Statistical offices collect and use massive volumes of highly valuable survey and administrative data to fulfil their public service mission. These data underpin the production of official statistics and supports policy and scientific research across various sectors. To fully utilize the immense potential for research and increase the societal value of these data, statistical organizations have started to explore new data partnership models such as novel data dissemination services, (distributed) research hubs, and Data Spaces.
However, privacy concerns loom large over the landscape of data partnerships. In the last five years there has been rapid development and evolution of methodologies and approaches to mitigating privacy risks when using sensitive or confidential data, which are collectively referred to as privacy-enhancing technologies (PETs). Secure multi-party computation, homomorphic encryption, differential privacy, synthetic data, distributed learning, trusted execution environments and other technologies applicable to various data partnership scenarios (for example in centralized or distributed settings) offer the promise of driving responsible innovation – unlocking data analysis and research across sectors, whilst upholding the highest standards of data privacy and security.
Successful implementation of PETs in new data partnership scenarios will require much more than technology. Some challenges are methodological and can be solved by experimentation with different PET techniques across different use cases. But there are other important aspects – such as legal and policy considerations, organizational capacity, collaboration with partners, engagement with users and culture (gap) – that must also be addressed.
The session will be divided into three sections starting with a keynote-like presentation that will introduce the importance of data partnerships in official statistics, in the context of privacy concerns in today’s data-driven world, and offer holistic insights into how statisticians can steer data partnerships towards privacy-first, ethically sound and trustworthy practices.
This will be followed by three presentations delving into different aspects of PETs implementation: data governance and balancing utility and privacy; legal and policy aspects including confidentiality, integrity, data minimization, accuracy, and other types of privacy law obligations; and the role of PETs in Data Spaces, presenting a framework for controlled and interoperable data sharing.
Finally, the case studies part of this session will feature a panel discussion with panelists from national statistical organizations (NSOs) who participate in real-world projects that are investigating and implementing PETs in various data partnership scenarios. Examples include a platform that enhances access to confidential data by enabling authorized users to securely analyze it without ever seeing the original information; an experimental protocol for private linkage of sensitive international datasets in a cross-NSO collaboration scenario; and testing PET methods in the health data setting through a collaboration between an NSO and a public health agency. This discussion will be moderated by a discussant with the opportunity for questions from the audience.
Speakers and panelists are experts working in statistical organizations, academia and private sector and are involved in national and international collaboration activities such as National Innovation Centre for PETs in the Netherlands, the UN PET task team and the UN PET Lab community of practice.