Guardado en:
Detalles Bibliográficos
Autor principal: Rusert, Jonathan
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2404.08538
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910407785971712
author Rusert, Jonathan
author_facet Rusert, Jonathan
contents Text classification systems have continuously improved in performance over the years. However, nearly all current SOTA classifiers have a similar shortcoming, they process text in a horizontal manner. Vertically written words will not be recognized by a classifier. In contrast, humans are easily able to recognize and read words written both horizontally and vertically. Hence, a human adversary could write problematic words vertically and the meaning would still be preserved to other humans. We simulate such an attack, VertAttack. VertAttack identifies which words a classifier is reliant on and then rewrites those words vertically. We find that VertAttack is able to greatly drop the accuracy of 4 different transformer models on 5 datasets. For example, on the SST2 dataset, VertAttack is able to drop RoBERTa's accuracy from 94 to 13%. Furthermore, since VertAttack does not replace the word, meaning is easily preserved. We verify this via a human study and find that crowdworkers are able to correctly label 77% perturbed texts perturbed, compared to 81% of the original texts. We believe VertAttack offers a look into how humans might circumvent classifiers in the future and thus inspire a look into more robust algorithms.
format Preprint
id arxiv_https___arxiv_org_abs_2404_08538
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle VertAttack: Taking advantage of Text Classifiers' horizontal vision
Rusert, Jonathan
Computation and Language
I.2.7
Text classification systems have continuously improved in performance over the years. However, nearly all current SOTA classifiers have a similar shortcoming, they process text in a horizontal manner. Vertically written words will not be recognized by a classifier. In contrast, humans are easily able to recognize and read words written both horizontally and vertically. Hence, a human adversary could write problematic words vertically and the meaning would still be preserved to other humans. We simulate such an attack, VertAttack. VertAttack identifies which words a classifier is reliant on and then rewrites those words vertically. We find that VertAttack is able to greatly drop the accuracy of 4 different transformer models on 5 datasets. For example, on the SST2 dataset, VertAttack is able to drop RoBERTa's accuracy from 94 to 13%. Furthermore, since VertAttack does not replace the word, meaning is easily preserved. We verify this via a human study and find that crowdworkers are able to correctly label 77% perturbed texts perturbed, compared to 81% of the original texts. We believe VertAttack offers a look into how humans might circumvent classifiers in the future and thus inspire a look into more robust algorithms.
title VertAttack: Taking advantage of Text Classifiers' horizontal vision
topic Computation and Language
I.2.7
url https://arxiv.org/abs/2404.08538