Deep neural networks (DNNs) are vulnerable to adversarial attacks, in which a small perturbation to samples can cause misclassification. However, how to select important words for textual attack models is a big challenge. Therefore, in this paper, an innovative score-based attack model is proposed to solve the important words selection problem for textual attack models. To this end, the generation of semantically adversarial examples in this model is adopted to mislead a text classification model. Then, this model integrates the self-attention mechanism and confidence probabilities for the selection of the important words. Moreover, an alternative model similar to the transfer attack is introduced to reflect the correlation degree of words inside the texts. Finally, adversarial training experimental results demonstrate the superiority of the proposed model. 

Aliasing black box adversarial attack with joint self-attention distribution and confidence probability

Despite great success on many machine learning tasks, deep neural networks are still vulnerable to adversarial samples. While gradient-based adversarial attack methods are well-explored in the field of computer vision, it is impractical to directly apply them in natural language processing due to the discrete nature of text. To bridge this gap, we propose a general framework to adapt existing gradient-based methods to craft textual adversarial samples. In this framework, gradient-based continuous perturbations are added to the embedding layer and are amplified in the forward propagation process. Then the final perturbed latent representations are decoded with a mask language model head to obtain potential adversarial samples. In this paper, we instantiate our framework with \textbf{T}extual \textbf{P}rojected \textbf{G}radient \textbf{D}escent (\textbf{TPGD}). We conduct comprehensive experiments to evaluate our framework by performing transfer black-box attacks on BERT, RoBERTa and ALBERT on three benchmark datasets. Experimental results demonstrate our method achieves an overall better performance and produces more fluent and grammatical adversarial samples compared to strong baseline methods. All the code and data will be made public.

Bridge the Gap Between CV and NLP! A Gradient-based Textual Adversarial Attack Framework.

AbstractWord substitution based textual adversarial attack is actually a combinatorial optimization problem. Existing greedy search methods are time-consuming due to extensive unnecessary victim model calls in word ranking and substitution. In this work, we propose a learnable attack method which uses neural networks to guide the greedy search to reduce victim model calls. Specifically, we use one network to predict the importance of each word, without accessing the victim model. It avoids the victim model calls proportional to the length of the text, which may be in hundreds. Moreover, we use the other network to score each substitute word for a specific substitute position and filter out the attack-inefficient and out-of-context ones, so as to reduce substitution attempts. We evaluate our method on sentiment analysis, natural language inference and paraphrase identification tasks. Experimental results show that our method can achieve higher attack success rate and adversarial example quality than the baseline methods, while requiring less computation overhead. The code is public at https://github.com/CMACH508/PredictiveTextualAttack.KeywordsTextual adversarial attackLearnable adversarial attackAdversarial robustnessDeep learning 

Learning to Generate Textual Adversarial Examples

Adversarial attacks expose the vulnerability of deep neural networks. Compared to image adversarial attacks, textual adversarial attacks are more challenging due to the discrete nature of texts. Recent synonym‐based methods achieve the current state‐of‐the‐art results. However, these methods introduce new words against the original text, leading to that humans easily perceive the difference between the adversarial example and the original text. Motivated by the fact that humans are usually unaware of chaotic word order in some cases, we propose exchange‐attack (EA), a concise and effective word‐level textual adversarial attack model. Specifically, the EA model generates adversarial examples by exchanging words of the original text itself according to the contributions that these words make regarding classification results. Intuitively, the smaller the distance between the two exchanged words, the more difficult the chaotic word order to be perceived by humans. We thus take the word distance into consideration when generating the chaotic word orders. Extensive experiments on several text classification data sets show that the EA model consistently outperforms the selected baselines in terms of averaged after‐attack accuracy, modification rate, query number, and semantic similarity. And human evaluation results reveal that humans difficultly perceive the adversarial examples generated by the EA model. In addition, quantitative and qualitative analyses further validate the effectiveness of the EA model, including that the generated adversarial examples are grammatically correct and semantically preserved.

Textual adversarial attacks by exchanging text‐self words

Recently, the textual adversarial attack models become increasingly popular due to their successful in estimating the robustness of NLP models. However, existing works have obvious deficiencies. (1) They usually consider only a single granularity of modification strategies (e.g. word-level or sentence-level), which is insufficient to explore the holistic textual space for generation; (2) They need to query victim models hundreds of times to make a successful attack, which is highly inefficient in practice. To address such problems, in this paper we propose MAYA, a Multi-grAnularitY Attack model to effectively generate high-quality adversarial samples with fewer queries to victim models. Furthermore, we propose a reinforcement-learning based method to train a multi-granularity attack agent through behavior cloning with the expert knowledge from our MAYA algorithm to further reduce the query times. Additionally, we also adapt the agent to attack black-box models that only output labels without confidence scores. We conduct comprehensive experiments to evaluate our attack models by attacking BiLSTM, BERT and RoBERTa in two different black-box attack settings and three benchmark datasets. Experimental results show that our models achieve overall better attacking performance and produce more fluent and grammatical adversarial samples compared to baseline models. Besides, our adversarial attack agent significantly reduces the query times in both attack settings. Our codes are released at this https URL.

Multi-granularity Textual Adversarial Attack with Behavior Cloning

Recently, the textual adversarial attack models become increasingly popular due to their successful in estimating the robustness of NLP models. However, existing works have obvious deficiencies. (1)They usually consider only a single granularity of modification strategies (e.g. word-level or sentence-level), which is insufficient to explore the holistic textual space for generation; (2) They need to query victim models hundreds of times to make a successful attack, which is highly inefficient in practice. To address such problems, in this paper we propose MAYA, a Multi-grAnularitY Attack model to effectively generate high-quality adversarial samples with fewer queries to victim models. Furthermore, we propose a reinforcement-learning based method to train a multi-granularity attack agent through behavior cloning with the expert knowledge from our MAYA algorithm to further reduce the query times. Additionally, we also adapt the agent to attack black-box models that only output labels without confidence scores. We conduct comprehensive experiments to evaluate our attack models by attacking BiLSTM, BERT and RoBERTa in two different black-box attack settings and three benchmark datasets. Experimental results show that our models achieve overall better attacking performance and produce more fluent and grammatical adversarial samples compared to baseline models. Besides, our adversarial attack agent significantly reduces the query times in both attack settings. Our codes are released at https://github.com/Yangyi-Chen/MAYA.

Yangyi Chen

Papers

Multi-granularity Textual Adversarial Attack with Behavior Cloning

Multi-granularity Textual Adversarial Attack with Behavior Cloning.

Bridge the Gap Between CV and NLP! A Gradient-based Textual Adversarial Attack Framework.