Audit-log integrity using redactable signatures with pseudonyms

In this paper we describe a new approach for the integrity of audit records. We show how to simultaneously establish the integrity of an entire audit data set and of any derived subsets, adapting techniques that have been used before for redactable signatures. In addition, our algorithms allow for the pseudonymization of data fields, cryptographically enforcing the consistency of chosen pseudonyms. The resulting schemes do not add significant computational overhead to a practical system, and are shown to be secure under reasonable cryptographic assumptions. We believe these algorithms can be a helpful tool to meet audit and reporting needs, in order to comply with such regulations as the US Sarbanes-Oxley Act (SOX). The algorithms enable proofs of the integrity of audit data and derived reports, while simultaneously providing means to protect privacy-sensitive information against internal and external consumers of these reports.