Disputation: He Gu

The PhD defence will be partially digital, in Kristen Nygaards sal (5370), Ole Johan Dahls hus, and streamed directly using Zoom. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.

Ex auditorio questions: the chair of the defence will invite the attending audience at Kristen Nygaards sal (5370), Ole Johan Dahls hus to ask ex auditorio questions.

Join the disputation

Download Zoom

Trial lecture

“Threats to Privacy-Preserving Machine Learning: Inference Attacks, Model Leakage and Mitigation Techniques”

Time and place: June 29, 2026 10:15 AM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus/ Zoom.

Main research findings

Internet of Things and data-driven applications have become increasingly ubiquitous, raising great concerns in potential privacy disclosures in data streams. This makes privacy protection one of the most essential research topics. However, protecting privacy means to conceal private information, which may also distort other useful information and reduce data utility. Therefore, this thesis presents novel approaches that protect privacy in data streams with less data distortion and higher utility than existing research. The privacy protection is based on Differential Privacy – a de facto standard of privacy protection in data management. Our approaches have the potential to surpass other existing solutions because we only focus on protecting private patterns that may reveal private information instead of protecting all data. We also present a semi-automatic approach which detects private patterns in databases or data streams. This approach can be implemented with only marginal privacy expertise, or even with no privacy expertise if a generally accepted regulation/law is appointed to be followed, e.g., The General Data Protection Regulation (GDPR). This approach simplifies the procedure of privacy-aware data-sharing in both human life and research. All our approaches are evaluated with public and synthetic datasets and are proven better than existing solutions for these datasets.

Adjudication committee

• Professor David Michael Eyers, University of Otago, School of Computing, New Zealand
• Professor Edmundo Monteiro, University of Coimbra, Dept. of Informatics Engineering, Portugal
• Professor Martin Giese, Department of Informatics, University of Oslo, Norway

Supervisors

• Professor Vera Hermine Goebel, Department of Informatics, University of Oslo, Norway

• Professor Boris Koldehofe, Marburg University, Germany

Chair of defence

• Silvia Lizeth Tapia Tarifa, Department of Informatics, University of Oslo, Norway

Contact information at Department: Pernille Adine Nordby