Select-Then-Compute: Encrypted Label Selection and Analytics over Distributed Datasets using FHE
Published in Network and Distributed System Security (NDSS)., 2026
Recommended citation: N. Koirala, S. Paik, S. Martin, H. Berens, T. Januszewicz, J. Takeshita, J.H. Seo, & T. Jung. (2026, February). Select-Then-Compute: Encrypted Label Selection and Analytics over Distributed Datasets using FHE. In 33rd Annual Network and Distributed System Security (NDSS 2026). https://eprint.iacr.org/2025/2023
This work presents the first encrypted label selection and analytics protocol construction, enabling a querier to privately obtain the outcomes of downstream functions evaluated on the labels tied to intersected identifiers. The protocol is built on approximate CKKS fully homomorphic encryption (FHE) to support efficient label retrieval and computation over real numbers, making it suitable for workloads involving floating-point arithmetic, such as machine learning inference. To scale to realistic deployments, the paper introduces techniques for handling identifiers from large domains (e.g., 64-bit or 128-bit IDs) while preserving the precision needed for accurate downstream results.
Recommended citation: Koirala, Nirajan, Seunghun Paik, Sam Martin, Helena Berens, Tasha Januszewicz, Jonathan Takeshita, Jae Hong Seo, and Taeho Jung. “Select-Then-Compute: Encrypted Label Selection and Analytics over Distributed Datasets using FHE.” In 33rd Annual Network and Distributed System Security (NDSS). 2026.