PLI-Assess: A Behavior Profile-Based Approach for Privacy-Preserving Log Assessment

Event logs of business processes provide a valuable starting point for process mining, and high-quality event logs can significantly enhance the quality of process mining. However, event logs often contain a substantial amount of sensitive and personal information. Therefore, the release of event logs should prioritize the model’s quality while minimizing the risk of privacy exposure. Specifically, quantifying performance indicators between the original event logs and the released ones enables the operational goals. To date, privacy benefit and utility loss are two main target performance indicators, especially from the perspective of structural similarity comparison of mined process models. To the best of our knowledge, no study aims to measure the privacy-preserving performance indicators from the point of behavior differentiation between the original event logs and released ones. In this paper, we propose an approach to quantify the behavior differentiation between the original event logs and the corresponding released ones. Specifically, an approach of event log release mechanism that effectively combines behavior privacy gain and behavior utility loss is presented in this paper. Firstly, we discuss challenges in scenarios where event data is released without privacy preservation, and describe potential attacks that could occur when third-party businesses perform process mining techniques. Based on these potential attacks, we present a behavior differentiation-based event log release mechanism named PLI-Assess to combat these threats. Finally, we conduct experiments on four groups of practical event logs for comparisons with the baseline methods.The experimental results suggest feasibility of privacy-utility trade-offs.