An entropy-based framework for database inference

Abstract

In this paper, we discuss the database inference problem. We look at both query-based and partial view-based cases of the problem, concentrating our efforts on classification rules related to the partial view-based case. Based on this analysis, we develop a theoretical formulation to quantify the amount of private information that may be inferred from a public database and we discuss ways to mitigate that inference. Finally, we apply this formulation to actual downgrading issues. Our results are dependent upon the knowledge engine used to derive classification rules. We use C4.5 since it is a well-known and popular robust software tool.