Collaborative Scheduling: Threats and Promises

For people to entrust their private information to modern collaborative tools, they will want some assurance of the extent to which this information will be protected. Accurately measuring the amount of privacy lost by different collaborative technologies is a critical step in providing such assurances. Unfortunately, the identification of intuitive and useful metrics for privacy loss has proven difficult. In this paper, we analyze the problems with the existing metrics for distributed constraint optimization, and based on our findings propose a new metric for measuring privacy loss that avoids these problems. In particular, our metric quantifies the amount of private information definitively lost when given an economic adversarial model. We show the importance of a defined threat model in measuring privacy loss and the dangers of metrics based on too much uncertainty. We develop and demonstrate our metric for the specific problem of collaborative meeting scheduling and use it to show the effects on privacy of network topology in distributed constraint optimization.