Skip to content

SoK references

The 107 references cited by the SoK; files are numbered by reference number.

Folder: papers/references/sok-references/

# Paper Year Local PDF Links
1 Large language models in medicine 2023 yes DOI
2 (a)i am not a lawyer, but...: Engaging legal experts towards responsible llm policies for legal advice 2024 yes DOI · arXiv:2402.01864
3 Integrating large language models in financial investments and market analysis: A survey 2025 yes arXiv:2507.01990
4 Beyond memorization: Violating privacy via inference with large language models 2024 yes arXiv:2310.07298
5 Homomorphic Encryption for Arithmetic of Approximate Numbers 2017 yes DOI · ePrint 2016/421 · Semantic Scholar
6 Approximate homomorphic encryption based privacy-preserving machine learning: a survey 2025 yes DOI · link
7 Collaborative privacy-preserving analysis of oncological data using multiparty homomorphic encryption 2023 yes ePrint 2023/1203
8 Privft: Private and fast text classification with homomorphic encryption 2020 yes DOI · arXiv:1908.06972
9 Private svm inference on encrypted data 2024 yes link
10 Private pathological assessment via machine learning and homomorphic encryption 2024 yes DOI · link
11 Transformer-based language models and homomorphic encryption: An intersection with bert-tiny 2024 no DOI
12 Pritran: Privacy-preserving inference for transformer-based language models under fully homomorphic encryption 2026 no DOI
13 ARION: Attention-optimized transformer inference on encrypted data 2025 yes ePrint 2025/2271
14 MOAI: Module-optimizing architecture for non-interactive secure transformer inference 2026 yes ePrint 2025/991
15 Secure transformer inference made non-interactive 2025 yes DOI · ePrint 2024/136
16 THOR: Secure transformer inference with homomorphic encryption 2025 yes ePrint 2024/1881
17 EncryptedLLM: Privacy-Preserving Large Language Model Inference via GPU-Accelerated Fully Homomorphic Encryption 2024 yes link
18 A scalable multi-GPU framework for encrypted large-model inference 2025 yes DOI · arXiv:2512.11269
19 Euston: Efficient and User-Friendly Secure Transformer Inference with Non-Interactivity 2026 yes ePrint 2026/46 · link
20 STIP: Efficient and secure non-interactive transformer inference via compact packing 2026 yes ePrint 2026/174
21 Fast and Accurate Homomorphic Softmax Evaluation 2024 yes DOI · arXiv:2410.11184
22 Demystifying bootstrapping in fully homomorphic encryption 2023 yes ePrint 2023/149
23 FIDESlib: A Fully-Fledged Open-Source FHE Library for Efficient CKKS on GPUs 2025 yes arXiv:2507.04775
24 SoK: Privacy-preserving computation techniques for deep learning 2021 yes DOI
25 SoK: Cryptographic neural-network computation 2023 no DOI
26 Privacypreserving llm inference in practice: A comparative survey of techniques, trade-offs, and deployability 2026 yes ePrint 2026/105
27 SoK: Connecting the dots in privacy-preserving ML - systematization of MPC protocols and conversions between secret sharing schemes 2025 yes ePrint 2025/1679
28 SoK: Private transformer-based model inference 2026 yes ePrint 2026/491
29 Deep learning on private data 2019 no DOI
30 M. Azraoui, M. Bahram, B. Bozdemir, S. Canard, E. Ciceri, O. Ermis, R. Masalha, M. Mosconi, M. ¨Onen, M. Paindavoine, B. Rozenberg, B. Vialla, and S. 2020 no link
31 Cryptography-based privacy-preserving large language models: a lifecycle survey of frameworks, methods, and future directions 2025 yes DOI
32 Towards efficient privacy-preserving machine learning: A systematic review from protocol, model, and system perspectives 2025 yes ePrint 2025/1365
33 On protecting the data privacy of large language models (llms) and llm agents: A literature review 2025 no DOI
34 SoK: Wildest dreams: Reproducible research in privacy-preserving neural network training 2024 yes DOI · link
35 Sok: Security and privacy in machine learning 2018 no DOI
36 Measuring computational universality of fully homomorphic encryption 2026 no link
37 Attention is all you need 2017 yes arXiv:1706.03762
38 BERT: Pre-training of deep bidirectional transformers for language understanding 2019 yes link
39 Improving language understanding by generative pretraining 2018 yes link
40 Llama: Open and efficient foundation language models 2023 yes arXiv:2302.13971
41 GLUE: A multi-task benchmark and analysis platform for natural language understanding 2018 yes link
42 Recursive deep models for semantic compositionality over a sentiment treebank 2013 yes link
43 The pascal recognising textual entailment challenge 2005 no DOI
44 SQuAD: 100,000+ questions for machine comprehension of text 2016 yes arXiv:1606.05250
45 Automatically constructing a corpus of sentential paraphrases 2005 yes link
46 Pointer sentinel mixture models 2016 yes arXiv:1609.07843
47 The LAMBADA dataset: Word prediction requiring a broad discourse context 2016 yes link
48 HellaSwag: Can a machine really finish your sentence? 2019 yes link
49 Think you have solved question answering? try ARC, the AI2 reasoning challenge 2018 yes arXiv:1803.05457
50 One server for the price of two: Simple and fast single-server private information retrieval 2023 yes ePrint 2022/949
51 SPIRAL: Fast, high-rate single-server PIR via FHE composition 2022 yes ePrint 2022/368
52 Stealing machine learning models via prediction APIs 2016 yes arXiv:1609.02943
53 Neuracrypt is not private 2021 yes arXiv:2108.07256
54 A comprehensive survey on secure outsourced computation and its applications 2019 yes DOI · link
55 On the Security of Homomorphic Encryption on Approximate Numbers 2021 yes ePrint 2020/1533
56 Security guidelines for implementing homomorphic encryption 2024 yes ePrint 2024/463
57 Approximate Homomorphic Encryption over the Conjugate-invariant Ring 2019 yes ePrint 2018/952
58 TFHE Gets Real: An Efficient and Flexible Homomorphic Floating-Point Arithmetic 2025 yes ePrint 2025/257
59 Secure numerical simulations using fully homomorphic encryption 2026 yes DOI · arXiv:2410.21824
60 Privacy-preserving computations of predictive medical models with minimax approximation and non-adjacent form 2017 no DOI
61 Faster bootstrapping with polynomial error 2014 yes DOI · ePrint 2014/94 · link · Semantic Scholar
62 On the number of nonscalar multiplications necessary to evaluate polynomials 1973 no DOI
63 (Leveled) Fully Homomorphic Encryption without Bootstrapping 2012 yes ePrint 2011/277
64 Fully Homomorphic Encryption without Modulus Switching from Classical GapSVP 2012 yes ePrint 2012/78 · link
65 Somewhat Practical Fully Homomorphic Encryption 2012 yes ePrint 2012/144
66 On the precision loss in approximate homomorphic encryption 2024 yes ePrint 2022/162
67 General Functional Bootstrapping using CKKS 2025 yes ePrint 2024/1623
68 DaCapo: Automatic Bootstrapping Management for Efficient Fully Homomorphic Encryption 2024 yes link
69 Secretvoice1.0: Non-interactive secure transformer inference 2026 no link
70 fairseq: A fast, extensible toolkit for sequence modeling 2019 yes arXiv:1904.01038
71 Powerformer: Efficient and high-accuracy privacy-preserving language model with homomorphic encryption 2025 yes link
72 Algorithms in HElib 2014 yes ePrint 2014/106
73 TRICYCLE: Private transformer inference with tricyclic encodings 2025 yes ePrint 2025/1200
74 Cachemir: Fully Homomorphic Encrypted Inference of Generative Large Language Model with KV Cache 2026 yes arXiv:2602.11470
75 ENSI: Efficient non-interactive secure inference for large language models 2025 yes arXiv:2509.09424
76 Bitnet: Scaling 1-bit transformers for large language models 2023 yes arXiv:2310.11453
77 ELLMo: Packing- and Depth-Aware Encrypted Transformer Inference 2026 yes ePrint 2026/198 · Semantic Scholar
78 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) 2026 yes arXiv:2601.18511
79 Secure outsourced matrix computation and application to neural networks 2018 yes ePrint 2018/1041
80 Encryption-friendly LLM architecture 2025 yes arXiv:2410.02486
81 Private and reliable neural network inference 2022 yes DOI · arXiv:2210.15614
82 Converting transformers to polynomial form for secure inference over homomorphic encryption 2024 yes arXiv:2311.08610
83 [Skyformer: Remodel selfattention with gaussian kernel and nystr](../papers/skyformer-remodel-selfattention-with-gaussian-kernel-and-nystr.md) 2021 yes arXiv:2111.00035
84 Theory, analysis, and best practices for sigmoid selfattention 2025 yes arXiv:2409.04431
85 Orion: A Fully Homomorphic Encryption Framework for Deep Learning 2025 yes DOI · arXiv:2311.03470
86 OpenFHE: Open-Source Fully Homomorphic Encryption Library 2022 yes ePrint 2022/915
87 Lattigo v6 2024 no link
88 Phantom: A CUDA-Accelerated Word-Wise Homomorphic Encryption Library 2024 yes ePrint 2023/49 · link
89 Liberate.FHE: A New FHE Library for Bridging the Gap between Theory and Practice with a Focus on Performance and Accuracy 2023 no link
90 Practical secure inference algorithm for a fine-tuned large language model based on fully homomorphic encryption 2026 yes arXiv:2501.01672
91 Simple Encrypted Arithmetic Library - SEAL 2017 no
92 HeLayers: A Tile Tensors Framework for Large Neural Networks on Encrypted Data 2023 yes DOI · arXiv:2011.01805
93 Phantom: A CUDA-Accelerated Word-Wise Homomorphic Encryption Library 2024 yes ePrint 2023/49 · link
94 Cost-efficient llm training with lifetimeaware tensor offloading via gpudirect storage 2025 yes arXiv:2506.06472
95 Homomorphic encryption standard 2019 yes ePrint 2019/939
96 ByteTransformer: A High-Performance Transformer Boosted for Variable-Length Inputs 2023 yes arXiv:2210.03052
97 ByteTransformer: A High-Performance Transformer Boosted for Variable-Length Inputs 2026 yes arXiv:2210.03052
98 Large language model inference acceleration: A comprehensive hardware perspective 2025 yes arXiv:2410.04466
99 NVIDIA NIM LLMs benchmarking: Performance – Llama-3.1-8B-Instruct results 2026 no link
100 SHIP: A Shallow and Highly Parallelizable CKKS Bootstrapping Algorithm 2025 yes ePrint 2025/784
101 PaCo: Bootstrapping for CKKS via Partial CoeffToSlot 2026 yes ePrint 2025/886
102 F1: A Fast and Programmable Accelerator for Fully Homomorphic Encryption 2021 yes arXiv:2109.05371
103 Heracles: 8192-way simd programmable scalable fully-homomorphic encryption soc for privacy-preserving cloud computing in intel 3 cmos 2026 no DOI
104 TREBUCHET: Fully Homomorphic Encryption Accelerator for Deep Computation 2023 yes arXiv:2304.05237
105 BASALISC: Programmable Hardware Accelerator for BGV Fully Homomorphic Encryption 2023 yes arXiv:2205.14017 · ePrint 2022/657
106 Efficient Homomorphic Evaluation on Large Intervals 2022 yes DOI · ePrint 2022/280
107 Fully Homomorphic Encryption for Matrix Arithmetic yes ePrint 2025/1935