<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="tsymboi-etal-2023-layerwise">
    <titleInfo>
        <title>Layerwise universal adversarial attack on NLP models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Olga</namePart>
        <namePart type="family">Tsymboi</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Danil</namePart>
        <namePart type="family">Malaev</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Andrei</namePart>
        <namePart type="family">Petrovskii</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ivan</namePart>
        <namePart type="family">Oseledets</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2023-07</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: ACL 2023</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Anna</namePart>
            <namePart type="family">Rogers</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Jordan</namePart>
            <namePart type="family">Boyd-Graber</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Naoaki</namePart>
            <namePart type="family">Okazaki</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Toronto, Canada</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>In this work, we examine the vulnerability of language models to universal adversarial triggers (UATs). We propose a new white-box approach to the construction of layerwise UATs (LUATs), which searches the triggers by perturbing hidden layers of a network. On the example of three transformer models and three datasets from the GLUE benchmark, we demonstrate that our method provides better transferability in a model-to-model setting with an average gain of 9.3% in the fooling rate over the baseline. Moreover, we investigate triggers transferability in the task-to-task setting. Using small subsets from the datasets similar to the target tasks for choosing a perturbed layer, we show that LUATs are more efficient than vanilla UATs by 7.1% in the fooling rate.</abstract>
    <identifier type="citekey">tsymboi-etal-2023-layerwise</identifier>
    <identifier type="doi">10.18653/v1/2023.findings-acl.10</identifier>
    <location>
        <url>https://aclanthology.org/2023.findings-acl.10</url>
    </location>
    <part>
        <date>2023-07</date>
        <extent unit="page">
            <start>129</start>
            <end>143</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Layerwise universal adversarial attack on NLP models
%A Tsymboi, Olga
%A Malaev, Danil
%A Petrovskii, Andrei
%A Oseledets, Ivan
%Y Rogers, Anna
%Y Boyd-Graber, Jordan
%Y Okazaki, Naoaki
%S Findings of the Association for Computational Linguistics: ACL 2023
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F tsymboi-etal-2023-layerwise
%X In this work, we examine the vulnerability of language models to universal adversarial triggers (UATs). We propose a new white-box approach to the construction of layerwise UATs (LUATs), which searches the triggers by perturbing hidden layers of a network. On the example of three transformer models and three datasets from the GLUE benchmark, we demonstrate that our method provides better transferability in a model-to-model setting with an average gain of 9.3% in the fooling rate over the baseline. Moreover, we investigate triggers transferability in the task-to-task setting. Using small subsets from the datasets similar to the target tasks for choosing a perturbed layer, we show that LUATs are more efficient than vanilla UATs by 7.1% in the fooling rate.
%R 10.18653/v1/2023.findings-acl.10
%U https://aclanthology.org/2023.findings-acl.10
%U https://doi.org/10.18653/v1/2023.findings-acl.10
%P 129-143

Markdown (Informal)

[Layerwise universal adversarial attack on NLP models](https://aclanthology.org/2023.findings-acl.10) (Tsymboi et al., Findings 2023)

• Layerwise universal adversarial attack on NLP models (Tsymboi et al., Findings 2023)

ACL

• Olga Tsymboi, Danil Malaev, Andrei Petrovskii, and Ivan Oseledets. 2023. Layerwise universal adversarial attack on NLP models. In Findings of the Association for Computational Linguistics: ACL 2023, pages 129–143, Toronto, Canada. Association for Computational Linguistics.