<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="silva-etal-2024-look">
    <titleInfo>
        <title>Look Who’s Talking Now: Covert Channels From Biased LLMs</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Daniel</namePart>
        <namePart type="family">Silva</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Frederic</namePart>
        <namePart type="family">Sala</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Ryan</namePart>
        <namePart type="family">Gabrys</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2024-11</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Findings of the Association for Computational Linguistics: EMNLP 2024</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Yaser</namePart>
            <namePart type="family">Al-Onaizan</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Mohit</namePart>
            <namePart type="family">Bansal</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Yun-Nung</namePart>
            <namePart type="family">Chen</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Miami, Florida, USA</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Large language model-based steganography encodes hidden messages into model-generated tokens. The key tradeoff is between how much hidden information can be introduced and how much the model can be perturbed. To address this tradeoff, we show how to adapt strategies previously used for LLM watermarking to encode large amounts of information. We tackle the practical (but difficult) setting where we do not have access to the full model when trying to recover the hidden information. Theoretically, we study the fundamental limits in how much steganographic information can be inserted into LLM-created outputs. We provide practical encoding schemes and present experimental results showing that our proposed strategies are nearly optimal.</abstract>
    <identifier type="citekey">silva-etal-2024-look</identifier>
    <location>
        <url>https://aclanthology.org/2024.findings-emnlp.971</url>
    </location>
    <part>
        <date>2024-11</date>
        <extent unit="page">
            <start>16648</start>
            <end>16658</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Look Who’s Talking Now: Covert Channels From Biased LLMs
%A Silva, Daniel
%A Sala, Frederic
%A Gabrys, Ryan
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Findings of the Association for Computational Linguistics: EMNLP 2024
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F silva-etal-2024-look
%X Large language model-based steganography encodes hidden messages into model-generated tokens. The key tradeoff is between how much hidden information can be introduced and how much the model can be perturbed. To address this tradeoff, we show how to adapt strategies previously used for LLM watermarking to encode large amounts of information. We tackle the practical (but difficult) setting where we do not have access to the full model when trying to recover the hidden information. Theoretically, we study the fundamental limits in how much steganographic information can be inserted into LLM-created outputs. We provide practical encoding schemes and present experimental results showing that our proposed strategies are nearly optimal.
%U https://aclanthology.org/2024.findings-emnlp.971
%P 16648-16658

Markdown (Informal)

[Look Who’s Talking Now: Covert Channels From Biased LLMs](https://aclanthology.org/2024.findings-emnlp.971) (Silva et al., Findings 2024)

• Look Who’s Talking Now: Covert Channels From Biased LLMs (Silva et al., Findings 2024)

ACL

• Daniel Silva, Frederic Sala, and Ryan Gabrys. 2024. Look Who’s Talking Now: Covert Channels From Biased LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 16648–16658, Miami, Florida, USA. Association for Computational Linguistics.