Skip to main content

Distributed/Federated data analytics allow to improve the utility of data from multiple sources through various techniques, ranging from statistics to machine learning. For example, collaborative cybersecurity systems can offer significant advantages over the local systems by detecting cyberattacks early. In such use cases, since data belong to different data sources and are considered as being personal or business confidential data, confidentiality becomes a mandatory requirement. Traditional data encryption solutions unfortunately do not appear to be very conducive to such an advanced federated data sharing technology since they prevent parties to operate over encrypted data. While recent developments in cryptography and the rise of homomorphic encryption and secure multi-party computation allow operations without revealing the underlying data in cleartext, these technologies still suffer from a noticeable overhead in terms of computation and communication. The ambition of PATTERN is therefore to enhance existing, and develop novel cryptographic mechanisms that allow for federated computations on sensitive data in an efficient and scalable manner, while at the same time also giving integrity guarantees on the obtained computations. The privacy and integrity tools developed in PATTERN will be prototyped and integrated in two real-world use cases: i) a mail analysis platform where PATTERN solutions will be used to perform some analysis on multiple servers without the disclosure of any sensitive data. Analytics will range from simple statistics such as counting to some machine learning inference schemes such as decision trees, support vector machines or neural networks;(ii) a cybersecurity platform that will leverage privacy-preserving federated machine learning inference solutions in order to enable several parties to collaborate in providing an ensemble model more robust for the detection of infected files.

Call Topic: Security and Privacy in Decentralised and Distributed Systems (SPiDDS), Call 2022
Start date: (36 months)
Funding support: 875708 €

Project partners

  • EURECOM - France (coordinator)
  • ORANGE SA - France
  • Luxembourg Institute of Science and Technology - Luxembourg
  • Ruxandra Olimid - Romania