Defense against adversarial attacks: robust and efficient compressed optimized neural networks

<p dir="ltr">In the ongoing battle against adversarial attacks, adopting a suitable strategy to enhance model efficiency, bolster resistance to adversarial threats, and ensure practical deployment is crucial. To achieve this goal, a novel four-component methodology is introduced. Fir...

وصف كامل

محفوظ في:
التفاصيل البيبلوغرافية
المؤلف الرئيسي: Insaf Kraidia (19198012) (author)
مؤلفون آخرون: Afifa Ghenai (19198015) (author), Samir Brahim Belhaouari (9427347) (author)
منشور في: 2024
الموضوعات:
الوسوم: إضافة وسم
لا توجد وسوم, كن أول من يضع وسما على هذه التسجيلة!
_version_ 1864513510325616640
author Insaf Kraidia (19198012)
author2 Afifa Ghenai (19198015)
Samir Brahim Belhaouari (9427347)
author2_role author
author
author_facet Insaf Kraidia (19198012)
Afifa Ghenai (19198015)
Samir Brahim Belhaouari (9427347)
author_role author
dc.creator.none.fl_str_mv Insaf Kraidia (19198012)
Afifa Ghenai (19198015)
Samir Brahim Belhaouari (9427347)
dc.date.none.fl_str_mv 2024-03-01T00:00:00Z
dc.identifier.none.fl_str_mv 10.1038/s41598-024-56259-z
dc.relation.none.fl_str_mv https://figshare.com/articles/journal_contribution/Defense_against_adversarial_attacks_robust_and_efficient_compressed_optimized_neural_networks/26355052
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
Adversarial attacks
Generative pre-trained transformer (GPT)
Compression
Multi expert
dc.title.none.fl_str_mv Defense against adversarial attacks: robust and efficient compressed optimized neural networks
dc.type.none.fl_str_mv Text
Journal contribution
info:eu-repo/semantics/publishedVersion
text
contribution to journal
description <p dir="ltr">In the ongoing battle against adversarial attacks, adopting a suitable strategy to enhance model efficiency, bolster resistance to adversarial threats, and ensure practical deployment is crucial. To achieve this goal, a novel four-component methodology is introduced. First, introducing a pioneering batch-cumulative approach, the exponential particle swarm optimization (ExPSO) algorithm was developed for meticulous parameter fine-tuning within each batch. A cumulative updating loss function was employed for overall optimization, demonstrating remarkable superiority over traditional optimization techniques. Second, weight compression is applied to streamline the deep neural network (DNN) parameters, boosting the storage efficiency and accelerating inference. It also introduces complexity to deter potential attackers, enhancing model accuracy in adversarial settings. This study compresses the generative pre-trained transformer (GPT) by 65%, saving time and memory without causing performance loss. Compared to state-of-the-art methods, the proposed method achieves the lowest perplexity (14.28), the highest accuracy (93.72%), and an 8 × speedup in the central processing unit. The integration of the preceding two components involves the simultaneous training of multiple versions of the compressed GPT. This training occurs across various compression rates and different segments of a dataset and is ultimately associated with a novel multi-expert architecture. This enhancement significantly fortifies the model's resistance to adversarial attacks by introducing complexity into attackers' attempts to anticipate the model's prediction integration process. Consequently, this leads to a remarkable average performance improvement of 25% across 14 different attack scenarios and various datasets, surpassing the capabilities of current state-of-the-art methods.</p><h2>Other Information</h2><p dir="ltr">Published in: Scientific Reports<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1038/s41598-024-56259-z" target="_blank">https://dx.doi.org/10.1038/s41598-024-56259-z</a></p>
eu_rights_str_mv openAccess
id Manara2_f95853b9211a9a391f998a374deb322e
identifier_str_mv 10.1038/s41598-024-56259-z
network_acronym_str Manara2
network_name_str Manara2
oai_identifier_str oai:figshare.com:article/26355052
publishDate 2024
repository.mail.fl_str_mv
repository.name.fl_str_mv
repository_id_str
rights_invalid_str_mv CC BY 4.0
spelling Defense against adversarial attacks: robust and efficient compressed optimized neural networksInsaf Kraidia (19198012)Afifa Ghenai (19198015)Samir Brahim Belhaouari (9427347)Information and computing sciencesArtificial intelligenceMachine learningAdversarial attacksGenerative pre-trained transformer (GPT)CompressionMulti expert<p dir="ltr">In the ongoing battle against adversarial attacks, adopting a suitable strategy to enhance model efficiency, bolster resistance to adversarial threats, and ensure practical deployment is crucial. To achieve this goal, a novel four-component methodology is introduced. First, introducing a pioneering batch-cumulative approach, the exponential particle swarm optimization (ExPSO) algorithm was developed for meticulous parameter fine-tuning within each batch. A cumulative updating loss function was employed for overall optimization, demonstrating remarkable superiority over traditional optimization techniques. Second, weight compression is applied to streamline the deep neural network (DNN) parameters, boosting the storage efficiency and accelerating inference. It also introduces complexity to deter potential attackers, enhancing model accuracy in adversarial settings. This study compresses the generative pre-trained transformer (GPT) by 65%, saving time and memory without causing performance loss. Compared to state-of-the-art methods, the proposed method achieves the lowest perplexity (14.28), the highest accuracy (93.72%), and an 8 × speedup in the central processing unit. The integration of the preceding two components involves the simultaneous training of multiple versions of the compressed GPT. This training occurs across various compression rates and different segments of a dataset and is ultimately associated with a novel multi-expert architecture. This enhancement significantly fortifies the model's resistance to adversarial attacks by introducing complexity into attackers' attempts to anticipate the model's prediction integration process. Consequently, this leads to a remarkable average performance improvement of 25% across 14 different attack scenarios and various datasets, surpassing the capabilities of current state-of-the-art methods.</p><h2>Other Information</h2><p dir="ltr">Published in: Scientific Reports<br>License: <a href="https://creativecommons.org/licenses/by/4.0" target="_blank">https://creativecommons.org/licenses/by/4.0</a><br>See article on publisher's website: <a href="https://dx.doi.org/10.1038/s41598-024-56259-z" target="_blank">https://dx.doi.org/10.1038/s41598-024-56259-z</a></p>2024-03-01T00:00:00ZTextJournal contributioninfo:eu-repo/semantics/publishedVersiontextcontribution to journal10.1038/s41598-024-56259-zhttps://figshare.com/articles/journal_contribution/Defense_against_adversarial_attacks_robust_and_efficient_compressed_optimized_neural_networks/26355052CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/263550522024-03-01T00:00:00Z
spellingShingle Defense against adversarial attacks: robust and efficient compressed optimized neural networks
Insaf Kraidia (19198012)
Information and computing sciences
Artificial intelligence
Machine learning
Adversarial attacks
Generative pre-trained transformer (GPT)
Compression
Multi expert
status_str publishedVersion
title Defense against adversarial attacks: robust and efficient compressed optimized neural networks
title_full Defense against adversarial attacks: robust and efficient compressed optimized neural networks
title_fullStr Defense against adversarial attacks: robust and efficient compressed optimized neural networks
title_full_unstemmed Defense against adversarial attacks: robust and efficient compressed optimized neural networks
title_short Defense against adversarial attacks: robust and efficient compressed optimized neural networks
title_sort Defense against adversarial attacks: robust and efficient compressed optimized neural networks
topic Information and computing sciences
Artificial intelligence
Machine learning
Adversarial attacks
Generative pre-trained transformer (GPT)
Compression
Multi expert