microaggregation

Microaggregation is a data anonymization technique used to protect privacy in statistical databases. The core idea is to group similar records together and replace each record with the aggregated values of its group. This process creates a k-anonymous dataset, where k is the size of each group. The goal is to make it difficult to identify individuals within the dataset by ensuring that any record in the anonymized dataset belongs to a group of at least k individuals with similar characteristics.

The process typically involves defining a similarity metric and a grouping algorithm. Records are considered similar if they share common attribute values or fall within a certain distance in a multi-dimensional space. Once records are grouped, their identifying attributes are replaced by the mean, median, or mode of the group's values for those attributes. This generalized data retains statistical utility while obscuring individual details.

Microaggregation is particularly useful for protecting sensitive information in datasets where direct identifiers are present. It offers a trade-off between privacy and data utility, as the aggregation process can lead to some loss of precision in the data. The effectiveness of microaggregation depends on the choice of similarity metric, the grouping strategy, and the value of k. Larger values of k provide stronger privacy guarantees but may result in greater data distortion. It is a widely used method in statistical disclosure control.