| An Open-Source End-to-End FHE Implementation for Privacy-Preserving Llama 3 8B Inference |
Fan et al. |
2026 |
arXiv 2609.12378 |
system |
arXiv:2609.12378 |
| ATLAS: Automated Approximation of Transformers for Efficient Homomorphic Inference in One Hour |
Xie et al. |
2026 |
IACR ePrint 2026/1817 |
system |
ePrint 2026/1817 |
| Cachemir: Fully Homomorphic Encrypted Inference of Generative Large Language Model with KV Cache |
Yu et al. |
2026 |
arXiv 2602.11470 |
system |
arXiv:2602.11470 |
| CGF-Softmax: A Cumulant-Based Softmax Reformulation for Efficient Inference under Homomorphic Encryption |
Park et al. |
2026 |
arXiv 2602.01621 |
system |
arXiv:2602.01621 |
| CRAfT: Constant-Round Non-Polynomial Approximation and P2P Network for Secure Transformer Inference |
Zhao et al. |
2026 |
IACR ePrint 2026/1292 |
system |
ePrint 2026/1292 |
| ELLMo: Packing- and Depth-Aware Encrypted Transformer Inference |
Guzelhan et al. |
2026 |
IACR ePrint 2026/198 |
system |
ePrint 2026/198 · Semantic Scholar |
| EncFormer: Secure and Efficient Transformer Inference over Encrypted Data |
Zhu et al. |
2026 |
arXiv 2604.09975 |
system |
arXiv:2604.09975 |
| Error Propagation-Aware Scale Design for Efficient Homomorphic Encryption-Based LLM Inference |
Seo and Min |
2026 |
IACR ePrint 2026/2242 |
system |
ePrint 2026/2242 |
| Euston: Efficient and User-Friendly Secure Transformer Inference with Non-Interactivity |
Gao et al. |
2026 |
IACR ePrint 2026/046 |
system |
ePrint 2026/46 · link |
| HEAT: Faster Fully Homomorphic Inference via Approximations-Weights Co-Adaptation |
Zirilli et al. |
2026 |
IACR ePrint 2026/1862 |
system |
ePrint 2026/1862 |
| MOAI: Module-optimizing architecture for non-interactive secure transformer inference |
Zhang et al. |
2026 |
ICLR 2026; IACR ePrint 2025/991 |
system |
ePrint 2025/991 |
| ROSETTA: Efficient and Accurate Privacy-Preserving LLM Decoding via Hybrid CKKS/TFHE Evaluation |
Yu et al. |
2026 |
IACR ePrint 2026/2047 |
system |
DOI · ePrint 2026/2047 |
| Scaling up FHE-based Privacy-Preserving ML: Higher Throughput, Longer Inputs for LLama-3-8B (v1: Scaling up Privacy-Preserving ML: A CKKS Implementation of Llama-2-7B) |
Park et al. |
2026 |
arXiv 2601.18511 |
system |
arXiv:2601.18511 |
| SoK: Private LLM Inference using Approximate Homomorphic Encryption |
Al Badawi et al. |
2026 |
SoK paper (IACR ePrint 2026/935) |
system |
ePrint 2026/935 |
| STIP: Efficient and secure non-interactive transformer inference via compact packing |
Wang et al. |
2026 |
IACR ePrint 2026/174 |
system |
ePrint 2026/174 |
| Terrazzo: Memory-Aware GPU Framework for Private Transformer Inference |
Shokri and Tsoutsos |
2026 |
IACR ePrint 2026/1870 |
system |
ePrint 2026/1870 |
| THEMIS: A Co-Designed System for Encrypted Transformer Inference |
Li et al. |
2026 |
IACR ePrint 2026/2264 |
system |
ePrint 2026/2264 |
| Zephyr: GPU-Efficient Homomorphic Encryption for Privacy-Preserving Transformer Inference |
Seo and Min |
2026 |
IACR ePrint 2026/932 |
system |
ePrint 2026/932 |
| A scalable multi-GPU framework for encrypted large-model inference |
Jayashankar et al. |
2025 |
arXiv 2512.11269 |
system |
DOI · arXiv:2512.11269 |
| An Efficient Architecture for Edge AI Federated Learning With Homomorphic Encryption |
Rahbari et al. |
2025 |
IEEE Access |
system |
DOI |
| ARION: Attention-optimized transformer inference on encrypted data |
Yang et al. |
2025 |
IACR ePrint 2025/2271 |
system |
ePrint 2025/2271 |
| BlindFL: Segmented Federated Learning with Fully Homomorphic Encryption |
Gronberg et al. |
2025 |
arXiv preprint |
system |
arXiv:2501.11659 |
| BumbleBee: Secure Two-party Inference Framework for Large Transformers |
Lu et al. |
2025 |
NDSS 2025 |
system |
DOI · ePrint 2023/1678 |
| Centaur: Bridging the Impossible Trinity of Privacy, Efficiency, and Performance in Privacy-Preserving Transformer Inference |
Luo et al. |
2025 |
ACL Findings 2025 |
system |
arXiv:2412.10652 |
| Cost-efficient llm training with lifetimeaware tensor offloading via gpudirect storage |
Yuan et al. |
2025 |
arXiv preprint |
system |
arXiv:2506.06472 |
| Encryption-friendly LLM architecture |
Rho et al. |
2025 |
ICLR 2025 |
system |
arXiv:2410.02486 |
| ENSI: Efficient non-interactive secure inference for large language models |
He et al. |
2025 |
arXiv |
system |
arXiv:2509.09424 |
| Federated Learning: An approach with Hybrid Homomorphic Encryption |
Correia et al. |
2025 |
arXiv preprint |
system |
arXiv:2509.03427 |
| Powerformer: Efficient and high-accuracy privacy-preserving language model with homomorphic encryption |
Park et al. |
2025 |
ACL 2025 (long) |
system |
link |
| Practical and Private Hybrid ML Inference with Fully Homomorphic Encryption |
Biswas et al. |
2025 |
arXiv preprint |
system |
arXiv:2509.01253 |
| Practical secure inference algorithm for a fine-tuned large language model based on fully homomorphic encryption |
Ruoyan et al. |
2025 |
arXiv 2025 |
system |
arXiv:2501.01672 |
| Secure transformer inference made non-interactive |
Zhang et al. |
2025 |
NDSS 2025 |
system |
DOI · ePrint 2024/136 |
| SHAFT: Secure, Handy, Accurate, and Fast Transformer Inference |
Kei and Chow |
2025 |
IACR ePrint 2025/2324 |
system |
DOI · ePrint 2025/2324 |
| SpyKing—Privacy-preserving framework for Spiking Neural Networks |
Nikfam et al. |
2025 |
Frontiers in Neuroscience |
system |
DOI |
| Swift: Fast Secure Neural Network Inference With Fully Homomorphic Encryption |
Fu et al. |
2025 |
IEEE Transactions on Information Forensics and Security |
system |
DOI |
| THOR: Secure transformer inference with homomorphic encryption |
Moon et al. |
2025 |
IACR ePrint / CCS 2025 |
system |
ePrint 2024/1881 |
| TRICYCLE: Private transformer inference with tricyclic encodings |
Lim et al. |
2025 |
IACR ePrint 2025/1200 |
system |
ePrint 2025/1200 |
| BOLT: Privacy-Preserving, Accurate and Efficient Inference for Transformers |
Pang et al. |
2024 |
IEEE S&P 2024 |
system |
ePrint 2023/1893 |
| Confidential Computing on NVIDIA Hopper GPUs: A Performance Benchmark Study |
Zhu et al. |
2024 |
arXiv 2024 |
system |
arXiv:2409.03992 |
| Converting transformers to polynomial form for secure inference over homomorphic encryption |
Zimerman et al. |
2024 |
arXiv |
system |
arXiv:2311.08610 |
| Ditto: Quantization-aware Secure Inference of Transformers upon MPC |
Wu et al. |
2024 |
ICML 2024 |
system |
arXiv:2405.05525 |
| EncryptedLLM: Privacy-Preserving Large Language Model Inference via GPU-Accelerated Fully Homomorphic Encryption |
Castro et al. |
2024 |
NeurIPS 2024 workshop; ICML 2025 (EncryptedLLM) |
system |
link |
| MPC-Minimized Secure LLM Inference |
Rathee et al. |
2024 |
arXiv 2024 |
system |
arXiv:2408.03561 |
| Nimbus: Secure and Efficient Two-Party Inference for Transformers |
Li et al. |
2024 |
NeurIPS 2024 |
system |
arXiv:2411.15707 |
| Private Cloud Compute: A new frontier for AI privacy in the cloud |
Apple |
2024 |
Apple Security Research |
system |
link |
| SecFormer: Fast and Accurate Privacy-Preserving Inference for Transformer Models via SMPC |
Luo et al. |
2024 |
ACL Findings 2024 |
system |
arXiv:2401.00793 |
| SIGMA: Secure GPT Inference with Function Secret Sharing |
Gupta et al. |
2024 |
PETS 2024 |
system |
DOI · ePrint 2023/1269 · Semantic Scholar |
| Split-and-Denoise: Protect Large Language Model Inference with Local Differential Privacy |
Mai et al. |
2024 |
ICML 2024 |
system |
arXiv:2310.09130 |
| zkLLM: Zero Knowledge Proofs for Large Language Models |
Sun et al. |
2024 |
CCS 2024 |
system |
arXiv:2404.16109 |
| CipherGPT: Secure Two-Party GPT Inference |
Hou et al. |
2023 |
IACR ePrint 2023/1147 |
system |
ePrint 2023/1147 |
| EZKL: zkSNARKs for ONNX models |
Zkonduit |
2023 |
Open-source toolkit |
system |
link |
| FedML-HE: An Efficient Homomorphic-Encryption-Based Privacy-Preserving Federated Learning System |
Jin et al. |
2023 |
arXiv preprint |
system |
arXiv:2303.10837 |
| InferDPT: Privacy-Preserving Inference for Closed-box Large Language Model |
Tong et al. |
2023 |
arXiv 2023 |
system |
arXiv:2310.12214 |
| MPCFormer: Fast, Performant and Private Transformer Inference with MPC |
Li et al. |
2023 |
ICLR 2023 |
system |
arXiv:2211.01452 |
| NVIDIA Confidential Computing |
NVIDIA |
2023 |
Product |
system |
link |
| PPFLHE: A privacy-preserving federated learning scheme with homomorphic encryption for healthcare data |
Wang et al. |
2023 |
Applied Soft Computing |
system |
DOI |
| PrivateLoRA for Efficient Privacy Preserving LLM |
Wang et al. |
2023 |
arXiv 2023 |
system |
arXiv:2311.14030 |
| PUMA: Secure Inference of LLaMA-7B in Five Minutes |
Dong et al. |
2023 |
arXiv 2023 |
system |
DOI · arXiv:2307.12533 |
| SecretFlow-SPU: A Performant and User-Friendly Framework for Privacy-Preserving Machine Learning |
Ma et al. |
2023 |
USENIX ATC 2023 |
system |
link |
| Cheetah: Lean and Fast Secure Two-Party Deep Neural Network Inference |
Huang et al. |
2022 |
USENIX Security 2022 |
system |
ePrint 2022/207 · Semantic Scholar |
| Iron: Private Inference on Transformers |
Hao et al. |
2022 |
NeurIPS 2022 |
system |
link |
| Privacy-Preserving Machine Learning With Fully Homomorphic Encryption for Deep Neural Network |
Lee et al. |
2022 |
IEEE Access |
system |
DOI |
| Private and reliable neural network inference |
Jovanovic et al. |
2022 |
Proceedings of the 2022 ACM SIGSAC Conference on Computer an |
system |
DOI · arXiv:2210.15614 |
| THE-X: Privacy-Preserving Transformer Inference with Homomorphic Encryption |
Chen et al. |
2022 |
ACL Findings 2022 |
system |
arXiv:2206.00216 |
| Confidential VMs (Intel TDX, AMD SEV-SNP) |
Intel and AMD |
2021 |
Product |
system |
link |
| CrypTen: Secure Multi-Party Computation Meets Machine Learning |
Knott et al. |
2021 |
NeurIPS 2021 |
system |
arXiv:2109.00984 |
| Towards the AlexNet Moment for Homomorphic Encryption: HCNN, the First Homomorphic CNN on Encrypted Data With GPUs |
Badawi et al. |
2021 |
IEEE Transactions on Emerging Topics in Computing |
system |
DOI |
| CrypTFlow2: Practical 2-Party Secure Inference |
Rathee et al. |
2020 |
CCS 2020 |
system |
ePrint 2020/1002 |
| Delphi: A Cryptographic Inference Service for Neural Networks |
Mishra et al. |
2020 |
USENIX Security 2020 |
system |
DOI · ePrint 2020/50 · link · Semantic Scholar |
| MP-SPDZ: A Versatile Framework for Multi-Party Computation |
Keller |
2020 |
CCS 2020 |
system |
ePrint 2020/521 |
| Privft: Private and fast text classification with homomorphic encryption |
Badawi et al. |
2020 |
IEEE Access, 2020 |
system |
DOI · arXiv:1908.06972 |
| Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware |
Tramèr and Boneh |
2019 |
ICLR 2019 |
system |
arXiv:1806.03287 |
| GAZELLE: A Low Latency Framework for Secure Neural Network Inference |
Juvekar et al. |
2018 |
USENIX Security 2018 |
system |
arXiv:1801.05507 |