Insights and Broader Deliberations
The last entry in the collection that commenced with reflections and lessons from the initial US-UK cooperation involving Privacy Enhancing Technologies (PETs). Following the PETs Prize Challenges, the landscape encompassing these technologies has continuously evolved, progressing from more abstract and scholarly dialogs to increased acceptance and examination of PETs.
Since our maiden post in December of 2023, this collection has delved into various pragmatic considerations pertinent to working with Secrecy-Preserving Collaborative Learning (SPCL), exploring a spectrum of privacy intrusion techniques and methods to alleviate their hazards to individuals’ confidentiality, scrutinizing the significance of privacy violations and means to safeguarding entry and output privacy. Through recent posts, we also engaged with guest contributors who emerged victorious in the PETs Prize Challenges or served as judges. They put forth reflections and thoughts on aspects such as expansion, execution, data conduit hurdles, and more.
The expanse of this blog collection mirrors the wealth of insights and considerations spawned by the PETs Prize Challenges, yet there exist additional facets of dealing with SPCL that the challenges – and this series – have not tackled. This includes deliberations on dealing with actual data across numerous jurisdictions. SPCL can eliminate the necessity to shift, stock, or process real data centrally; a valuable trait when datasets must persist in a decentralized state (owing to policy or technical rationales). Nonetheless, we necessitate research to shape our comprehension of practical applications of SPCL and PETs in a universal context. NIST is delving into the explorations and intricacies of real-world SPCL deployments via the PETs Testbed, administered by the NIST National Cybersecurity Center of Excellence (NCCoE).
Prospective Partnership
As a rather fresh and burgeoning technique for managing data in a secrecy-driven manner, SPCL harbors the prospect of bolstering innovation and nurturing collaboration in the times ahead.
The UK and US have sanctioned further cooperative endeavors encompassing PETs. Building atop this allegiance, and leveraging insights from our antecedent collaboration, the UK National Disease Registration Service and the US National Cancer Institute are jointly exploring how PETs can catalyze inventive investigations into rare pediatric cancers through secure, confidentiality-preserving data alliance between national ailment registries. Harnessing PETs, researchers can execute cross-border data scrutiny sans the prerequisite for data relocation or direct admittance; they can extract deeper discernments sans jeopardizing data secrecy. This venture is being substantiated by the UK’s Department for Science, Innovation and Technology, the White House’s Office for Science and Technology Policy, NIST, and the US Department of Energy, along with synchronization with the National Science Foundation.
Collaboration utilizing PETs, akin to SPCL, will empower researchers to maneuver challenges stemming from the scarceness of data in distinct nations. Employing a federated modus operandi as a tool for data querying and/or model creation will authorize researchers to scrutinize data concerning uncommon pediatric cancers in a secrecy-maintaining manner. This will also empower researchers to assess information in approaches that were hitherto unattainable owing to the inadequate data availability (currently, no singular jurisdiction has access to a voluminous enough dataset encompassing ultra-rare tumor categories to independently conduct such scrutiny).
This approach also unveils the potential to extend research efforts in the longer run, to encompass additional data from other jurisdictions in the analysis. Escalating to encompass more nations could buttress a broader universal initiative to engender cooperation on pediatric cancer.
Ongoing Examination
The broader environment surrounding SPCL is ever-evolving, with the emergence of additional platforms for discourse and more avenues for policymakers and researchers to align and collaborate looming on the horizon.
To delve deeper into PETs and their respective applicability for specific scenarios, NIST has propelled the PETs Testbed. Collaborating with the U.S. Census Bureau XD team and extended through the NCCoE, the inaugural model quandary embodies a SPCL model framework featuring a genomics usage scenario. The model solution will permit the rationalization of the intricacies in deploying SPCL to tackle real-world predicaments. NIST is formulating privacy and efficacy metrics to underpin comprehension of their nexus (e.g., trade-offs) concerning federated learning. The architecture will undergo a privacy menace assessment encompassing the utilization of tools such as the NIST Privacy Framework and the results of a privacy challenge analysis. These evaluations will establish a framework to aid establishments in navigating the compromises in a SPCL setup.
Supplementary Information
As an emerging modus operandi for handling data, there is still abundant to assimilate and delve into concerning SPCL, and PETs more broadly, both theoretically and practically. The enlisted links proffer supplementary insights on this, incorporating instances of real-world utilization:
Should you wish to relay feedback or additional concepts, please contact us at pets [at] dsit.gov.uk (pets[at]dsit[dot]gov[dot]uk) or privacyeng [at] nist.gov (privacyeng[at]nist[dot]gov).
About Author
