. Then, itgains the purpose to protect personal information.But, the defect of k-anonymity is that it has not any constraints on sensitive data.The attackers can discover private data by exploiting homogeneity attack andbackground attack. In order to solve this problem, scholar proposed l-diversity; itrequires at least l distinct values in every equivalence group, effectively reducing theprobability of exposing of privacy. In this thesis, a micro-aggregation algorithm isproposed to realize the diversity of sensitive attributes, which is based onmicro-aggregation and l-diversity model. This algorithm ensures data availability whilethe safe of the released data effectively. The main work is as follows:①This thesis studies the basic principle, the realization method and the evaluationmethod of the k-anonymity. Also, this thesis studies the theoretical knowledge ofmicro-aggregation algorithm, including the relevant calculations method ofmicro-aggregation. By studying the step that how micro-aggregation achievek-anonymity and traditional micro-aggregation algorithm achieved k-anonymity, thisthesis analyzes theirs advantages and disadvantages, to find how to improve them andenhance its ability to protect privacy while ensuring the availability of the release data.②This thesis proposes micro-aggregation algorithm realized the diversity ofsensitive attributes. Through the study of k-anonymity and l-diversity model, this thesisproposes a sensitive attribute diversity Micro-aggregation algorithm based on traditionalmicro-aggregation MDAV algorithm, the algorithm groups l nearest tuples to clustercenter into one group, which has l different sensitive values. Then, it extends this clusterbased on satisfying the l-diversity, so the anonymous table l-diversity yield by thealgorithm satisfies sensitive attribute l-diversity can resist homogeneity attack andbackground knowledge attack. Meanwhile, the algorithm keeps the Micro-aggregation’s features of simplicity and effective, also ensures low information loss and highefficiency.③This thesis compares this algorithm with traditional MDAV algorithm byexperiment to verify the performance of the algorithm proposed by this thesis.Experimental results show: the algorithm can generate anonymity table to satisfy theneeds of sensitive attribute diversity. It effectively reduces the average leakageprobability of sensitive attribute in single sensitive attribute conditions, enhances theability of release data to resist homogeneity attack and background knowledge attack,and improved data security. At the same time, data yield by this algorithm has lowinformation loss and good availability. In addition, this high efficiency algorithm hasgood time performance. It should be noted that the information loss and timeperformance of this algorithm are not ideal in multi-sensitive attributes conditions. Itneeds further improvement and optimization.Experimental results show that the algorithm can generate anonymity table tosatisfy the needs of sensitive attribute diversity, and to ensure the availability ofanonymous table,...