mikroaggregation

Mikroaggregation is a privacy-preserving data anonymization technique. It aims to protect individual privacy within a dataset by grouping similar records together and then replacing the original values of sensitive attributes with aggregate values for the group. The core idea is to ensure that each individual record is indistinguishable from at least k-1 other records in the dataset, where k is a predefined parameter.

The process typically involves several steps. First, the dataset is divided into a set of disjoint partitions, where each partition contains k records. Then, for each partition, sensitive attributes are replaced with their mean, median, or mode within that partition. This ensures that any query targeting an individual within a group will receive an aggregated value, making it difficult to infer the exact value of that individual's sensitive attribute.

Mikroaggregation offers a balance between privacy and data utility. By maintaining group-level statistics, it allows for aggregate analysis of the data while obscuring individual identities. However, the effectiveness of mikroaggregation depends on the choice of k and the similarity metric used for grouping records. A larger k generally provides stronger privacy but can lead to a greater loss of data utility. It is particularly useful in scenarios where tabular data needs to be shared or analyzed while minimizing the risk of re-identification.