Tamp-X: Attacking explainable natural language classifiers through tampered activations

<p>While the technique of Deep Neural Networks (DNNs) has been instrumental in achieving state-of-the-art results for various Natural Language Processing (NLP) tasks, recent works have shown that the decisions made by DNNs cannot always be trusted. Recently Explainable Artificial Intelligence...

Full description

Saved in:
Bibliographic Details
Main Author: Hassan Ali (3348749) (author)
Other Authors: Muhammad Suleman Khan (17562612) (author), Ala Al-Fuqaha (4434340) (author), Junaid Qadir (16494902) (author)
Published: 2022
Subjects:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1864513535792381952
author Hassan Ali (3348749)
author2 Muhammad Suleman Khan (17562612)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
author2_role author
author
author
author_facet Hassan Ali (3348749)
Muhammad Suleman Khan (17562612)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
author_role author
dc.creator.none.fl_str_mv Hassan Ali (3348749)
Muhammad Suleman Khan (17562612)
Ala Al-Fuqaha (4434340)
Junaid Qadir (16494902)
dc.date.none.fl_str_mv 2022-09-01T00:00:00Z
dc.identifier.none.fl_str_mv 10.1016/j.cose.2022.102791
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Tamp-X_Attacking_explainable_natural_language_classifiers_through_tampered_activations/24745086
dc.rights.none.fl_str_mv CC BY 4.0
info:eu-repo/semantics/openAccess
dc.subject.none.fl_str_mv Information and computing sciences
Artificial intelligence
Machine learning
Explainable artificial intelligence (XAI)
Natural language processing
Attacking XAI
Adversarial attacks
Model tampering
dc.title.none.fl_str_mv Tamp-X: Attacking explainable natural language classifiers through tampered activations
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p>While the technique of Deep Neural Networks (DNNs) has been instrumental in achieving state-of-the-art results for various Natural Language Processing (NLP) tasks, recent works have shown that the decisions made by DNNs cannot always be trusted. Recently Explainable Artificial Intelligence (XAI) methods have been proposed as a method for increasing DNN’s reliability and trustworthiness. These XAI methods are however open to attack and can be manipulated in both white-box (gradient-based) and black-box (perturbation-based) scenarios. Exploring novel techniques to attack and robustify these XAI methods is crucial to fully understand these vulnerabilities. In this work, we propose Tamp-X—a novel attack which tampers the activations of robust NLP classifiers forcing the state-of-the-art white-box and black-box XAI methods to generate misrepresented explanations. To the best of our knowledge, in current NLP literature, we are the first to attack both the white-box and the black-box XAI methods simultaneously. We quantify the reliability of explanations based on three different metrics—the descriptive accuracy, the cosine similarity, and the L p norms of the explanation vectors. Through extensive experimentation, we show that the explanations generated for the tampered classifiers are not reliable, and significantly disagree with those generated for the untampered classifiers despite that the output decisions of tampered and untampered classifiers are almost always the same. Additionally, we study the adversarial robustness of the tampered NLP classifiers, and find out that the tampered classifiers which are harder to explain for the XAI methods, are also harder to attack by the adversarial attackers.</p><h2>Other Information</h2> <p> Published in: Computers & Security<br> License: <a href="http://creativecommons.org/licenses/by/4.0/" target="_blank">http://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1016/j.cose.2022.102791" target="_blank">https://dx.doi.org/10.1016/j.cose.2022.102791</a></p>
eu_rights_str_mv openAccess
id Manara2_6c5c6d3df74258de5b1ced3167df0e8e
identifier_str_mv 10.1016/j.cose.2022.102791
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/24745086
publishDate 2022
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Tamp-X: Attacking explainable natural language classifiers through tampered activationsHassan Ali (3348749)Muhammad Suleman Khan (17562612)Ala Al-Fuqaha (4434340)Junaid Qadir (16494902)Information and computing sciencesArtificial intelligenceMachine learningExplainable artificial intelligence (XAI)Natural language processingAttacking XAIAdversarial attacksModel tampering<p>While the technique of Deep Neural Networks (DNNs) has been instrumental in achieving state-of-the-art results for various Natural Language Processing (NLP) tasks, recent works have shown that the decisions made by DNNs cannot always be trusted. Recently Explainable Artificial Intelligence (XAI) methods have been proposed as a method for increasing DNN’s reliability and trustworthiness. These XAI methods are however open to attack and can be manipulated in both white-box (gradient-based) and black-box (perturbation-based) scenarios. Exploring novel techniques to attack and robustify these XAI methods is crucial to fully understand these vulnerabilities. In this work, we propose Tamp-X—a novel attack which tampers the activations of robust NLP classifiers forcing the state-of-the-art white-box and black-box XAI methods to generate misrepresented explanations. To the best of our knowledge, in current NLP literature, we are the first to attack both the white-box and the black-box XAI methods simultaneously. We quantify the reliability of explanations based on three different metrics—the descriptive accuracy, the cosine similarity, and the L p norms of the explanation vectors. Through extensive experimentation, we show that the explanations generated for the tampered classifiers are not reliable, and significantly disagree with those generated for the untampered classifiers despite that the output decisions of tampered and untampered classifiers are almost always the same. Additionally, we study the adversarial robustness of the tampered NLP classifiers, and find out that the tampered classifiers which are harder to explain for the XAI methods, are also harder to attack by the adversarial attackers.</p><h2>Other Information</h2> <p> Published in: Computers & Security<br> License: <a href="http://creativecommons.org/licenses/by/4.0/" target="_blank">http://creativecommons.org/licenses/by/4.0/</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1016/j.cose.2022.102791" target="_blank">https://dx.doi.org/10.1016/j.cose.2022.102791</a></p>2022-09-01T00:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1016/j.cose.2022.102791https://figshare.com/articles/journal_contribution/Tamp-X_Attacking_explainable_natural_language_classifiers_through_tampered_activations/24745086CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/247450862022-09-01T00:00:00Z
spellingShingle Tamp-X: Attacking explainable natural language classifiers through tampered activations
Hassan Ali (3348749)
Information and computing sciences
Artificial intelligence
Machine learning
Explainable artificial intelligence (XAI)
Natural language processing
Attacking XAI
Adversarial attacks
Model tampering
status_str publishedVersion
title Tamp-X: Attacking explainable natural language classifiers through tampered activations
title_full Tamp-X: Attacking explainable natural language classifiers through tampered activations
title_fullStr Tamp-X: Attacking explainable natural language classifiers through tampered activations
title_full_unstemmed Tamp-X: Attacking explainable natural language classifiers through tampered activations
title_short Tamp-X: Attacking explainable natural language classifiers through tampered activations
title_sort Tamp-X: Attacking explainable natural language classifiers through tampered activations
topic Information and computing sciences
Artificial intelligence
Machine learning
Explainable artificial intelligence (XAI)
Natural language processing
Attacking XAI
Adversarial attacks
Model tampering