SoK: Can Fully Homomorphic Encryption Support General AI Computation? A Functional and Cost Analysis#

Authors: Jiaqi Xue (University of Central Florida), Xin Xin (University of Central Florida), Wei Zhang (University of Central Florida), Mengxin Zheng (University of Central Florida), Qianqian Song (University of Florida), Minxuan Zhou (Illinois Institute of Technology), Yushun Dong (Florida State University), Dongjie Wang (The University of Kansas), Xun Chen (Samsung Research America), Jiafeng Xie (Villanova University), Liqiang Wang (University of Central Florida), David Mohaisen (University of Central Florida), Hongyi Wu (University of Arizona), Qian Lou (University of Central Florida)

• Why: You wouldn't search for an FHE SoK, but the question it answers is one your roadmap will eventually force: is encrypted computation of AI workloads anywhere near economic? This is the field's cost-accounting checkpoint, with a Samsung Research co-author signaling industry attention — and it frames the budget conversation for every "privacy-preserving inference" pitch you'll be pitched next year.
• What: A systematization of running general AI computation under FHE, centered on a functional-completeness and cost analysis — what class of computation is even expressible, and what it costs in time and money (title/author line only — no abstract in program).
• How: Expect a taxonomy of FHE schemes and compiler/frameworks for neural inference, benchmarked across operators and models, with the cost analysis giving concrete latency/overhead figures per workload class.
• Where applicable: Private inference services, confidential ML in regulated environments; the SoK's value is telling you where the boundary of practicality currently sits — treat its conclusions as the reality check for vendor FHE claims.