Explain2Attack: Text Adversarial Attacks via Cross-Domain Interpretability

Hossam, Mahmoud; Le, Trung; Zhao, He; Phung, Dinh

doi:10.1109/icpr48806.2021.9412526

Cited by 5 publications

(7 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We employ target classifier predictions under these two settings on sentiment classification task using WordCNN and WordLSTM target classifiers. For all of the experiments, we used the same datasets, substitute datasets, and target classifier parameters and hyperparameters that were used in (Hossam et al, 2020).…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…To evaluate the performance of the two proposed methods described above, we report the adversarial accuracy (Adv Acc), the average number of queries (Adv Queries), and compare with the original Explain2Attack reported in (Hossam et al, 2020). In detail, in tables 1 and 2 we report the results for incorporating the target model predictions for training Explain2Attack selector without access to the target test set and with access to it, respectively.…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…For the first method, we use the exact setup as in (Hossam et al, 2020) for the experiments and the datasets, except that the substitute labels Y b were replaced with the target classifier predictions from the substitute sentences, and Eq. ( 1) was used for training the selector.…”

Section: Experiments and Discussionmentioning

confidence: 99%

“…In addition, the increased number of queries to the target classifier is not desirable in blackbox settings, where it can raise the suspicion towards the attacking agent. Recently, several methods were introduced to improve the query efficiency, including (Garg & Ramakrishnan, 2020;Li et al, 2020) and Explain2Attack (Hossam et al, 2020), where an interpretable model was trained to learn word importance ranking for synonym replacement.…”

Section: Introductionmentioning

confidence: 99%

“…In this paper, we propose two modifications to this interpretable approach to better train the selector network while benefiting from available target information. Specifically, we do not use the substitute dataset labels used for selector training as in (Hossam et al, 2020). Instead, we replace them with output predictions from the target classifier under two settings: the target model is given the substitute dataset sentences X b ∈ D b , or the target model is given the target test set…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Improved and Efficient Text Adversarial Attacks using Target Information

Hossam,

Le,

Zhao

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

There has been recently a growing interest in studying adversarial examples on natural language models in the black-box setting. These methods attack natural language classifiers by perturbing certain important words until the classifier label is changed. In order to find these important words, these methods rank all words by importance by querying the target model word by word for each input sentence, resulting in high query inefficiency. A new interesting approach was introduced that addresses this problem through interpretable learning to learn the word ranking instead of previous expensive search. The main advantage of using this approach is that it achieves comparable attack rates to the state-of-the-art methods, yet faster and with fewer queries, where fewer queries are desirable to avoid suspicion towards the attacking agent. Nonetheless, this approach sacrificed the useful information that could be leveraged from the target classifier for the sake of query efficiency. In this paper we study the effect of leveraging the target model outputs and data on both attack rates and average number of queries, and we show that both can be improved, with a limited overhead of additional queries.

show abstract

Section: Experiments and Discussionmentioning

confidence: 99%

Section: Experiments and Discussionmentioning

confidence: 99%

Section: Experiments and Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Improved and Efficient Text Adversarial Attacks using Target Information

Hossam,

Le,

Zhao

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Adversarial Attacks in Explainable Machine Learning: A Survey of Threats Against Models and Humans

Vadillo,

Santana,

Lozano

2024

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Reliable deployment of machine learning models such as neural networks continues to be challenging due to several limitations. Some of the main shortcomings are the lack of interpretability and the lack of robustness against adversarial examples or out‐of‐distribution inputs. In this paper, we comprehensively review the possibilities and limits of adversarial attacks for explainable machine learning models. First, we extend the notion of adversarial examples to fit in explainable machine learning scenarios where a human assesses not only the input and the output classification, but also the explanation of the model's decision. Next, we propose a comprehensive framework to study whether (and how) adversarial examples can be generated for explainable models under human assessment. Based on this framework, we provide a structured review of the diverse attack paradigms existing in this domain, identify current gaps and future research directions, and illustrate the main attack paradigms discussed. Furthermore, our framework considers a wide range of relevant yet often ignored factors such as the type of problem, the user expertise or the objective of the explanations, in order to identify the attack strategies that should be adopted in each scenario to successfully deceive the model (and the human). The intention of these contributions is to serve as a basis for a more rigorous and realistic study of adversarial examples in the field of explainable machine learning.

show abstract

An adversarial-example generation method for Chinese sentiment tendency classification based on audiovisual confusion and contextual association

Tian

et al. 2023

Knowl Inf Syst

View full text Add to dashboard Cite

Explain2Attack: Text Adversarial Attacks via Cross-Domain Interpretability

Cited by 5 publications

References 5 publications

Improved and Efficient Text Adversarial Attacks using Target Information

Improved and Efficient Text Adversarial Attacks using Target Information

Adversarial Attacks in Explainable Machine Learning: A Survey of Threats Against Models and Humans

An adversarial-example generation method for Chinese sentiment tendency classification based on audiovisual confusion and contextual association

Contact Info

Product

Resources

About