CipherGenome: Homomorphic Inference for Genomic Mixture-of-Experts

Abstract

Genome foundation models are growing into sparse mixture-of-experts (MoE) networks whose expert weights no longer fit on the machines that hold the sequences, yet sending a private genome to rented accelerators exposes it: we show that a single server hosting one expert recovers the input nucleotides with 99.8% top-1 accuracy. We present CipherGenome, a protocol that keeps the embedding, attention and router of a 15.1B-parameter MoE genome model on a trusted thin client and outsources every expert projection, 95.8% of the parameters, to untrusted and possibly colluding GPU servers under module-LWE encryption. The design exploits three structural facts: expert layers are linear between two SwiGLU gates, expert weights are public, and GPU integer tensor cores can evaluate a ciphertext-weight product exactly modulo 2⁴⁸ in a single GEMM. The client evaluates the nonlinearity exactly and re-encrypts with fresh secrets, so no polynomial approximation or bootstrapping is ever needed. On 72 windows from 12 bacterial genomes, encryption adds 2.54 × 10⁻⁴ nats per token of KL divergence (95% CI upper bound 3.95 × 10⁻⁴), below a pre-registered non-inferiority margin and indistinguishable from bf16 inference, while the same inversion attack falls to chance level. A reusable public hint cuts end-to-end latency by 3.54 times, wire compression reduces traffic 6.8 times, per-layer padding reduces routing leakage from 54.9% to 8.9% accuracy, and HE-compatible int4 experts remain non-inferior to their plaintext counterparts. Per expert and token, the server-side cost is more than six orders of magnitude below a CKKS baseline.

ForceFlowAb: physics-aware mixture-of-experts flow matching model for antibody CDRs design
Z. Li, Z. Lv, G. Zhang
bioRxiv7 Sep 2026PreprintAll rights reservedFull textBioinformatics & computational biologybioRxiv
MI-PEFT: Mixture-of-Experts Integrated Parameter-Efficient Fine-Tuning Protein Language Models Improves Acidophilic Proteins Classification
Honghan Shen
arXiv7 Sep 2026PreprintCC BY 4.0Full textBioinformatics & computational biologyarXiv
Two blind spots in the demographic inference of human origins from genomic data
Ryan N Gutenkunst
arXiv27 Aug 2026PreprintarXiv licenceFull textEvolutionary biologyarXiv
BlockFormer : Transformer-based inference from genomic contact maps
Eloïse Touron, Pedro L. C. Rodrigues, Julyan Arbel +2
arXiv30 Sep 2026PreprintCC BY 4.0Full textBioinformatics & computational biologyarXiv
Robust Demographic Inference from Genomic and SNP Data
Laurent Excoffier, Isabelle Dupanloup, Emilia Huerta-Sánchez +2
PLoS Genetics28 Oct 2026Peer-reviewedCC BYFull text1,480 citedGenetics & genomicsHAL

Cite

BibTeX RIS

BibTeX

@misc{yang2026ciphergenome,
  title = {{CipherGenome: Homomorphic Inference for Genomic Mixture-of-Experts}},
  author = {Guang Yang and Fengchen Liu},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2609.35883},
  eprint = {2609.35883},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.35883},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - CipherGenome: Homomorphic Inference for Genomic Mixture-of-Experts
AU  - Guang Yang
AU  - Fengchen Liu
PY  - 2026
DA  - 2026/10/01
PB  - arXiv
DO  - 10.48550/arxiv.2609.35883
UR  - https://arxiv.org/abs/2609.35883
AB  - Genome foundation models are growing into sparse mixture-of-experts (MoE) networks whose expert weights no longer fit on the machines that hold the sequences, yet sending a private genome to rented accelerators exposes it: we show that a single server hosting one expert recovers the input nucleotides with 99.8% top-1 accuracy. We present CipherGenome, a protocol that keeps the embedding, attention and router of a 15.1B-parameter MoE genome model on a trusted thin client and outsources every expert projection, 95.8% of the parameters, to untrusted and possibly colluding GPU servers under module-LWE encryption. The design exploits three structural facts: expert layers are linear between two SwiGLU gates, expert weights are public, and GPU integer tensor cores can evaluate a ciphertext-weight product exactly modulo 2^48 in a single GEMM. The client evaluates the nonlinearity exactly and re-encrypts with fresh secrets, so no polynomial approximation or bootstrapping is ever needed. On 72 windows from 12 bacterial genomes, encryption adds 2.54 × 10^−4 nats per token of KL divergence (95% CI upper bound 3.95 × 10^−4), below a pre-registered non-inferiority margin and indistinguishable from bf16 inference, while the same inversion attack falls to chance level. A reusable public hint cuts end-to-end latency by 3.54 times, wire compression reduces traffic 6.8 times, per-layer padding reduces routing leakage from 54.9% to 8.9% accuracy, and HE-compatible int4 experts remain non-inferior to their plaintext counterparts. Per expert and token, the server-side cost is more than six orders of magnitude below a CKKS baseline.
ER  -

Abstract

Related

Cite