In recent years, it has become important for researchers, security

incident responders and educators to share network

logs, and many log anonymization tools and techniques have

been put forth to sanitize this sensitive data source in order

to enable more collaboration. Unfortunately, many new

attacks have been created, in parallel, that try to exploit

weaknesses in the anonymization process. In this paper, we

present a taxonomy that relates similar kinds of attacks in

a meaningful way. We also present a new adversarial model

which we can map into the taxonomy by the types of attacks

that can be perpetrated by a particular adversary. This has

helped us to negotiate the trade-offs between data utility

and trust, by giving us a way to specify the strength of an

anonymization scheme as a measure of the types of adversaries

it protects against.