Data plays an increasingly important role in how organizations make decisions. However, most data-driven organizations realize that their own data is not enough, and that they need access to more complete information in order to build an accurate worldview and make sound data-backed decisions.
Unfortunately, setting up collaborations around sensitive or confidential data is cumbersome and organizations are reluctant to collaborate with others because of risks of data leaks or because of data privacy concerns and regulations. A paradigmatic example of this issue is personalized medicine, where doctors have to make decisions on treatments based on prior patients of the same hospital/clinic, or on third party studies that are based on different populations that might not match the characteristics of a given patient. Normally, optimal treatments can only be found if the doctors have access to a large pool of evidence from similar patients and different treatments; this information can only be accessed through multi-site collaborations that involve personal patient information from several clinical sites, potentially across different regulatory frameworks, and are therefore difficult and time-consuming to set up and get approved.
Tune Insight B2B software provides infrastructure-agnostic solutions to enable secure collaborative analytics and machine learning in federated scenarios between several organizations, enabling them to extract collective insights from confidential data from the whole network of participating organizations, while they remain in full control of their own data. Our software can be deployed on-premise or in cloud instances, and it avoids any data transfer while keeping all data encrypted in transit and during computation, thanks to a unique combination of cryptographic techniques called multiparty homomorphic encryption.
This technology has a transformative potential to data collaborations, enabling business models and opportunities for competitive and collaborative relationships that could not take place without our secure framework and the possibility of “sharing insight without sharing data”. This brings a real disruption to data-centric organizations, accelerating and streamlining their data pipelines and enabling them to tap into larger collective data pools to extract valuable and actionable insights that can effectively improve their business decisions.
All this is possible thanks to the technological guarantees that ease compliance with data protection regulations and with the strictest data management and security policies, removing the need for complex data transfer agreements, lengthy ethics approvals, or data use policies, as they are already embedded in the technology itself.
This project combines this technology with advanced machine learning capabilities, which can exponentially increase the impact of these solutions, enabling deeper workflows and more advanced pipelines that can effectively cope with complex data models and extract the full potential of the underlying data.
This includes but is not limited to health, with better and more precise personalized care in hospitals and clinics and faster drug development in pharma, that will improve the overall quality of care and the life expectancy for society in general; financial services and insurance, with improved risk estimations and more accurate fraud detection enabling reduced losses and more precise risk-based pricing calculation, which result in overall better services for corporate and individual customers; cybersecurity and cyberdefense, with more effective and constantly updated defense mechanisms that use the full potential of extremely confidential vulnerability and incident data to timely respond to the ever-growing challenge of new cyberthreats.
The project will start by studying, analyzing, and prioritizing the advanced ML classifiers and models that should be supported on top of Multiparty Homomorphic Encryption (MHE). This process will based on early market feedback from Tune Insight customers and partners, and on the performance and scalability of the workflow implementation within our MHE framework.
The next steps involve designing functional approximations and optimized secure protocols based on the MHE secure Map-Reduce paradigm, that realize the prioritized advanced ML functionalities, that will be subsequently tested and validated with end users (e.g., data analysts and data scientists, physicians, bioinformaticians, security operators) in realistic business cases (e.g., personalized medicine, cyberintelligence, insurance and manufacturing).
Persons involved in the project
Last update to this project presentation 17.05.2022