@inproceedings{liu-etal-2026-privacy,
title = "Privacy-Preserving Text Sanitization for Distributed Agents Collaboration via Disentangled Representations",
author = "Liu, Xuan and
Zhou, Hefeng and
Chen, Sicheng and
Yang, Chao and
Xu, Xingcheng and
Qu, Jingjing and
Lou, Jiong and
LI, Jie and
Hu, Xia",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.1052/",
pages = "24148--24170",
ISBN = "979-8-89176-459-0",
abstract = "Distributed agents need to exchange textual evidence across organizational boundaries, but these exchanges can expose two types of private information. One is carried by explicit identifiers. The other appears as source signatures spread across formatting, vocabulary, and syntax. Existing PII detection and masking methods mainly handle localized identifiers, leaving these broader source signals largely intact. Even masking 19.2{\%} of tokens reduces TF-IDF stylometric attribution by only 18.6{\%}. To address this problem, we propose *DiSan*, which treats text sanitization as a representation learning problem. A two-stream encoder maps text into orthogonal role and style subspaces. The role subspace captures task semantics shared across sources, while the local style subspace captures variation associated with source identity. The style representation is never transmitted directly. Prototype alignment supports joint training without centralizing raw text, while adversarial regularization limits source leakage in shared representations and prototypes. We test leakage in sanitized text, learned representations, and uploaded prototypes. On a distributed multi-agent RAG benchmark, *DiSan* reduces PII exposure in answers by 20{\texttimes} while preserving 83{\%} answer faithfulness. On Enron emails, it reduces TF-IDF stylometric attribution by 73.2{\%}, compared with 18.6{\%} for the strongest masking baseline. Embedding and prototype attribution are also close to random. Together, these results show that *DiSan* reduces explicit and distributional leakage while retaining task utility."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-privacy">
<titleInfo>
<title>Privacy-Preserving Text Sanitization for Distributed Agents Collaboration via Disentangled Representations</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xuan</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hefeng</namePart>
<namePart type="family">Zhou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sicheng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chao</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xingcheng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingjing</namePart>
<namePart type="family">Qu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiong</namePart>
<namePart type="family">Lou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jie</namePart>
<namePart type="family">LI</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xia</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Distributed agents need to exchange textual evidence across organizational boundaries, but these exchanges can expose two types of private information. One is carried by explicit identifiers. The other appears as source signatures spread across formatting, vocabulary, and syntax. Existing PII detection and masking methods mainly handle localized identifiers, leaving these broader source signals largely intact. Even masking 19.2% of tokens reduces TF-IDF stylometric attribution by only 18.6%. To address this problem, we propose *DiSan*, which treats text sanitization as a representation learning problem. A two-stream encoder maps text into orthogonal role and style subspaces. The role subspace captures task semantics shared across sources, while the local style subspace captures variation associated with source identity. The style representation is never transmitted directly. Prototype alignment supports joint training without centralizing raw text, while adversarial regularization limits source leakage in shared representations and prototypes. We test leakage in sanitized text, learned representations, and uploaded prototypes. On a distributed multi-agent RAG benchmark, *DiSan* reduces PII exposure in answers by 20× while preserving 83% answer faithfulness. On Enron emails, it reduces TF-IDF stylometric attribution by 73.2%, compared with 18.6% for the strongest masking baseline. Embedding and prototype attribution are also close to random. Together, these results show that *DiSan* reduces explicit and distributional leakage while retaining task utility.</abstract>
<identifier type="citekey">liu-etal-2026-privacy</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1052/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>24148</start>
<end>24170</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Privacy-Preserving Text Sanitization for Distributed Agents Collaboration via Disentangled Representations
%A Liu, Xuan
%A Zhou, Hefeng
%A Chen, Sicheng
%A Yang, Chao
%A Xu, Xingcheng
%A Qu, Jingjing
%A Lou, Jiong
%A LI, Jie
%A Hu, Xia
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F liu-etal-2026-privacy
%X Distributed agents need to exchange textual evidence across organizational boundaries, but these exchanges can expose two types of private information. One is carried by explicit identifiers. The other appears as source signatures spread across formatting, vocabulary, and syntax. Existing PII detection and masking methods mainly handle localized identifiers, leaving these broader source signals largely intact. Even masking 19.2% of tokens reduces TF-IDF stylometric attribution by only 18.6%. To address this problem, we propose *DiSan*, which treats text sanitization as a representation learning problem. A two-stream encoder maps text into orthogonal role and style subspaces. The role subspace captures task semantics shared across sources, while the local style subspace captures variation associated with source identity. The style representation is never transmitted directly. Prototype alignment supports joint training without centralizing raw text, while adversarial regularization limits source leakage in shared representations and prototypes. We test leakage in sanitized text, learned representations, and uploaded prototypes. On a distributed multi-agent RAG benchmark, *DiSan* reduces PII exposure in answers by 20× while preserving 83% answer faithfulness. On Enron emails, it reduces TF-IDF stylometric attribution by 73.2%, compared with 18.6% for the strongest masking baseline. Embedding and prototype attribution are also close to random. Together, these results show that *DiSan* reduces explicit and distributional leakage while retaining task utility.
%U https://aclanthology.org/2026.emnlp-main.1052/
%P 24148-24170
Markdown (Informal)
[Privacy-Preserving Text Sanitization for Distributed Agents Collaboration via Disentangled Representations](https://aclanthology.org/2026.emnlp-main.1052/) (Liu et al., EMNLP 2026)
ACL
- Xuan Liu, Hefeng Zhou, Sicheng Chen, Chao Yang, Xingcheng Xu, Jingjing Qu, Jiong Lou, Jie LI, and Xia Hu. 2026. Privacy-Preserving Text Sanitization for Distributed Agents Collaboration via Disentangled Representations. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 24148–24170, Budapest, Hungary. Association for Computational Linguistics.