Reasoning on privacy policies

Privacy policies stated in natural language are often ambiguous and not amenable to direct machine processing. Controlled Natural Languages (CNLs) provide a simplified, structured way to express policy rules that is both easier for natural language processing (NLP) to parse and closer to formal representations used in automated reasoning (AR). By translating further CNL statements into logical systems, we can connect NLP with rigorous AR to verify privacy policies and related regulations. This paper proposes a three-layer methodology – based on NLP, CNL, and AR – for the formal verification of privacy policies, and describes our ongoing development of this approach.