Personal Attributes and Privacy

The aggregation of personal attributes into user profiles is a significant privacy concern. Existing attribute management systems support the controlled release of attributes and unlinkability between session protocols, but do not address the problem that attributes distributed in this way may be data-mined for features that allow user profiles to be reconstructed. This paper identifies the aggregation problem as the missing element in the protection of personal attributes, and introduces attribute management principles that are sufficient to provide an overall framework to protect users from profile aggregation. The principles are clarified by formalizing them as constraints on primitive operations in a service-based architecture, and this analysis is the basis for a proof that they support system wide privacy. These results are of particular value in the decomposition of business processes into services, and a location-privacy problem is used to show how they can be applied in practice.