8. June 2022
The project is also represented at this year’s IFIP SEC conference by Christian Burkert with the paper “Data Minimisation Potential for Timestamps in Git: An Empirical Analysis of User Configurations”.
The paper was written with co-authors Johanna Ansohn McDougall and Hannes Federrath. It will be presented on the second day of the conference by Christian Burkert (see programme). The authors want to get closer to the question of what users of DevOps tools expect in terms of precision from the timestamps stored there. A better understanding of these expectations and needs is a prerequisite for consistent data minimisation, also for timestamps, where timestamps should only be recorded and stored as detailed as is necessary for the purposes of the processing. However, which purposes these are and what is necessary for them is as yet unexplored and presumably strongly dependent on the working method and individual preferences.
Using an extensive dataset of over 20,000 configuration files of the Git tool, the authors evaluated the precision needs of users,
as far as they are reflected in Git’s configuration options.
This revealed that a large proportion of users actually make use of settings with reduced precision.
For example, it turned out that around 90% of all output formatting for git log and related commands use reduced or adaptive precision rather than the default dates collected in seconds precision.
Reducing a timestamp from seconds precision to, say, 5 minutes precision experimentally reduces a user’s distinguishable activity dates to almost 50% (for the median user) in an analysis of over 360 million real Git commits.
The authors see the study results as a sign that data minimisation is useful and possible even with timestamps, and even if the purpose is to inform the user.
The paper is available in the Proceedings of the conference and as Manuscript version.