Inference-Aware Aggregation Control in Data Warehouses to Prevent Statistical Disclosure

This paper presents an inference-aware aggregation control mechanism for data warehouses to prevent statistical disclosure caused by small group sizes in aggregation queries. Although aggregate functions such as SUM, AVG, and COUNT are commonly assumed to preserve privacy, aggregation over small groups may lead to inference attacks and unintended leakage of sensitive information. The proposed approach evaluates the size of each group produced by GROUP BY queries and conditionally suppresses or generalizes aggregate results when the group size falls below a predefined threshold. The framework is implemented using standard SQL constructs in MySQL without requiring any modification to the database engine. Experimental evaluation on a synthetic employee data warehouse demonstrates that the proposed method significantly reduces inference risk while preserving analytical utility for sufficiently large groups. The approach is practical, lightweight, and suitable for real-world data warehouse environments.