Humboldt-Universität zu Berlin - Mathematisch-Naturwissenschaftliche Fakultät - Institut für Informatik

Abstract:
Process Mining is an emerging subfield of data mining that focuses on the data-driven analysis of business processes. It uses event data recorded while executing a business process, where each execution of an activity is captured by an event. A sequence of events, referred to as a trace, captures the behavior of a single process instance. However, traces may reveal sensitive information about individuals such as patients, customers, or process workers. Anonymizing traces is challenging, as behavioral characteristics need to be preserved for process analysis. In this talk, we will present techniques to address this challenge and introduce novel algorithms that ensure established privacy guarantees, such as k-anonymity and differential privacy, for event logs, while also providing utility improvements over the state of the art.