aiqbench analysis
This research proposes selective homomorphic inference, a method that encrypts only sensitive data regions while processing the rest in plaintext to reduce computational costs. The study demonstrates that locality-preserving neural architectures can achieve significant speedups for this approach, whereas models with early global mixing offer little benefit. By identifying architectural locality as the critical factor, this work provides a practical pathway to deploying fully homomorphic encryption more efficiently in privacy-sensitive machine learning applications.
Encrypt What Matters: When Selective Homomorphic Inference Is Efficient
arXiv:2609.09357v1 Announce Type: cross Abstract: Fully homomorphic encryption (FHE) enables inference on private data without revealing it to the server, but evaluating an entire input under FHE is expensive. We study \emph{selective homomorphic inference}, where only a sensitive region of interest (ROI) is encrypted, and computations independent of that region are performed in plaintext. Selective evaluation produces the same output as full FHE on the same model, without retraining. Its efficiency depends on how quickly encrypted dependencies spread through the network. For small encrypted ROIs, locality-preserving architectures can achieve order-of-magnitude homomorphic-evaluation speedups, whereas architectures with early global mixing provide essentially no speedup. These results identify locality as the key architectural property governing the benefit of selective homomorphic inference.
Abstract:Fully homomorphic encryption (FHE) enables inference on private data without revealing it to the server, but evaluating an entire input under FHE is expensive. We study \emph{selective homomorphic inference}, where only a sensitive region of interest (ROI) is encrypted, and computations independent of that region are performed in plaintext. Selective evaluation produces the same output as full FHE on the same model, without retraining. Its efficiency depends on how quickly encrypted dependencies spread through the network. For small encrypted ROIs, locality-preserving architectures can achieve order-of-magnitude homomorphic-evaluation speedups, whereas architectures with early global mixing provide essentially no speedup. These results identify locality as the key architectural property governing the benefit of selective homomorphic inference.
Cryptography and Security (cs.CR); Machine Learning (cs.LG)
arXiv:2609.09357 [cs.CR]
(or arXiv:2609.09357v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2609.09357 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 211 words
Related coverage
A Solution to Cryptographic Boxes for Unfriendly AI
LessWrong · Jul 19, 2026
Lattice-Based Cryptography Explained
Quantum Zeitgeist · Aug 15, 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026
FEnc$^2$: Unifying Data Packing for Efficient Private Inference via Convolution and Architecture-Aware Fragment Encoding
arXiv · Jun 16, 2026