Distance-preserving pseudonymization for timestamps and spatial data

The need for privacy in intrusion detection data, such as audit logs is widely recognized. The prevalent method for privacy protection in audit logs is pseudonymization (and suppression). There is a clear trade-off between the privacy of a pseudonymization technique and its utility for intrusion detection. E.g., for IP addresses a method for prefix preserving pseudonymization has been developed, that allows pseudonymized IP addresses to be still grouped into subnets. This paper describes a pseudonymization technique for timestamps that is distance preserving. I.e. given two pseudonymized timestamps one can compute the distance δ, if d is below or equal to an agreed threshold d and one cannot compute δif δ = 2d. We extend our technique for two dimensional spatial data, e.g. location of objects or persons. We also evaluate the privacy any such distance-preserving technique can provide for timestamps theoretically and on real-world log data.