Modelling and Automatically Analysing Privacy Properties for Honest-but-Curious Adversaries

For certain communication protocols, undetectability and unlinkability of messages or information items are desirable properties, and are used to reason about anonymity and privacy. Previous work has formalized and analysed these properties using the notions of indistinguishability and observational equivalence. However, it is also possible to perform this analysis using a constructive definition of the adversary model an approach that has received less attention. The semihonest or honest-but-curious (HBC) adversary is commonly used in the analysis of these privacy properties. In this work, we develop a formal model of the capabilities of an HBC adversary with respect to undetectability and unlinkability. Our HBC model is defined as a deductive system consisting of a set of inference rules. We show that each rule is based on commonly accepted definitions and therefore claim that our overall model is a faithful representation of these definitions. The advantage of this constructive approach is that our HBC model can be directly integrated with methodologies for analysing security properties. We demonstrate this by integrating our HBC model with Casper/FDR, an established protocol analysis method based on the process algebra of CSP. We extend the Casper tool to also analyse undetectability and unlinkability properties for multiple adversaries based on a single description of the protocol. We demonstrate the effectiveness of our HBC model and Casper extension by analysing several protocols in terms of their security and privacy properties. In our case studies, we find new attacks as well as rediscover known attacks.