Medical researchers at two different institutions may need to compare patient data to advance drug discovery. Cybersecurity professionals may want to discuss intelligence related to attacks in their networks to defend against adversaries. Privacy concerns and safeguards, as well as legal regulations, however, can prevent them from sharing and collaborating on sensitive information, including personally identifiable information or intellectual property.
With $5 million from the U.S. National Science Foundation Office of Advanced Cyberinfrastructure, researchers from San Diego State University, in partnership with Arizona State University and University of Utah, will develop a national community resource for computing and sharing private data.
Over the next five years, the team will deploy and support a novel cyberinfrastructure called Prototype Architecture for Research Advances using Privacy Enhancing Technologies (PARAPET), that could be used in a range of fields that require data analysis, including healthcare, internet security, science, finance and the social sciences, to protect sensitive information such as medical records, financial information and census statistics.
Using privacy-enhancing technologies, PARAPET is designed to be a more secure way to share and collaborate on sensitive data, safeguard data at rest and during computation, and train AI models without exposing private data.
“There are a lot of examples where privacy aspects preclude researchers from working on data together, and that's an impediment to science,” said Robert Beverly, director of SDSU’s Cybersecurity Center for Research and Education (CSCRE) and professor of computer science. “Researchers are entrusted with sensitive data, so we need to protect it. Trying to get these technologies into production promises to move the whole field forward.”
How to secure sharable data
Today, data can be encrypted for transmission and storage, but performing computations on encrypted data remains challenging. This project aims to make that process easier.
To secure shared information, researchers plan to equip PARAPET with specialized software and hardware tools that allow one user to send encrypted data that another user can work on without decrypting it, ensuring the data is never revealed to anyone but the original user. In addition, these techniques could help protect against future quantum computing attacks, which pose a heightened threat to cyber systems. PARAPET will provide an important national-scale testbed for the research community to explore these tools and techniques.
PARAPET also seeks to preventatively protect data, so that it remains inaccessible in the event that a device is compromised. In fact, the entire system will be secured by locking away encrypted data so only approved users on secured networks can access it.
“Our long-term goal is, even if our infrastructure got compromised and an attacker got a copy of encrypted data, they can't do much about it,” said Joann Chen, SDSU assistant professor of computer science.
Finally, with AI spreading into virtually every discipline, it is becoming increasingly more essential to protect data as models are built. In response, PARAPET hopes to protect users’ data while fine-tuning central AI models that can be used nationwide.
One way to do this is users can train their local model on their own device and share only the model updates with PARAPET, so it can use the information from the updates to train its larger model without accessing the raw data.
“You can see any kind of AI model as a giant math function. Basically, we take the output from that function and average all of them from different, local models,” Chen said.
Unlike systems that require organizations to transfer raw data to a central service, PARAPET will test methods that allow participating institutions to retain greater control over sensitive research data.
Beverly points out that even anonymizing data is not secure, as many hackers can conduct re-identification attacks by reverse engineering data to link it back to a person. With secure computing hardware utilizing technologies such as differential privacy, homomorphic encryption and federated learning, PARAPET seeks to provide a vastly more secure option for supporting high performance computing needs.
Manish Parashar, chief AI officer for the University of Utah and executive director of its Scientific Computing and Imaging Institute, will help manage PARAPET’s hardware and ensure the infrastructure is ready for deployment at a national scale.
“Data-driven and AI-enabled research is revolutionizing science, but progress in critical fields such as cybersecurity and public health is hampered by our inability to obtain, create, compute on, and share regulated data,” Parashar said. “PARAPET takes an important step toward solving this problem by leveraging advances in privacy-enhancing technologies. The University of Utah looks forward to contributing expertise and leadership in designing and operating the compliant, regulated environments that PARAPET needs to safely accelerate research with sensitive data.”
An expert on regulatory and policy compliance and director of Arizona State University’s Research Technology Office, Carolyn Ellis will ensure PARAPET meets the rapidly evolving requirements for technologies like this and will lead efforts to recruit users from across the nation to test it once a prototype is ready.
“Privacy-enhancing technologies have enormous potential to change how we conduct research with sensitive or regulated datasets,” said Ellis, a co-PI on the project. “We first need to understand how these new capabilities fit within institutional policies, research workflows, and the privacy expectations established through data-sharing agreements. The ASU team is well-positioned to investigate these new research data use cases.”
Through ASU’s leadership of the national Regulated Research Community of Practice , the team will bring a vital institutional perspective to the PARAPET project. This role bridges the gap between new tech capabilities and the practical realities of protecting sensitive research data.
“We’re building the infrastructure around individual privacy-enhancing technologies, bringing together specialized hardware, software, networking, security and data workflows into a prototype researchers can actually use,” said Mike Farley, SDSU chief technology research officer who oversees the on-campus data center where PARAPET’s servers will be housed. “PARAPET helps us understand what it takes to move these technologies from the lab into practical research cyberinfrastructure.”
This project will build upon SDSU’s growing body of interdisciplinary cyber initiatives and development of new cybersecurity and data privacy curricula under the CSCRE.
This project is made possible through U.S. National Science Foundation Office of Advanced Cyberinfrastructure grant #2537035 .