Brief reminder, this talk is on Wednesday.
--- begin forwarded message ---
From: Christian Cachin christian.cachin@unibe.ch To: INF inf-staff@list.inf.unibe.ch, CRYPTO-Events crypto-events@list.inf.unibe.ch Cc: Judith Senn Judith.Senn@uibk.ac.at, aso@zurich.ibm.com, Alessandro Sorniotti ale@ftml.net Subject: Talk announcement: Privacy and the Detection of Money Laundering, by Judith Senn (1 July 2026) Date: Tue, 16 Jun 2026 12:02:53 +0200
*Date and Location* 1 July 2026, *15:00* Universität Bern, Hauptgebäude, Hochschulstrasse 4, Seminarraum 206
*Speaker* Judith Senn Security and Privacy Lab Universität Innsbruck
*Title* Privacy and the Detection of Money Laundering
*Abstract* The discussion of payment systems is often framed as a polarised debate between privacy and enforcement against financial crime. While many anti-money laundering mechanisms depend on transaction data to identify suspicious behaviour, payment records can expose sensitive information about individuals and groups. Current detection methods often assume access to detailed transaction data, but it remains unclear which specific attributes are necessary for reliable detection.
We address this question experimentally by evaluating two graph-based detectors of money laundering, FraudGT and MEGA-GNN, under different privacy regimes for a common benchmark of transaction data. We compare how coarsening, concealment, and anonymisation of selected transaction attributes affect detection performance. The evaluation considers three settings: testing pre-trained models on privacy-modified data, fine-tuning pre-trained models on privacy-modified data, and training models directly under the corresponding privacy scenario. The results suggest that privacy modifications preserving the transaction graph are largely compatible with money laundering detection. In particular, modifying transaction amounts has little to no effect on detection performance, while timestamp transformations have limited influence. Conversely, modifications that impair the reconstruction of the transaction graph result in a performance drop. This pattern is consistent across both detectors.
*Brief bio* Judith Senn is a research assistant in the Security and Privacy Lab at the Universität Innsbruck. Her research focuses on privacy in digital payments and its impact on compliance and performance.
---
Please note the later talk by Alessandro Sorniotti at 16:00.
See you there!
Christian Cachin
--- Christian Cachin email: christian.cachin@unibe.ch Cryptology and Data Security Group web: crypto.unibe.ch/cc Institute of Computer Science tel: +41 31 684 8560 University of Bern Neubrückstrasse 10, CH-3012 Bern, Switzerland
--- end forwarded message ---
--- Christian Cachin email: christian.cachin@unibe.ch Cryptology and Data Security Group web: crypto.unibe.ch/cc Institute of Computer Science tel: +41 31 684 8560 University of Bern Neubrückstrasse 10, CH-3012 Bern, Switzerland