SwissAnon in Tokyo: Statistical Disclosure Control Meets Cryptography

In mid-August 2026, SwissAnon lead Prof. Dr. Matthias Templ (FHNW) traveled to Tokyo, where he presented SwissAnon’s work on statistical disclosure control (SDC) at two venues: a research seminar at the Center for Spatial Information Science (CSIS), University of Tokyo, and an informal research exchange and planning session with the leadership of Project SDC4Society.

SDC4Society is a Japanese national research initiative funded by the Japan Science and Technology Agency (JST) under its Key and Advanced Technology R&D through Cross Community Collaboration Program. It is led by Prof. Kazuhiro Minami, Vice Director-General and Professor at the Institute of Statistical Mathematics (ISM) in Tokyo, and combines SDC with advanced cryptography, including computation on encrypted data, to protect data across its full lifecycle: generation, transmission, storage and analysis. Four research groups contribute, based at the Research Organization of Information and Systems, Gunma University, Chuo University and Kyoto Tachibana University.

The session, on August 15 near Tokyo Station, brought together Minami, Prof. Shinsuke Ito of Chuo University and Dr. Chanh Minh Tran alongside Templ. It was small and informal enough that the talk and the meeting were really the same conversation. Each brought a short research update. Ito, whose recent work examines how differential privacy can be adapted to official statistics, presented on applying it to the Japanese population census. Tran presented on estimating data utility through the predictive performance of downstream models, the same logic behind the train-on-synthetic, test-on-real (TSTR) family of utility metrics.

Templ’s own update walked through SwissAnon’s R toolbox for SDC — sdcMicro, simPop, the new design-aware synthesizer synvey, and riskutility (home of RAPID, the Risk of Attribute Prediction-Induced Disclosure measure) — alongside the centre’s current applied projects: the SNF Bridge Discovery grant on longitudinal and trajectory privacy, an anonymization engagement with Nestlé, and SBB SynPop, a nine-million-person geo-referenced synthetic population. A second part turned to AI on both sides of the disclosure-control table: an assistant that proposes and refines anonymization strategies inside sdcMicro (a paper under review at SoftwareX), and an adversary — a team of AI agents that re-identified the large majority of participants who were publicly findable online from “anonymised” mobility traces, for a few dollars and minutes of effort per target (to appear at PSD 2026). The throughline ran across all three talks: empirical, adversarial testing shows where you actually stand; formal guarantees (differential privacy, cryptography) show where you are provably protected. The two need each other, which is exactly SDC4Society’s own bet.

Two concrete ideas came out of the conversation that followed: a funded research and teaching visit, most likely through a host-applied JSPS Invitational Fellowship, combining an intensive lecture series on SDC and synthetic data in R with joint time to draft a paper and a grant proposal; and a joint research project on SDC under agentic-AI adversaries, sized either for the bilateral SNSF–JSPS Strategic Japanese–Swiss Programme or, at larger scale, a Horizon Europe Pillar II consortium for the 2027 call now that Japan is fully associable.

It’s an early conversation, not a signed agreement — but SDC4Society’s cryptographic guarantees and SwissAnon’s empirical, adversarial risk measurement are an unusually good fit, and both sides left with concrete next steps rather than good intentions. The slides and a fuller project rundown are on the companion page at swissanon.com/sdc4society.