A Novel Method for Micro-Aggregation in Secure Statistical Databases Using Association and Interaction

We consider the problem of micro-aggregation in secure statistical databases, by enhancing the primitive Micro-Aggregation Technique (MAT), which incorporates proximity information. The state-of-the-art MAT recursively reduces the size of the data set by excluding points which are farthest from the centroid, and those which are closest to these farthest points, while it ignores the mutual Interaction between the records. In this paper, we argue that inter-record relationships can be quantified in terms of two entities, namely their "Association" and "Interaction". Based on the theoretically sound principles of the neural networks (NN), we believe that the proximity information can be quantified using the mutual Association, and their mutual Interaction can be quantified by invoking transitive-closure like operations on the latter. By repeatedly invoking the inter-record Associations and Interactions, the records are grouped into sizes of cardinality "k", where k is the security parameter in the algorithm. Our experimental results, which are done on artificial data and on the benchmark data sets for real-life data, demonstrate that the newly proposed method is superior to the state-of-the-art by as much as 13%.