pllmR14: FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing
FuseFSS targets fusion and efficiency of secure nonlinear/helper operations in FSS-based LLM inference. PLLM status: tracked.
Status: tracked
Year: 2026
Authors: Yuhan Ma, Yong Li, Stefan Schmid
External source: FuseFSS: Efficient Secure LLM Inference with Function Secret Sharing
Source contribution
FuseFSS targets fusion and efficiency of secure nonlinear/helper operations in FSS-based LLM inference.
PLLM perspective
Acquire and pin the full source and artifact; port complete fused helpers with their offline key generation, remaining secure multiplication and rescaling.
Prior reference fused-polynomial experiments are alternative methods, not a FuseFSS implementation.
This paper remains tracked. Named modules below are implementation targets, not finished PLLM components; no reproduction, integration, assurance, or matched benchmark is claimed.
The source's boundary remains binding: A fused key lookup does not make subsequent private polynomial evaluation free. Public model support does not eliminate the second online role.
PLLM modules and documentation
Target module identifiers: gate.fss_fused, compiler.fss_fusion. These identifiers describe scoped work unless this page explicitly links an implementation source. Dependencies: R13.
Backlog
This source is assigned to reimplementation priority 8, Full-system comparison targets. Its source-registry order is 14. Follow current gates and unfinished work in the research backlog.
R13: SIGMA: Secure GPT Inference with Function Secret Sharing
SIGMA provides FSS-based secure transformer operations and GPU-backed inference with preprocessing. PLLM status: tracked.
R15: Efficient Pseudorandom Correlation Generators over Z/p^k Z
Galois-ring constructions extend pseudorandom correlation generation to integer rings, including multiplication-related correlations. PLLM status: tracked.