Enhancing Trusted Transatlantic Data Processing and Storage with Fully Homomorphic Encryption
Coach Name
Jordi Bosch i Garcia
EU Organization
Random Red (Croatia)
Members
- Kruno Miličević
- Davor Vinko
- Ivan Uglik
- Adrijan Djurin
US Organization
Penn State University – Beaver Campus (USA)
Members
- Richard Lomotey
- Madhurima Ray
Project Overview
TruPS explored the use of Fully Homomorphic Encryption (FHE) to enable privacy-preserving machine learning and data analytics across borders. By allowing computations directly on encrypted data, the project demonstrated that organisations can process sensitive information without ever decrypting it—ensuring confidentiality, security, and compliance with both EU (GDPR, eIDAS) and US (HIPAA, FISMA, GLBA) data protection frameworks.
The team built a functional prototype cloud service for encrypted inference on various machine learning models and benchmarked its performance against plaintext equivalents, achieving similar accuracy (~1% deviation) while maintaining complete data protection.
TruPS thus offers a powerful foundation for secure transatlantic collaboration—where data can remain encrypted “everywhere”, closing the last gap in cross-border data protection.
Methods and approaches
Fully Homomorphic Encryption (FHE) for Privacy-Preserving ML
TruPS implemented a cloud-based prototype that enables computations on encrypted data without decryption. Using the open-source Concrete-ML (TFHE-based) framework, the team integrated FHE into standard machine learning pipelines, proving that accurate predictions can be made while data remains encrypted end-to-end.
Legal–Technical Integration for Compliance-by-Design
The project combined cryptographic engineering with legal analysis of EU and US data protection frameworks (GDPR, HIPAA, FISMA). This interdisciplinary approach ensured that the system’s architecture not only delivered strong technical safeguards but also aligned with transatlantic regulatory requirements for encryption, data minimization, and security-by-design.
Key Achievements
Prototype Development and Deployment
Developed and deployed the TruPS prototype platform for privacy-preserving ML using FHE (trups.app)
Open-Source Code Publication
Published an open-source repository with code and documentation (GitHub)
Scientific and White Paper Releases
Released a white paper and two scientific papers integrating technical and legal perspectives on FHE
Public Dissemination and Events
Conducted a public webinar (May 2025) and presented results at Future Labs Live (Basel)
Community Engagement and User Validation
Engaged with the open-source community (Zama’s Concrete-ML) and conducted 45 stakeholder interviews for user validation
Compliance-by-Design Demonstration
Demonstrated compliance-by-design for transatlantic data sharing under GDPR and HIPAA
Impact & Results
Scientific Impact
TruPS advances scientific knowledge in privacy-preserving AI, demonstrating the practical use of FHE in real-world machine learning.
Economic Impact
It has economic potential for sectors like healthcare, finance, and cloud computing that need secure data collaboration across jurisdictions.
Social Impact
On a societal level, TruPS empowers data-driven innovation without compromising privacy, reinforcing public trust and transatlantic cooperation.
Publications and Open-Source Contributions
- White Paper: “TruPS – Enhancing Trusted Transatlantic Data Processing and Storage with Fully Homomorphic Encryption”
- Scientific Papers:
- Evaluation of Sector-Specific EU and US Regulations for Fully Homomorphic Encryption (under review)
- Security Overview and Experimental Performance Assessment of Using Fully Homomorphic Encryption on Machine Learning Algorithms (under review)
- Open-source Repository: https://github.com/RandomRedLtd/trups-public
- Webinar recording: youtu.be/xl_hTojuiMM
Future directions
Random Red plans to commercialize TruPS as part of its service portfolio, improving FHE performance and usability for enterprise adoption. The partnership with Penn State University will continue to evolve through joint R&D and potential follow-up projects.