Pseudonymizing Unix Log Files

Abstract

Unix systems in many cases record personal data in log files. We present tools that help in practice to retrofit privacy protection into existing Unix audit systems. Our tools are based on an approach to pseudonymizing Unix log files while balancing user requirements for anonymity and the service provider's requirements for accountability. By pseudonymizing identifying data in log files the association between the data and the real persons is hidden. Only upon good cause shown, such as a proceeding attack scenario, the identifying data behind the pseudonyms can be revealed. We develop a trust model as well as an architecture that integrates seamlessly with existing Unix systems. Finally, we provide performance measurements demonstrating that the tools are sufficiently fast for use at large sites.