Josep Maria Mateo-Sanz

Practical data-oriented microaggregation for statistical disclosure control

Microaggregation is a statistical disclosure control technique for microdata disseminated in statistical databases. Raw microdata (i.e., individual records or data vectors) are grouped into small aggregates prior to publication. Each aggregate should contain at least k data vectors to prevent disclosure of individual information, where k is a constant value preset by the data protector. No exact polynomial algorithms are known to date to microaggregate optimally, i.e., with minimal variability loss. Methods in the literature rank data and partition them into groups of fixed-size; in the multivariate case, ranking is performed by projecting data vectors onto a single axis. In this paper, candidate optimal solutions to the multivariate and univariate microaggregation problems are characterized. In the univariate case, two heuristics based on hierarchical clustering and genetic algorithms are introduced which are data-oriented in that they try to preserve natural data aggregates. In the multivariate case, fixed-size and hierarchical clustering microaggregation algorithms are presented which do not require data to be projected onto a single dimension; such methods clearly reduce variability loss as compared to conventional multivariate microaggregation on projected data.

Efficient multivariate data-oriented microaggregation

Microaggregation is a family of methods for statistical disclosure control (SDC) of microdata (records on individuals and/or companies), that is, for masking microdata so that they can be released while preserving the privacy of the underlying individuals. The principle of microaggregation is to aggregate original database records into small groups prior to publication. Each group should contain at least k records to prevent disclosure of individual information, where k is a constant value preset by the data protector. Recently, microaggregation has been shown to be useful to achieve k-anonymity, in addition to it being a good masking method. Optimal microaggregation (with minimum within-groups variability loss) can be computed in polynomial time for univariate data. Unfortunately, for multivariate data it is an NP-hard problem. Several heuristic approaches to microaggregation have been proposed in the literature. Heuristics yielding groups with fixed size k tends to be more efficient, whereas data-oriented heuristics yielding variable group size tends to result in lower information loss. This paper presents new data-oriented heuristics which improve on the trade-off between computational complexity and information loss and are thus usable for large datasets.

Grape-seed procyanidins modulate inflammation on human differentiated adipocytes in vitro

Flavonoids are functional constituents of many fruits and vegetables. Procyanidins are flavonoids with an oligomeric structure, and it has been shown that they can improve the pathological oxidative state of a diabetic situation. To evaluate whether procyanidins can modulate inflammation, an event strongly associated with obesity, diabetes and insulin resistance states, we used human adipocytes (SGBS) and macrophage-like (THP-1) cell lines and administered an extract of grape-seed procyanidins (GSPE). THP-1 and SGBS cells pre-treated with GSPE showed a reduction of IL-6 and MCP-1 expression after an inflammatory stimulus. GSPE stimuli alone modulate adipokine (APM1 and LEP) and cytokine (IL-6 and MCP-1) gene expression. GSPE partially inhibited NF-kappaB translocation to the nucleus in both cell lines. These preliminary findings demonstrate that GSPE reduces the expression of IL-6 and MCP-1 and enhances the production of the anti-inflammatory adipokine adiponectin suggesting that may have a beneficial effect on low-grade inflammatory diseases such obesity and type 2 diabetes.

Probabilistic Information Loss Measures in Confidentiality Protection of Continuous Microdata

Inference control for protecting the privacy of microdata (individual data) should try to optimize the tradeoff between data utility (low information loss) and protection against disclosure (low disclosure risk). Whereas risk measures are bounded between 0 and 1, information loss measures proposed in the literature for continuous data are unbounded, which makes it awkward to trade off information loss for disclosure risk. We propose in this paper to use probabilities to define bounded information loss measures for continuous microdata.

Post-Masking Optimization of the Tradeoff between Information Loss and Disclosure Risk in Masked Microdata Sets

Previous work by these authors has been directed to measuring the performance of microdata masking methods in terms of information loss and disclosure risk. Based on the proposed metrics, we show here how to improve the performance of any particular masking method. In particular, post-masking optimization is discussed for preserving as much as possible the moments of first and second order (and thus multivariate statistics) without increasing the disclosure risk. The technique proposed can also be used for synthetic microdata generation and can be extended to preservation of all moments up to m-th order, for any m.

On the security of microaggregation with individual ranking: analytical attacks

Microaggregation is a statistical disclosure control technique. Raw microdata (i.e. individual records) are grouped into small aggregates prior to publication. With fixed-size groups, each aggregate contains k records to prevent disclosure of individual information. Individual ranking is a usual criterion to reduce multivariate microaggregation to univariate case: the idea is to perform microaggregation independently for each variable in the record. Using distributional assumptions, we show in this paper how to find interval estimates for the original data based on the microaggregated data. Such intervals can be considerably narrower than intervals resulting from subtraction of means, and can be useful to detect lack of security in a microaggregated data set. Analytical arguments given in this paper confirm recent empirical results about the unsafety of individual ranking microaggregation.

Outlier Protection in Continuous Microdata Masking

Masking methods protect data sets against disclosure by perturbing the original values before publication. Masking causes some information loss (masked data are not exactly the same as original data) and does not completely suppress the risk of disclosure for the individuals behind the data set. Information loss can be measured by observing the differences between original and masked data while disclosure risk can be measured by means of record linkage and confidentiality intervals. Outliers in the original data set are particularly difficult to protect, as they correspond to extreme inviduals who stand out from the rest. The objective of our work is to compare, for different masking methods, the information loss and disclosure risk related to outliers. In this way, the protection level offered by different masking methods to extreme individuals can be evaluated.

Fast generation of accurate synthetic microdata

Generation of a synthetic microdata set that reproduces the statistical properties of an original microdata set is a promising approach to statistical disclosure control (SDC) of microdata. In this paper, a new method for generating continuous synthetic microdata is proposed. The covariance matrix and the univariate statistics of the original data set are exactly preserved. The method is non-iterative and its complexity grows linearly with the number of records to be protected.

Multivariate Microaggregation Based Genetic Algorithms

Microaggregation is a clustering problem with cardinality constraints that originated in the area of statistical disclosure control for micro data. This article presents a method for multivariate microaggregation based on genetic algorithms (GA). The adaptations required to characterize the multivariate microaggregation problem are explained and justified. Extensive experimentation has been carried out with the aim of finding the best values for the most relevant parameters of the modified GA: the population size and the crossover and mutation rates. The experimental results demonstrate that our method finds the optimal solution to the problem in almost all experiments when working with small data sets. Thus, for small data sets the proposed method performs better than known polynomial heuristics and can be combined with these for larger data sets. Moreover, a sensitivity analysis of parameter values is reported which shows the influence of the parameters and their best values

A 2/sup d/-tree-based blocking method for microaggregating very large data sets

Blocking is a well-known technique used to partition a set of records into several subsets of manageable size. The standard approach to blocking is to split the records according to the values of one or several attributes (called blocking attributes). This paper presents a new blocking method based on 2/sup d/-trees for intelligently partitioning very large data sets for micro aggregation. A number of experiments has been carried out in order to compare our method with the most typical univariate one.