New Semantics Improves Confidentiality-Preserving Query Answering in DL Ontologies

Controlled Query Evaluation (CQE) is a declarative method for ensuring data access adheres to confidentiality policies, particularly within Description Logic (DL) ontologies where policies are expressed via Epistemic Dependencies (EDs). Previous studies showed that query answering under established CQE semantics (GA- and IGA-entailment) is computationally intractable for DL-Lite_R TBoxes, and IGA semantics also failed to satisfy the crucial indistinguishability property for confidentiality preservation. To address these challenges, researchers propose a new CQE semantics based on the concept of Minimal Policy Violation (MPV). This novel approach aims to provide computationally easier and confidentiality-preserving forms of CQE. The MPV semantics is demonstrated to be a sound approximation of prior semantics and, importantly, satisfies the indistinguishability property, ensuring that sensitive information cannot be inferred even by observing query results. Furthermore, for DL-Lite_R ontologies, query entailment under MPV semantics can be decided in polynomial time concerning data complexity, a significant improvement in tractability. A software implementation of this framework has been developed and evaluated, confirming its feasibility using an existing OWL 2 QL benchmark. This work offers a practical path toward more secure and efficient data access in knowledge-based systems.