Towards the Quantification of Strategy Leakage

This paper reports first steps towards a formal model for strategy leakage. We generalize the representation of prior adversarial knowledge from a distribution on secrets to a distribution on strategies for generating secrets, which we call an environment. Applying information-theoretic techniques to environments allows us to disentangle the information leakage about a secret from the leakage about how users generate their secrets, i.e., their strategy. Such an approach can provide insight in situations when secrets change according to a strategy (e.g., location based on commuting patterns), or when multiple secrets are generated according to a strategy (e.g., passwords on different web sites), among others.