Fisch wins NSF CAREER Award for work to make private data secure and trustworthy in the cloud
For his work on building cryptographic systems that keep sensitive data private and computational results trustworthy, Ben Fisch has won a Faculty Early Career Development (CAREER) Award from the National Science Foundation (NSF).
Fisch, assistant professor of computer science, will use the $500,000, five-year grant to develop new cryptographic techniques that make it practical to process sensitive data securely in the cloud. The NSF CAREER Award is a prestigious honor for young faculty members and supports the early career activities of teachers and scholars who are considered most likely to become the academic leaders of the future.
As artificial intelligence systems take on a growing role in healthcare and finance, they increasingly depend on sensitive data – medical records, financial histories, private research datasets – that patients and institutions are rightly reluctant to share. When that data is sent to a remote server for processing, two things can go wrong: The data itself can be exposed, and there is no easy way to verify that the server actually performed the computation correctly.
“Cloud servers are powerful, but they are also magnets for attacks, from data breaches to application exploits,” Fisch said. “The goal of this work is to make it possible to outsource computation to an untrusted server while preserving data privacy and allowing the client to verify correctness. That is increasingly important as AI systems are used in higher-stakes settings across healthcare, finance, and scientific research.”
Those tools are known as Fully Homomorphic Encryption, which allows a server to perform calculations directly on encrypted data without ever decrypting it – like solving a puzzle through a locked box without opening it – and cryptographic proof systems, which allow a server to attach a short, verifiable receipt to its output proving the computation was performed correctly. Together they enable what researchers call verifiable computation on private data. In the end, the server learns nothing, and the client can trust the result.
The problem is that combining these two techniques efficiently has proven difficult. Current approaches work in principle but are too inefficient for real-world deployment.
Fisch's project attacks that gap on three fronts: building faster proof systems that work across the different numerical formats used by fully homomorphic encryption systems; constructing end-to-end pipelines that integrate proofs directly into encrypted computation; and developing streamlined protocols for specific high-demand tasks like neural network inference.
“The goal is to bring together two major advances in cryptography that have largely developed on separate tracks,” Fisch said. “Making these two powerful capabilities work together efficiently requires new cryptographic techniques.”
More Details
Published Date
Jul 6, 2026
