MPC company Roseman Labs puts academic results from PRIViLEDGE into practice

blog

TU Eindhoven researchers launched MPC company Roseman Labs during Horizon2020 program PRIViLEDGE, valorising the project’s open access research output.

“Anyone who comes into contact with data will soon grapple with a privacy problem, especially if you want to analyze data from different sources. This prospect makes collaborations within industries difficult. Companies and institutions do not dare to pool data because it can end up in the wrong hands.” wrote IT magazine Computable, who recently featured Roseman Labs (website, article) and summarized what the company is planning to solve.

Founded in March 2020 and headquartered in The Netherlands, Roseman Labs offers Cranmera, a general-purpose software engine for secure multi-party computation, or MPC. MPC is a cryptographic technique that enables organizations to perform computations on the combination of their sensitive data sets, without actually revealing those data sets to each other; only the result of the computation becomes known.

Niek Bouman, who holds a PhD from CWI in quantum cryptography and is a former postdoc from TU Eindhoven (TUE), teamed up with Toon Segers, a former strategy consultant and currently a cryptography PhD-candidate at TUE, and Roderick Rodenburg, a tech entrepreneur. Shortly after launch, the trio invited Berry Schoenmakers, an associate professor in cryptography at TUE, to their advisory board.  Schoenmakers, who is known in the international cryptography community for his contributions to Sigma-protocols, e-voting and publicly verifiable secret sharing among other things, currently works on the open-source MPC framework MPyC (github). MPyC was inspired by the older VIFF framework (Geissler et al.) and offers a rich and flexible MPC programming environment, which allows rapid deployment of a variety of use-cases; it also offers Roseman Labs a blueprint for parts of its privacy engine.  









Toon Segers (on the left), PRIViLEDGE researcher and CPO at Roseman Labs and Berry Schoenmakers (on the right), Associate Prof., PRIViLEDGE researcher partner and Advisory Board member of Roseman Labs.

During the first year of the company, results from the PRIViLEDGE project became directly relevant for Roseman Labs. For example, one of its clients was looking to conduct wage gap surveys like the one the Boston Women’s Workforce Council conducts already for several years in the wider Boston area, using MPC (link). The non-profit needed a performant linear regression technique in MPC, which Roseman Labs could immediately offer based on PRIViLEDGE sponsored research on ‘Efficient Secure Ridge Regression’ from 2019 (eprint link).

Another Roseman Labs client requires storage of data between secure computations. There are several techniques for doing this, but one of the directions the company considers is using the so-called ‘Secure Groups’ scheme by Segers and Schoenmakers to facilitate threshold cryptography within the MPC context. The work-in-progress Secure Groups scheme is a direct product of the PRIViLEDGE project and featured in Deliverable 2. See D2.3, Chapter 7 for a preliminary draft. The final version is planned for D2.4.

“The demand for MPC is real” according to Roderick Rodenburg. “In our first year, we successfully delivered projects for the Dutch National Cyber Security Centre, part of the Ministry of Justice, and for the national grid operator Stedin. They are both excited that we are able to deploy MPC rapidly and robustly in an enterprise environment.”

If you are interested in learning more about MPC for enterprise applications and the role H2020 projects can play in valorising research, please do not hesitate to reach out to Toon Segers, TUE member of the PRIViLEDGE consortium, or Liis Livin, head of communications for PRIViLEDGE. 


Written by A.J. M (Toon) Segers, The Eindhoven University of Technology.
Photos by Roseman Labs.


Links:

Roseman Labs website: https://rosemanlabs.com/  
MPyC: https://github.com/lschoe/mpyc
PRIViLEDGE: https://priviledge-project.eu/