In many applications (such as access control or internet privacy) it is important to know whether a specific natural person (e.g., a person registered to be living in Dresden) can uniquely be identified from information that is known about an anonymous individual (e.g., someone posting information in a social network). In addition to factual information about natural persons and anonymous individuals, also background information about the world (e.g., that a person can have only one mother and one residential address) can be used in this identification process. In order to formalize this problem as a logical inference problem, we assume that the available background information is expressed in an ontology (more formally, a Description Logic TBox) and that the information about persons and anonymous individuals are stored in a fact base (i.e., a Description Logic ABox). The research challenge is then (i) to see which formal inference problem corresponds to our identification problem; (ii) to find out which expressive power is needed to represent the necessary information; and (iii) to ensure practical answer times given that the ABox may represent information about a large number of persons.

In the simple form described above, the problem can be solved probably be solved using traditional Description Logic reasoning. In a more realistic setting, one also needs to take into account that the information about the natural persons and the anonymous individuals is known to hold only with a certain probability and that it may be of a temporal nature. This raises the additional challenges of extending traditional Description Logic reasoning to (iv) probabilistic and (v) temporal reasoning. Finally, one further advanced challenge for this problem is to consider (vi) context-based reasoning in both facts and background information may depend on the current context.