The TNO PET Lab is a cross-project initiative initiated to improve the overall quality, generality, and reusability in the development of Privacy-Enhancing Technologies (PETs) solutions developed in the numerous (past, ongoing, and future) TNO projects that involve different PETs. It consists of generic software components, procedures, and functionalities, grouped by technology (e.g. multi party computation, federated learning, ...); the code is maintained on a regular basis to facilitate and aid in the development of PETs solutions. The lab strives to facilitate and spread the use of privacy enhancing technologies, and decrease the time-to-market of new, innovative solutions.

PETs

Privacy-Enhancing Technologies are a family of technologies aimed to allow the collaboration between different parties of the sharing of information while protecting the privacy and securing personal or confidential data. Referring to the PET guide published by the United Nations in 2023, we can categorize PETs into:

  • Secure Multi-Party Computation: an umbrella term consisting of different cryptographic protocols allowing several parties to jointly compute a function while preserving the privacy of the input data;
  • Homomorphic Encryption: a cryptographic technology allows for computations on encrypted data;
  • Differential Privacy: a mathematical formulation of how much a method preserves and guarantees the privacy of the output;
  • Synthetic Data: a family of statistical of ML-based techniques aimed to generate artificial data that preserve the relevant statistical properties of the original data, without exposing any private information;
  • Distributed/Federated Learning: a family of protocols aiming to jointly train a ML model on data distributed among different parties, without the need of sharing or collecting the data;
  • Zero Knowledge proof: a class of cryptographic methods that allow one party to convince another party of the veracity of a statement, without exposing any of the (secrete) information used to demonstrate the statement;
  • Trusted Execution Environment: it is a secure area of modern CPUs; having code and data loaded inside this secure area helps protecting confidentiality and integrity.

TNO expertise and projects touch all of those branches, and focus also on how to combine different techniques, for example Federated Learning and MPC or synthetic data generation and Differential privacy.

TNO PET Lab GitHub

Our PET Lab codebase can be found here.

To GitHub

Open Source

We believe that open sourcing building blocks of several PETs serves the purpose of using in a more secure, privacy aware and responsible way confidential data. Moreover public and open source software allow to validate the theoretical/scientifical correctness of new methods and protocols as well as their implementations.

The published PETs building blocks and complete solutions can be found on GitHub. The mostly used license for TNO PET lab components is the Apache License, Version 2.0. This allows for easy adoption and flexible usage without enforcing a specific license to (end-)users and contributors of the codebase. We are always open to questions on, suggestions for and contributions to our codebase.

Open Source Code for Quality Evaluation of Synthetic Data

Introducing a tool to evaluate synthetic data.

Read more (pdf)

Federated survival analysis with Cox regression

We propose a federated survival analysis, so to facilitate collaboration among different medical centers, while preserving the sensitivity of patients data.

Read more

Bundling forces in money laundering detection using MPC

We examined how secured Multi-Party Computation allow banks to bundle forces and to fight money laundering collaboratively.

Read more

Privacy Enhancing Technologies in Practice

The big data era brought potential for a data-driven society, and led to a new market for PETs. In this article we explore the Dutch-scene of PETs in practice.

Read more

Predicting progression of medical status while preserving privacy

New insights into cancer are needed to help improving care and prevention. This requires broad and rich data, for instance to develop machine-learning models that can evaluate treatment outcomes.

Read more

Tool: explore privacy-enhancing technologies together

A public support tool for inspiring and facilitating multidisciplinary teams that are interested in applying PETs to their business challenges.

Read more

Secure and private statistics with distributed Paillier

We recently used distributed Paillier cryptography to do statistics on sensitive data with unparalleled security and privacy-preservation.

Read more

Identifying high-risk factors for diseases while preserving privacy

Multi-Party Computation (MPC) enables using more data from multiple sources to develop accurate models for health care predictions while preserving privacy.

Read more

Advanced data linking without breaching privacy

Linking distributed data while safeguarding privacy. An apparent contradiction. MPC technology shows that it can be done.

Read more

A targeted, yet privacy-friendly approach for battling poverty

Many citizens entitled to AIO provision are not using it. Multi-Party Computation (MPC) enables to proactive reach out to potential customers in a targeted way.

Read more