As individuals are increasingly benefiting from the use of online services, there are growing concerns about the treatment of personal information. Society’s ongoing response to these concerns often gives rise to privacy policies expressed in legislation and regulation. These policies are written in natural language (or legalese) as privacy agreements that users must agree to, or presented as a set of privacy settings and options that users must opt in or out of in order to receive the service they want. But comprehensibility of privacy policies and settings is becoming increasingly challenging as agreements become longer and there are many privacy options to choose from. Additionally, organizations face the challenge of assuring compliance with policies that govern collecting, using, and sharing of personal data. This thesis proposes frameworks for personalized privacy and privacy auditing to address these two problems.

In this thesis, we focus our investigation on the comprehensibility issues of personalized privacy using the concrete application domain of personal health data as recorded in systems known as personal health records (PHR). We develop the Privacy Goals and Settings Mediator (PGSM) model, which is based on i* multi-agent modelling techniques, as a way to help users comprehend privacy settings when employing multiple services over a web platform. Additionally, the PGSM model helps privacy experts contribute their privacy knowledge to the users’ privacy decision-making task. To address the privacy auditing problem, we propose two light-weight ontologies, L2TAP and SCIP, that are designed for deployment as Linked Data, an emerging standard for representing and publishing web data. L2TAP (Linked Data Log to Transparency, Accountability and Privacy) provides flexible and extensible provenance-enabled logging of privacy events. SCIP (Simple Contextual Integrity Privacy) provides a simple target for mapping the key concepts of Contextual Integrity and enables SPARQL query-based solutions for two important privacy processes:​ compliance checking and obligation derivation. This thesis validates the premise of PHR users’ privacy concerns, attitudes and behaviour through an empirical study. The usefulness of the PGSM model for privacy experts is evaluated through interviews with experts. Finally, the scalability and practical benefits of L2TAP+SCIP for log-based privacy auditing are validated experimentally.