Analyzing Causes of Privacy Mismatches in Service Oriented Architecture

Internet users want controlled disclosure of their private data. They are concerned about what personal information they may reveal inadvertently while accessing websites. Intelligent systems can alleviate user’s concern by assessing website’s data practices automatically, assuming machine readable privacy policies. In case of mismatch with user expectations, these systems can also help both parties reviewing their privacy statements by providing useful information. In the context of the collaborative research project PrimeLife (Privacy and Identity Management for Europe in Life), IBM, SAP, ULD, W3C and European Microsoft Innovation Center (EMIC) are working on new languages to define privacy policies. Specifying logic-based languages is important to enable reasoning on mismatches, i.e. understanding why service’s privacy policy does not match user’s privacy preferences. This master thesis, done with EMIC, uses domain specific language to specify privacy and focuses on mechanisms to analyze mismatches and to propose modifications for getting a match, at a higher abstraction level, e.g. DSL. In case of mismatch, this guidance permits the user judging the required amendments and make the right choice thereby, i.e. reject service’s policy or modify her preference accordingly. Another concern of this work is separating different aspects of a privacy management system and link them effectively as required. The proposed approach is validated by developing a proof-of-concept prototype implementation with Microsoft’s textual DSL tool, MGrammar and an existing formal language, Formula.