Weaving privacy and mechanism design

Abstract

Today's information regime allows rapid collection of data from phones, social networks, internet activity etc., providing industry, government, and researchers with information about human behavior in scale that were previously unimaginable. This use of individual information introduces a threat to the privacy of individuals, pronouncing the problem of balancing between utility extracted from sensitive data and loss of privacy.

We will review some of the recent research towards modelling privacy-aware agents and constructing mechanisms for them. Two main directions have emerged in this literature. In one, individuals are compensated monetarily for their loss of privacy; In the other, individuals have a stake in the outcome, and their influence of the outcome is balanced with their loss of privacy, without resorting to monetary transfers. In both research directions, the notion of differential privacy plays a central role in the modelling of privacy loss and integrating it into the agents' utility functions, and in the construction of privacy-aware mechanisms.

The talk will be self-contained - no background knowledge in game theory or differential privacy will be assumed.