---
title: "SecureGate — paper by Shaaban and Elmahallawy (2026)"
description: "Federated fine-tuning lets organizations adapt a shared LLM without pooling raw data. 4.20% accuracy. 39 references."
authors: ["Mohamed Shaaban", "Mohamed Elmahallawy"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.1972/"
doi: "10.18653/v1/2026.acl-long.1972"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/shaaban-2026-securegate-learning-when-reveal-pii/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/shaaban-2026-securegate-learning-when-reveal-pii/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/42/) / SecureGate: Learning When to Reveal PII Safely via Token-Gated Dual-Adapters for Federated LLMs

# SecureGate: Learning When to Reveal PII Safely via Token-Gated Dual-Adapters for Federated LLMs

**Authors:** [Mohamed Shaaban](https://md.research.elamur.ai/free/en/authors/mohamed-shaaban/) and Mohamed Elmahallawy · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.1972/) · [PDF](https://aclanthology.org/2026.acl-long.1972.pdf) · [DOI 10.18653/v1/2026.acl-long.1972](https://doi.org/10.18653/v1/2026.acl-long.1972) · [OpenAlex](https://openalex.org/W7166871266)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Federated fine-tuning lets organizations adapt a shared LLM without pooling raw data. However, LLMs can memorize personally identifiable information (PII). Standard protections such as differential privacy or data scrubbing also tend to reduce task performance, and aggregation struggles to balance global and local utility. SecureGate addresses this with two LoRA adapters per organization. A "secure" adapter is trained on sanitized data (via noise injection or [MASK]-style scrubbing) and holds knowledge that can be shared globally. A "revealing" adapter keeps sensitive, organization-specific content. A lightweight gating module checks whether a prompt carries a learned special token and routes it accordingly. Authorized requests reach the revealing adapter, and all others fall back to the secure one, with no retraining required. Experiments on decoder-only models (Llama, Qwen, Gemma) using the ECHR and Yelp Reviews datasets show that unauthorized requests leak far less PII. Authorized users still get near-full utility at a small computational cost. This makes per-request access control practical for privacy-sensitive collaborative LLM training.

## Key points
- Unauthorized requests see up to 31.66× lower inference-attack accuracy and 17.07× lower extraction recall. Routing to the correct adapter was 100% reliable.
- Authorized access reaches 25.20% inference accuracy and 6.32 PPL. Unauthorized access drops to 4.20% accuracy and 15.89 PPL, close to the secure-only baseline of 16.03 PPL.
- On Llama-1B with ECHR, a valid key yields 20.00% attack accuracy versus 2.44% without one, an 8.20× gap. Using the wrong dataset's key exposes little: 4.94% on ECHR and at most 6.76% on Yelp.
- The gating pass adds about 197 ms, roughly 1.13× latency per inference. Fusion-stage FLOPs grow from 11.12 to 41.05 × 10^15 as more revealing adapters are added.
- Stated limitations:
  - added prefill latency;
  - extra memory for multiple adapters;
  - dependence on the secrecy of key tokens;
  - evaluation limited to decoder-only models.

  Code is available on GitHub (wsu-cyber-security-lab-ai/SecureGate).

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

Also on the byline: [Washington State University](https://ror.org/05dk0ce17).

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,772 works · [🤝 Sociology and Political Science](https://md.research.elamur.ai/free/en/shelves/social-sciences/sociology-and-political-science/) · 577 works

🧭 Topics (1): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works

## Cite

Cite as: Mohamed Shaaban and Mohamed Elmahallawy. 2026. SecureGate: Learning When to Reveal PII Safely via Token-Gated Dual-Adapters for Federated LLMs. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{shaaban2026securegate,
  title = {{SecureGate: Learning When to Reveal PII Safely via Token-Gated Dual-Adapters for Federated LLMs}},
  author = {Mohamed Shaaban and Mohamed Elmahallawy},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1972},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1972}
}
```

- [← previous sebis at CRF Filling 2026: A Two-Stage Local LLM Pipeline for Medical CRF Filling](https://md.research.elamur.ai/free/en/papers/sommer-2026-sebis-crf-filling-2026-two/)
- [next → Seeing Isn't Believing: Mitigating Belief Inertia via Active Intervention in Embodied Agents](https://md.research.elamur.ai/free/en/papers/wang-2026-seeing-isn-t-believing-mitigating/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/42/)

## References

39 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Alan Akbik, Tanja Bergmann, Duncan Blythe, Kashif Rasul, Stefan Schweter, and Roland Vollgraf. 2019. [Flair: An easy-to-use framework for state-of-the-art nlp](https://md.research.elamur.ai/free/en/papers/akbik-2019-flair-easy-use-framework-state/). In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics (demonstrations), pages…
2. Atilla Akkus, Masoud Poorghaffar Aghdam, Mingjie Li, Junjie Chu, Michael Backes, Yuyang Zhang, and Sinem Sav. 2025. Generated data with fake privacy: Hidden dangers of fine-tuning large language models on generated data. In 34th USENIX Security Symposium (USENIX Security 25), pages 8075–8093. [find](https://scholar.google.com/scholar?q=Atilla+Akkus%2C+Masoud+Poorghaffar+Aghdam%2C+Mingjie+Li%2C+Junjie+Chu%2C+Michael+Backes%2C+Yuyang+Zhang%2C+and+Sinem+Sav%2E+2025%2E+Generated+data+with+fake+privacy%3A+Hidden+dangers+of+fine%2Dtuning+large+language+model)
3. Youssef Allouah, Rachid Guerraoui, and John Stephan. 2025. Towards trustworthy federated learning with untrusted participants. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Youssef+Allouah%2C+Rachid+Guerraoui%2C+and+John+Stephan%2E+2025%2E+Towards+trustworthy+federated+learning+with+untrusted+participants%2E+In+Forty%2Dsecond+International+Conference+on+Machine+Learning%2E)
4. Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, and 1 others. 2021. [Extracting training data from large language models](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/). In 30th USENIX security symposium (USENIX Security 21), pages 2633–2650…
5. Ilias Chalkidis, Ion Androutsopoulos, and Nikolaos Aletras. 2019. [Neural legal judgment prediction in english](https://md.research.elamur.ai/free/en/papers/chalkidis-2019-neural-legal-judgment-prediction-english/). arXiv preprint arXiv:1906.02059.
6. Shuai Cheng and 1 others. 2025. Understanding pii leakage in large language models: A systematic survey. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25, pages 10409–10417. International Joint Conferences on Artificial Intelligence Organization… [find](https://scholar.google.com/scholar?q=Shuai+Cheng+and+1+others%2E+2025%2E+Understanding+pii+leakage+in+large+language+models%3A+A+systematic+survey%2E+In+Proceedings+of+the+Thirty%2DFourth+International+Joint+Conference+on+Artificial+Intelligence%2C+)
7. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv e-prints, pages arXiv–2407.
8. Muneeb Ul Hassan, Mubashir Husain Rehmani, and Jinjun Chen. 2019. Differential privacy techniques for cyber physical systems: A survey. IEEE Communications Surveys & Tutorials, 22(1):746–789. [find](https://scholar.google.com/scholar?q=Muneeb+Ul+Hassan%2C+Mubashir+Husain+Rehmani%2C+and+Jinjun+Chen%2E+2019%2E+Differential+privacy+techniques+for+cyber+physical+systems%3A+A+survey%2E+IEEE+Communications+Surveys+%26+Tutorials%2C+22%281%29%3A746%E2%80%93789%2E)
9. Tzu-Ming Harry Hsu, Hang Qi, and Matthew Brown. 2019. Measuring the effects of non-identical data distribution for federated visual classification. arXiv preprint arXiv:1909.06335. — [arXiv:1909.06335](https://arxiv.org/abs/1909.06335) [find](https://scholar.google.com/scholar?q=Tzu%2DMing+Harry+Hsu%2C+Hang+Qi%2C+and+Matthew+Brown%2E+2019%2E+Measuring+the+effects+of+non%2Didentical+data+distribution+for+federated+visual+classification%2E+arXiv+preprint+arXiv%3A1909%2E06335%2E+%E2%80%94+%5BarXiv%3A1909%2E06335)
10. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). ICLR, 1(2):3.
11. Chengsong Huang, Qian Liu, Bill Yuchen Lin, Tianyu Pang, Chao Du, and Min Lin. 2023. [Lorahub: Efficient cross-task generalization via dynamic lora composition](https://md.research.elamur.ai/free/en/papers/huang-2023-lorahub-efficient-cross-task-generalization/). arXiv preprint arXiv:2307.13269.
12. Meirui Jiang and 1 others. 2024. Heterogeneous personalized federated learning by local-global updates mixing via convergence rate. In The Twelfth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Meirui+Jiang+and+1+others%2E+2024%2E+Heterogeneous+personalized+federated+learning+by+local%2Dglobal+updates+mixing+via+convergence+rate%2E+In+The+Twelfth+International+Conference+on+Learning+Representations%2E)
13. Peter Kairouz, H Brendan McMahan, Brendan Avent, Aurélien Bellet, Mehdi Bennis, Arjun Nitin Bhagoji, Kallista Bonawitz, Zachary Charles, Graham Cormode, Rachel Cummings, and 1 others. 2021. Advances and open problems in federated learning. Foundations and trends® in machine learning, 14(1– 2):1–210. [find](https://scholar.google.com/scholar?q=Peter+Kairouz%2C+H+Brendan+McMahan%2C+Brendan+Avent%2C+Aur%C3%A9lien+Bellet%2C+Mehdi+Bennis%2C+Arjun+Nitin+Bhagoji%2C+Kallista+Bonawitz%2C+Zachary+Charles%2C+Graham+Cormode%2C+Rachel+Cummings%2C+and+1+others%2E+2021%2E+Advances+a)
14. Bill Yuchen Lin and 1 others. 2022. FedNLP: Benchmarking federated learning methods for natural language processing tasks. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 157–175, Seattle, United States. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Bill+Yuchen+Lin+and+1+others%2E+2022%2E+FedNLP%3A+Benchmarking+federated+learning+methods+for+natural+language+processing+tasks%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+NAACL+2022%2C+pag)
15. Chen Ling, Xujiang Zhao, Jiaying Lu, Chengyuan Deng, Can Zheng, Junxiang Wang, Tanmoy Chowdhury, Yun Li, Hejie Cui, Xuchao Zhang, and 1 others. 2023. Domain specialization as the key to make large language models disruptive: A comprehensive survey. ACM Computing Surveys. [find](https://scholar.google.com/scholar?q=Chen+Ling%2C+Xujiang+Zhao%2C+Jiaying+Lu%2C+Chengyuan+Deng%2C+Can+Zheng%2C+Junxiang+Wang%2C+Tanmoy+Chowdhury%2C+Yun+Li%2C+Hejie+Cui%2C+Xuchao+Zhang%2C+and+1+others%2E+2023%2E+Domain+specialization+as+the+key+to+make+large+lan)
16. Bingyan Liu, Nuoyan Lv, Yuanchun Guo, and Yawen Li. 2024. Recent advances on federated learning: A systematic survey. Neurocomputing, 597:128019. [find](https://scholar.google.com/scholar?q=Bingyan+Liu%2C+Nuoyan+Lv%2C+Yuanchun+Guo%2C+and+Yawen+Li%2E+2024%2E+Recent+advances+on+federated+learning%3A+A+systematic+survey%2E+Neurocomputing%2C+597%3A128019%2E)
17. Guodong Long, Tao Shen, Jing Jiang, Michael Blumen-stein, and 1 others. 2024. Dual-personalizing adapter for federated foundation models. Advances in Neural Information Processing Systems, 37:39409–39433. [find](https://scholar.google.com/scholar?q=Guodong+Long%2C+Tao+Shen%2C+Jing+Jiang%2C+Michael+Blumen%2Dstein%2C+and+1+others%2E+2024%2E+Dual%2Dpersonalizing+adapter+for+federated+foundation+models%2E+Advances+in+Neural+Information+Processing+Systems%2C+37%3A39409%E2%80%9339)
18. Ilya Loshchilov and Frank Hutter. 2017. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101. [find](https://scholar.google.com/scholar?q=Ilya+Loshchilov+and+Frank+Hutter%2E+2017%2E+Decoupled+weight+decay+regularization%2E+arXiv+preprint+arXiv%3A1711%2E05101%2E)
19. Wei Lu, Rachel K Luu, and Markus J Buehler. 2025. Fine-tuning large language models for domain adaptation: Exploration of training strategies, scaling, model merging and synergistic capabilities. npj Computational Materials, 11(1):84. [find](https://scholar.google.com/scholar?q=Wei+Lu%2C+Rachel+K+Luu%2C+and+Markus+J+Buehler%2E+2025%2E+Fine%2Dtuning+large+language+models+for+domain+adaptation%3A+Exploration+of+training+strategies%2C+scaling%2C+model+merging+and+synergistic+capabilities%2E+npj+)
20. Nils Lukas, Ahmed Salem, Robert Sim, Shruti Tople, Lukas Wutschitz, and Santiago Zanella-Béguelin. 2023. Analyzing leakage of personally identifiable information in language models. In 2023 IEEE Symposium on Security and Privacy (SP), pages 346–363. IEEE. [find](https://scholar.google.com/scholar?q=Nils+Lukas%2C+Ahmed+Salem%2C+Robert+Sim%2C+Shruti+Tople%2C+Lukas+Wutschitz%2C+and+Santiago+Zanella%2DB%C3%A9guelin%2E+2023%2E+Analyzing+leakage+of+personally+identifiable+information+in+language+models%2E+In+2023+IEEE+Sympo)
21. Mohamad Mansouri, Melek Önen, Wafa Ben Jaballah, and Mauro Conti. 2023. Sok: Secure aggregation based on cryptographic schemes for federated learning. Proceedings on Privacy Enhancing Technologies. [find](https://scholar.google.com/scholar?q=Mohamad+Mansouri%2C+Melek+%C3%96nen%2C+Wafa+Ben+Jaballah%2C+and+Mauro+Conti%2E+2023%2E+Sok%3A+Secure+aggregation+based+on+cryptographic+schemes+for+federated+learning%2E+Proceedings+on+Privacy+Enhancing+Technologies%2E)
22. Brendan McMahan and 1 others. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273–1282. PMLR. [find](https://scholar.google.com/scholar?q=Brendan+McMahan+and+1+others%2E+2017%2E+Communication%2Defficient+learning+of+deep+networks+from+decentralized+data%2E+In+Artificial+intelligence+and+statistics%2C+pages+1273%E2%80%931282%2E+PMLR%2E)
23. David Nadeau and Satoshi Sekine. 2007. A survey of named entity recognition and classification. Lingvisticae Investigationes, 30(1):3–26. [find](https://scholar.google.com/scholar?q=David+Nadeau+and+Satoshi+Sekine%2E+2007%2E+A+survey+of+named+entity+recognition+and+classification%2E+Lingvisticae+Investigationes%2C+30%281%29%3A3%E2%80%9326%2E)
24. Venkatesh Balavadhani Parthasarathy, Ahtsham Zafar, Aafaq Khan, and Arsalan Shahid. 2024. The ultimate guide to fine-tuning llms from basics to breakthroughs: An exhaustive review of technologies, research, best practices, applied research challenges and opportunities. arXiv preprint arXiv:2408.1329… — [arXiv:2408.13296](https://arxiv.org/abs/2408.13296) [find](https://scholar.google.com/scholar?q=Venkatesh+Balavadhani+Parthasarathy%2C+Ahtsham+Zafar%2C+Aafaq+Khan%2C+and+Arsalan+Shahid%2E+2024%2E+The+ultimate+guide+to+fine%2Dtuning+llms+from+basics+to+breakthroughs%3A+An+exhaustive+review+of+technologies%2C+res)
25. Yuefeng Peng, Junda Wang, Hong Yu, and Amir Houmansadr. 2024. Data extraction attacks in retrieval-augmented generation via backdoors. arXiv preprint arXiv:2411.01705. — [arXiv:2411.01705](https://arxiv.org/abs/2411.01705) [find](https://scholar.google.com/scholar?q=Yuefeng+Peng%2C+Junda+Wang%2C+Hong+Yu%2C+and+Amir+Houmansadr%2E+2024%2E+Data+extraction+attacks+in+retrieval%2Daugmented+generation+via+backdoors%2E+arXiv+preprint+arXiv%3A2411%2E01705%2E+%E2%80%94+%5BarXiv%3A2411%2E01705%5D%28https%3A%2F%2Farx)
26. Ildikó Pilán, Pierre Lison, Lilja Øvrelid, Anthi Papadopoulou, David Sánchez, and Montserrat Batet. 2022. [The text anonymization benchmark (tab): A dedicated corpus and evaluation framework for text anonymization](https://md.research.elamur.ai/free/en/papers/pilan-2022-text-anonymization-benchmark-tab-dedicated/). Computational Linguistics, 48(4):1053–1101.
27. Samar Pratap and 1 others. 2025. The fine art of fine-tuning: A structured review of advanced llm fine-tuning techniques. Natural Language Processing Journal, 11:100144. [find](https://scholar.google.com/scholar?q=Samar+Pratap+and+1+others%2E+2025%2E+The+fine+art+of+fine%2Dtuning%3A+A+structured+review+of+advanced+llm+fine%2Dtuning+techniques%2E+Natural+Language+Processing+Journal%2C+11%3A100144%2E)
28. Sashank Reddi and 1 others. 2020. Adaptive federated optimization. arXiv preprint arXiv:2003.00295. — [arXiv:2003.00295](https://arxiv.org/abs/2003.00295) [find](https://scholar.google.com/scholar?q=Sashank+Reddi+and+1+others%2E+2020%2E+Adaptive+federated+optimization%2E+arXiv+preprint+arXiv%3A2003%2E00295%2E+%E2%80%94+%5BarXiv%3A2003%2E00295%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2003%2E00295%29)
29. Lei Shen, Zhenheng Tang, Lijun Wu, Yonggang Zhang, Xiaowen Chu, Tao Qin, and Bo Han. Hot-pluggable federated learning: Bridging general and personalized fl via dynamic selection. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Lei+Shen%2C+Zhenheng+Tang%2C+Lijun+Wu%2C+Yonggang+Zhang%2C+Xiaowen+Chu%2C+Tao+Qin%2C+and+Bo+Han%2E+Hot%2Dpluggable+federated+learning%3A+Bridging+general+and+personalized+fl+via+dynamic+selection%2E+In+The+Thirteenth+Int)
30. Ilya Sutskever, James Martens, George Dahl, and Geoffrey Hinton. 2013. On the importance of initialization and momentum in deep learning. In International conference on machine learning, pages 1139– 1147. pmlr. [find](https://scholar.google.com/scholar?q=Ilya+Sutskever%2C+James+Martens%2C+George+Dahl%2C+and+Geoffrey+Hinton%2E+2013%2E+On+the+importance+of+initialization+and+momentum+in+deep+learning%2E+In+International+conference+on+machine+learning%2C+pages+1139%E2%80%93+1)
31. Gemma Team. 2024. Gemma: Open models based on gemini research and technology. arXiv preprint arXiv:2403.08295. [find](https://scholar.google.com/scholar?q=Gemma+Team%2E+2024%2E+Gemma%3A+Open+models+based+on+gemini+research+and+technology%2E+arXiv+preprint+arXiv%3A2403%2E08295%2E)
32. Xinghao Wu, Xuefeng Liu, Jianwei Niu, Haolin Wang, Shaojie Tang, and Guogang Zhu. 2024. Fedlora: When personalized federated learning meets low-rank adaptation. [find](https://scholar.google.com/scholar?q=Xinghao+Wu%2C+Xuefeng+Liu%2C+Jianwei+Niu%2C+Haolin+Wang%2C+Shaojie+Tang%2C+and+Guogang+Zhu%2E+2024%2E+Fedlora%3A+When+personalized+federated+learning+meets+low%2Drank+adaptation%2E)
33. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, and et al. 2025. [Qwen3: Large language models with hybrid thinking modes and extensive multilingual support](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). arXiv preprint arXiv:2505.09388.
34. Rui Ye, Wenhao Wang, Jingyi Chai, Dihan Li, Zexi Li, Yinda Xu, Yaxin Du, Yanfeng Wang, and Siheng Chen. 2024. Openfedllm: Training large language models on decentralized private data via federated learning. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, page… [find](https://scholar.google.com/scholar?q=Rui+Ye%2C+Wenhao+Wang%2C+Jingyi+Chai%2C+Dihan+Li%2C+Zexi+Li%2C+Yinda+Xu%2C+Yaxin+Du%2C+Yanfeng+Wang%2C+and+Siheng+Chen%2E+2024%2E+Openfedllm%3A+Training+large+language+models+on+decentralized+private+data+via+federated+lea)
35. Liping Yi and 1 others. 2023. pfedlora: Model-heterogeneous personalized federated learning with lora tuning. arXiv preprint arXiv:2310.13283. — [arXiv:2310.13283](https://arxiv.org/abs/2310.13283) [find](https://scholar.google.com/scholar?q=Liping+Yi+and+1+others%2E+2023%2E+pfedlora%3A+Model%2Dheterogeneous+personalized+federated+learning+with+lora+tuning%2E+arXiv+preprint+arXiv%3A2310%2E13283%2E+%E2%80%94+%5BarXiv%3A2310%2E13283%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2310%2E13283%29)
36. Xiang Zhang, Junbo Zhao, and Yann LeCun. 2015. [Character-level convolutional networks for text classification](https://md.research.elamur.ai/free/en/papers/zhang-2015-character-level-convolutional-networks-text/). Advances in neural information processing systems, 28.
37. Zeyuan Zuo, Ningxin Su, Baochun Li, and Teng Zhang. 2024. Pack: Towards communication-efficient homomorphic encryption in federated learning. In Proceedings of the 2024 ACM Symposium on Cloud Computing, pages 470–486. [find](https://scholar.google.com/scholar?q=Zeyuan+Zuo%2C+Ningxin+Su%2C+Baochun+Li%2C+and+Teng+Zhang%2E+2024%2E+Pack%3A+Towards+communication%2Defficient+homomorphic+encryption+in+federated+learning%2E+In+Proceedings+of+the+2024+ACM+Symposium+on+Cloud+Computin)
38. Lang Pu, Jingjing Gu, Chao Lin, and Xinyi Huang. 2025. Janus: Dual-server multi-round secure aggregation with verifiability for federated learning. In Forty-second International Conference on Machine Learning. [find](https://scholar.google.com/scholar?q=Lang+Pu%2C+Jingjing+Gu%2C+Chao+Lin%2C+and+Xinyi+Huang%2E+2025%2E+Janus%3A+Dual%2Dserver+multi%2Dround+secure+aggregation+with+verifiability+for+federated+learning%2E+In+Forty%2Dsecond+International+Conference+on+Machine+)
39. Jiaxing Qi, Zhongzhi Luan, Shaohan Huang, Carol Fung, Hailong Yang, and Depei Qian. 2024. Fdlora: Personalized federated learning of large language model via dual lora tuning. arXiv preprint arXiv:2406.07925. *Table 5: Inference attack accuracy (%) for Llama-1B across non-IID clients (Clients 1–5: E… — [arXiv:2406.07925](https://arxiv.org/abs/2406.07925) [find](https://scholar.google.com/scholar?q=Jiaxing+Qi%2C+Zhongzhi+Luan%2C+Shaohan+Huang%2C+Carol+Fung%2C+Hailong+Yang%2C+and+Depei+Qian%2E+2024%2E+Fdlora%3A+Personalized+federated+learning+of+large+language+model+via+dual+lora+tuning%2E+arXiv+preprint+arXiv%3A240)

*Updated 1 November 2026*