<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vakili-dalianis-2023-using">
    <titleInfo>
        <title>Using Membership Inference Attacks to Evaluate Privacy-Preserving Language Modeling Fails for Pseudonymizing Data</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Thomas</namePart>
        <namePart type="family">Vakili</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hercules</namePart>
        <namePart type="family">Dalianis</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2023-05</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa)</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Tanel</namePart>
            <namePart type="family">Alumäe</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mark</namePart>
            <namePart type="family">Fishel</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>University of Tartu Library</publisher>
            <place>
                <placeTerm type="text">Tórshavn, Faroe Islands</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Large pre-trained language models dominate the current state-of-the-art for many natural language processing applications, including the field of clinical NLP. Several studies have found that these can be susceptible to privacy attacks that are unacceptable in the clinical domain where personally identifiable information (PII) must not be exposed. However, there is no consensus regarding how to quantify the privacy risks of different models. One prominent suggestion is to quantify these risks using membership inference attacks. In this study, we show that a state-of-the-art membership inference attack on a clinical BERT model fails to detect the privacy benefits from pseudonymizing data. This suggests that such attacks may be inadequate for evaluating token-level privacy preservation of PIIs.</abstract>
    <identifier type="citekey">vakili-dalianis-2023-using</identifier>
    <location>
        <url>https://aclanthology.org/2023.nodalida-1.33</url>
    </location>
    <part>
        <date>2023-05</date>
        <extent unit="page">
            <start>318</start>
            <end>323</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Using Membership Inference Attacks to Evaluate Privacy-Preserving Language Modeling Fails for Pseudonymizing Data
%A Vakili, Thomas
%A Dalianis, Hercules
%Y Alumäe, Tanel
%Y Fishel, Mark
%S Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa)
%D 2023
%8 May
%I University of Tartu Library
%C Tórshavn, Faroe Islands
%F vakili-dalianis-2023-using
%X Large pre-trained language models dominate the current state-of-the-art for many natural language processing applications, including the field of clinical NLP. Several studies have found that these can be susceptible to privacy attacks that are unacceptable in the clinical domain where personally identifiable information (PII) must not be exposed. However, there is no consensus regarding how to quantify the privacy risks of different models. One prominent suggestion is to quantify these risks using membership inference attacks. In this study, we show that a state-of-the-art membership inference attack on a clinical BERT model fails to detect the privacy benefits from pseudonymizing data. This suggests that such attacks may be inadequate for evaluating token-level privacy preservation of PIIs.
%U https://aclanthology.org/2023.nodalida-1.33
%P 318-323

Markdown (Informal)

[Using Membership Inference Attacks to Evaluate Privacy-Preserving Language Modeling Fails for Pseudonymizing Data](https://aclanthology.org/2023.nodalida-1.33) (Vakili & Dalianis, NoDaLiDa 2023)

• Using Membership Inference Attacks to Evaluate Privacy-Preserving Language Modeling Fails for Pseudonymizing Data (Vakili & Dalianis, NoDaLiDa 2023)

ACL

• Thomas Vakili and Hercules Dalianis. 2023. Using Membership Inference Attacks to Evaluate Privacy-Preserving Language Modeling Fails for Pseudonymizing Data. In Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa), pages 318–323, Tórshavn, Faroe Islands. University of Tartu Library.