t-Closeness through Microaggregation:  Strict Privacy with Enhanced Utility  Preservation(2015)

ABSTRACT:Microaggregation is a technique for disclosure limitation aimed at protecting the privacy of data subjects in microdata releases. It has been used as an alternative to generalization and suppression to generate k-anonymous data sets, where the identity of each subject is hidden within a group of k subjects. Unlike generalization, microaggregation perturbs the data and this additional masking freedom allows improving data utility in several ways, such as increasing data granularity, reducing the impact of outliers and avoiding discretization of numerical data. k-Anonymity, on the other side, does not protect against attribute disclosure, which occurs if the variability of the confidential values in a group of k subjects is too small. To address this issue, several refinements of k-anonymity have been proposed, among which t-closeness stands out as providing one of the strictest privacy guarantees. Existing algorithms to generate t-close data sets are based on generalization and suppression (they are extensions of k-anonymization algorithms based on the same principles). This paper proposes and shows how to use microaggregation to generate k-anonymous t-close data sets. The advantages of microaggregation are analyzed, and then several microaggregation algorithms for k-anonymous t-closeness are presented and empirically evaluated.

EXISTING SYSTEM:

1.  Same as for k-anonymity, the most common way to attain t-closeness is to use generalization and suppression. In fact, the algorithms for k-anonymity based on those principles can be adapted to yield t-closeness by adding the t-closeness constraint in the search for a feasible minimal generalization: in the Incognito algorithm and in the Mondrian algorithm are respectively adapted to t-closeness.
2.  SABRE is another interesting approach specifically designed for t-closeness. In SABRE the data set is first partitioned into a set of buckets and then the equivalence classes are generated by taking an appropriate number of records from each of the buckets.

DISADVANTAGES OF EXISTING SYSTEM:

1.  The buckets in SABRE are generated in an iterative greedy manner which may yield more buckets than our algorithm (which analytically determines the minimal number of required buckets). A greater number of buckets leads to equivalence classes with more records and, thus, to more information loss.
2.  It is not known how to optimally combine generalization and local suppression.
3.  There is no agreement in the literature on how suppression should be performed: one can suppress at the record level (entire record suppressed), or suppress particular attributes in some records; furthermore, suppression can be done by either blanking a value or replacing it by a neutral value (i.e.some kind of average).
4.  Last but not least, and no matter how suppression is performed, it complicates data analysis (users need to resort to software dealing with censored data).

PROPOSED SYSTEM:

1.  A first contribution of this paper is to identify the strong points of microaggregation to achieve k-anonymous t-closeness. The second contribution consists of three new microaggregation-based algorithms for t-closeness, which are presented and evaluated.
2.  We propose three different algorithms to reconcile conflicting goals.
3.  The first algorithm is based on performing microaggregation in the usual way, and then merging clusters as much as needed to satisfy the t-closeness condition. This first algorithm is simple and it can be combined with any microaggregation algorithm, yet it may perform poorly regarding utility because clusters may end up being quite large.
4.  The other algorithms modify the microaggregation algorithm for it to take t-closeness into account, in an attempt to improve the utility of the anonymized data set.
5.  Two variants are proposed: k-anonymity-first (which generates each cluster based on the quasi-identifiers and then refines it to satisfy t-closeness) and t-closenessfirst (which generates each cluster based on both quasiidentifier attributes and confidential attributes, so that it satisfies t-closeness by design from the very beginning).

ADVANTAGES OF PROPOSED SYSTEM:

1.  Microaggregation has several advantages over generalization/recoding for k-anonymity that are mostly related to data utility preservation:
2.  Global recoding may recode some records that do not need it, hence causing extra information loss. On the other hand, local recoding makes data analysis more complex, as values corresponding to various different levels of generalization may co-exist in the anonymized data. Microaggregation is free from either drawback.
3.  Data generalization usually results in a significant loss of granularity, because input values can only be replaced by a reduced set of generalizations, which are more constrained as one moves up in the hierarchy. Microaggregation, on the other hand, does not reduce the granularity of values, because they are replaced by numerical or categorical averages.
4.  If outliers are present in the input data, the need to generalize them results in very coarse generalizations and, thus, in a high loss of information. For microaggregation, the influence of an outlier in the calculation of averages/centroids is restricted to the outlier’s equivalence class and hence is less noticeable.
5.  For numerical attributes, generalization discretizes input numbers to numerical ranges and thereby changes the nature of data from continuous to discrete. In contrast, microaggregation maintains the continuous nature of numbers.
6.  In microaggregation one seeks to maximize the homogeneity of records within a cluster, which is beneficial for the utility of the resultant k-anonymous data set.

MODULES:

 Dataset Extraction

 Pre-processing

 Cluster Formation

 t-Closeness Aware Microaggregation

 Evaluation Result

MODULES DESCSRIPTION:

Dataset Extraction:

The first module is Dataset Extraction. First we load the data set. The dataset extraction module can be used to extract the dataset and it will be stored in the database for future use. Initially the dataset was selected, after that it will be split separate data and it can be stored in the table to the user database.

Pre-processing

In this module, have to pre-process the dataset. Here the adult dataset is taken. Train and test dataset is considered. This data set consists of 15 attributes (class attribute). The prediction task associated with the Adult data set is to determine based on census and demographic information about people. The data set contains both categorical and numerical attributes. Although the Age attribute in the Adult data set is numerical. And they filter the incomplete records from the adult dataset.

Cluster Formation

This module is used to measure the success of the method in removing all evidence of direct and/or indirect discrimination from the original data set; on the other hand, have to measure the impact of the method in terms of information loss (i.e., data quality loss). In a first test we used as evaluation data the Census data set, which is usual to test privacy protection methods record with numerical attributes.

Because k-anonymity and t-closeness pursue goals (the former clusters records with similar quasi identifiers while the latter requires clusters with a distribution of confidential attributes similar to the one of the entire data set), we defined data sets according to the correlation between the values of quasi-identifier and confidential attributes

t-Closeness Aware Microaggregation:

This module is deal with the key problem of transforming data with minimum information loss to prevent at the same time both direct and indirect discrimination. Global recoding may recode some records that do not need it, hence causing extra information loss. On the other hand, local recoding makes data analysis more complex, as values corresponding to various different levels of generalization may co-exist in the anonymized data. Microaggregation is free from either drawback. Generating a t-close data set via generalization is essentially an optimization problem: one must find a minimal generalization that satisfies t-closeness. A common way to find a solution is to iteratively generalize one of the attributes (selected according to some quality criterion) until the resulting data set satisfies t-closeness. Our first proposal to attain t-closeness via microaggregation follows a similar approach. We microaggregate and then merge clusters of records in the microaggregated data set; we use the distance between the quasi-identifiers of the microaggregated clusters as the quality criterion to select which groups are to be merged. Initially, the microaggregation algorithm is run on the quasi-identifier attributes of the original data set; this step produces a k-anonymous data set. Then, clusters of microaggregated records are merged until t-closeness is satisfied. We iteratively improve the level of t-closeness by: i) selecting the cluster whose confidential attribute distribution is most different from the confidential attribute distribution in the entire data set (that is, the cluster farthest from satisfying t-closeness); and ii) merging it with the cluster closest to it in terms of quasiidentifiers.

Evaluation Result

In this module, have to evaluate the whole result and also displayed in table. The above mentioned methods have to evaluate and finally they have to present with the parameter alpha which is the fixed value.

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

 System                          :         Pentium IV 2.4 GHz.

 Hard Disk                      :         40 GB.

 Floppy Drive                 :         1.44 Mb.

 Monitor                         :         15 VGA Colour.

 Mouse                            :         Logitech.

 Ram                               :         512 Mb.

SOFTWARE REQUIREMENTS:

 Operating system           :         Windows XP/7.

 Coding Language :         JAVA/J2EE

 IDE                      :         Netbeans 7.4

 Database              :         MYSQL

REFERENCE:

Jordi Soria-Comas, Josep Domingo-Ferrer, Fellow, IEEE, David S´anchez and Sergio Mart´inez, “t-Closeness through Microaggregation: Strict Privacy with Enhanced Utility Preservation”, IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2015.