MUSKETEER · Project

Privacy-Preserving Machine Learning That Lets Companies Share Data Without Exposing It

digitalTestedTRL 6

Imagine several hospitals or factories each sitting on valuable data, but none of them can share it because of privacy rules or competitive concerns. MUSKETEER built a platform where machine learning models travel to the data instead of the data traveling to a central place — so everyone benefits from collective insights without anyone seeing anyone else's raw information. Think of it like a group study session where everyone learns the answers but nobody shows their notes. The platform was tested in both smart manufacturing and healthcare settings with real industrial data.

By the numbers

EUR 4,380,335

EU funding for platform development

consortium partners involved

countries represented in consortium

industrial scenarios validated (manufacturing and health)

total project deliverables produced

demo deliverables with working software

64%

industry ratio in consortium

The business problem

What needed solving

Companies sitting on valuable data cannot share it with partners, suppliers, or even their own subsidiaries due to privacy regulations, competitive concerns, and security risks. This means machine learning models are trained on limited, siloed datasets — producing weaker insights than what collective data could deliver. The result: missed optimization opportunities, duplicated R&D efforts, and an inability to build industry-wide intelligence for problems like fraud detection, quality control, or clinical outcomes.

The solution

What was built

MUSKETEER built a complete federated machine learning platform with client connectors, multiple privacy operating modes (from encrypted centralized processing to fully distributed learning where data never leaves client premises), pre-processing and data alignment libraries, data value estimation algorithms, and a dashboard for monitoring. The final prototype demonstrated end-to-end execution across all use cases and all privacy modes.

Audience

Who needs this

Hospital networks needing to train AI models across patient data from multiple sites without violating GDPRManufacturers wanting cross-plant machine learning for quality control without exposing proprietary production dataBanks and insurers building shared fraud detection models without sharing customer recordsPharmaceutical companies running federated clinical trial analysis across multiple countriesIndustry consortia wanting to pool data for benchmarking while keeping individual contributions confidential

Business applications

Who can put this to work

Healthcare & Pharmaceuticals

enterprise

Target: Hospital networks or pharma companies needing multi-site clinical data analysis

If you are a hospital network or pharmaceutical company struggling to pool patient data across sites due to GDPR and patient privacy rules — this project developed a federated machine learning platform that trains predictive models across distributed datasets without any patient record ever leaving its original hospital. The platform was validated in an industrial health scenario with 11 consortium partners across 7 countries.

Smart Manufacturing

enterprise

Target: Manufacturers wanting to benchmark or optimize across multiple plant locations

If you are a manufacturer running multiple production lines and want to use machine learning for quality control or predictive maintenance but cannot centralize sensitive production data — this project built a platform with privacy-preserving algorithms that learn patterns across all your plants without exposing proprietary process data. MUSKETEER was specifically validated in the smart manufacturing scenario as one of 2 industrial use cases.

Financial Services & Insurance

enterprise

Target: Banks or insurers needing cross-institutional fraud detection or risk modeling

If you are a bank or insurance company that would benefit from shared fraud detection models but cannot legally share customer transaction data with competitors — this project created federated learning algorithms operating across encrypted and distributed data. The platform supports multiple privacy operating modes so you can choose the level of data protection that matches your regulatory requirements.

Frequently asked

Quick answers

What would it cost to adopt this technology?

The MUSKETEER platform was developed as an open research output with EUR 4,380,335 in EU funding across 11 partners. The platform code and connectors were delivered as software artifacts. Licensing terms would need to be discussed with IBM Ireland (coordinator) or other consortium members — costs would depend on deployment scope and support needs.

Can this scale to handle enterprise-level data volumes?

The project explicitly targeted scalability and efficiency for real use cases. The final prototype demonstrated end-to-end execution across all privacy operating modes and all use cases. However, production-scale deployment beyond the 2 validated industrial scenarios would require further engineering.

Who owns the intellectual property and how is it licensed?

IBM Ireland coordinated the project with 7 industry partners and 3 universities across 7 countries. IP ownership follows the Horizon 2020 grant agreement, typically shared among contributing partners. Contact the coordinator to discuss licensing for specific components like the ML algorithms or platform connectors.

Has this been tested with real industrial data or just lab experiments?

MUSKETEER was validated in 2 different industrial scenarios: smart manufacturing and health. The final platform prototype demonstrated complete end-to-end execution of data sharing and federated machine learning across all use cases and all privacy operating modes, with dashboard reporting included.

How does this handle different privacy regulations across countries?

The platform supports multiple privacy operating modes (POM1 through POM3 and beyond), ranging from scenarios where encrypted data is processed centrally to fully federated modes where datasets never leave each client's premises. This flexibility lets you match the privacy mode to your specific regulatory requirements, whether GDPR, sector-specific rules, or internal policies.

How long would integration take for an existing IT system?

MUSKETEER delivered dedicated client connectors (both first and final prototype versions) designed to link existing data systems to the platform. The 3-year project produced 28 deliverables including pre-processing, normalization, and data alignment tools. Integration timelines depend on your data infrastructure, but the connector architecture was built for interoperability.

Is there ongoing support or has the project ended?

The project closed in November 2021 after 3 years of development. The project website musketeer.eu may still host documentation and outputs. For ongoing support or commercial arrangements, contact IBM Ireland or check which consortium partners continue development of the technology.

Consortium

Who built it

The MUSKETEER consortium is led by IBM Ireland — a global technology leader with deep expertise in enterprise AI and data platforms — which adds significant credibility to the technology's commercial potential. With 11 partners across 7 countries (Belgium, Germany, Greece, Spain, Ireland, Italy, UK), the project has strong European coverage. The 64% industry ratio (7 out of 11 partners from industry) signals this was built with real-world deployment in mind, not just academic research. The mix includes 3 universities providing research depth and 2 SMEs bringing agility. This industry-heavy composition means the platform was shaped by organizations that understand enterprise data challenges, making it more likely to address actual business pain points around privacy-preserving analytics.

IBM IRELAND LIMITEDCoordinator · IE
TREE TECHNOLOGY SAparticipant · ES
IMPERIAL COLLEGE OF SCIENCE TECHNOLOGY AND MEDICINEparticipant · UK
INTERNATIONAL DATA SPACES EVparticipant · DE
DIAGNOSTIKON KAI THERAPEFTIKON KENTRON ATHINON YGEIA ANONYMOS ETAIREIAparticipant · EL
BIOTRONICS 3D LIMITEDparticipant · UK
ENGINEERING - INGEGNERIA INFORMATICA SPAparticipant · IT
COMAU SPAparticipant · IT
KATHOLIEKE UNIVERSITEIT LEUVENparticipant · BE
UNIVERSIDAD CARLOS III DE MADRIDparticipant · ES

How to reach the team

IBM Ireland Limited coordinated MUSKETEER. Reach out through the project website or IBM's research division for licensing and partnership inquiries.

Order a report on this consortium See IBM IRELAND LIMITED profile →

Next steps

Talk to the team behind this work.

Want to explore how MUSKETEER's federated learning platform could solve your data sharing challenges? SciTransfer can connect you directly with the right consortium partner for your use case.

Order a report on this topic More in Digital & ICT