Privacy-Preserving Data Publishing in Process Mining
This project implements a web-based application for Privacy-Preserving Data Publishing (PPDP) in process mining. The application is witten in Python using Django framework.
At the moment, the application has four main modules:
Event data management
In this module, you can upload and manage your event data. Both standard XES event logs and non-standard event data, called Event Log Abstraction (ELA) resulting from some privacy preservation techniques can be handeled in this module.
This module implements the decomposition method proposed in the paper Mining Roles From Event Logs While Preserving Privacy. The result of applying this technique to an event log is another event log which preserves the data utility for mining roles (similar task social networks) based on the resource and activity information without revealing who performs what.
This module implements the TLKC-privacy model proposed in the paper TLKC-Privacy Model for Process Mining. This privacy model provides group-based privacy guarantees assuming four types of background knowledge: set, multiset, sequence, and relative. T refers to the accuracy of timestamps in the privacy-aware event log, L refers to the power of background knowledge, K refers to the k in the k-anonymity privacy model, and C refers to the bound of confidence regarding the sensitive attribute values in an equivalence class. Applying this method results in a privacy-aware event log in the XES format that preserves data utility for process discovery and performance analysis.
Privacy metadata are also embedded into the developed privacy preservation techniques.
Each privacy preservation technique in the tool is implemented as a Django application that enables the simultaneous running of different techniques on an event log.
New techniques can simply be integrated as independent applications.
The outputs for the privacy preservation techniques are provided independently for each technique and can be downloaded or stored in the event data repository.
The tool is designed in a way that provides a cycle of privacy preservation techniques, i.e., the privacy-aware event data, added to the event data repository, can be set as the input for the techniques again as long as they are in the form of standard XES event logs.
To keep the process analysts aware of the modifications applied to the privacy-aware event logs, the privacy metadata specify the order of the applied privacy preservation techniques.
A naming approach is followed to uniquely identify the privacy-aware event data based on name of the technique, the creation time, and name of the event log.
In this video, you can watch a screencast of the tool which demonstrates the main functionalities of our Python-based infrastructure for privacy preservation in process mining. Note that you may need to download the file to watch.
To simplify the usage, and to run the appication without going throgh the installation phase, a Docker container has been provide that can be run on your local system using the following docker commands:
docker pull m4jid/ppdp-pm
docker run -d -p 8000:8000 m4jid/ppdp-pm
Note that for using docker commands, first you need to install Docker accourding to your operation system.
After running the docker, use your browser and enter the following address to run the web-based application:
http://127.0.0.1:8000/
The introduced privacy preservation techniques have also been integrated into PM4Py-WS (PMTK) as an open-source web-based
application for process mining. Where process mining algorithms can directly be applied to the privacy-aware event logs. Use the following docker commands to run this application:
docker pull m4jid/pm4pyws:privacyIntegration
docker run -d -p 5000:80 m4jid/pm4pyws:privacyIntegration
After running the docker, use your browser and enter the following address to run the web-based application:
http://127.0.0.1:5000/index.html