Contextualized Perturbation for Textual Adversarial Attack
arXiv:2009.07502
Abstract
Adversarial examples expose the vulnerabilities of natural language processing (NLP) models, and can be used to evaluate and improve their robustness. Existing techniques of generating such examples are typically driven by local heuristic rules that are agnostic to the context, often resulting in unnatural and ungrammatical outputs. This paper presents CLARE, a ContextuaLized AdversaRial Example generation model that produces fluent and grammatical outputs through a mask-then-infill procedure. CLARE builds on a pre-trained masked language model and modifies the inputs in a context-aware manner. We propose three contextualized perturbations, Replace, Insert and Merge, allowing for generating outputs of varied lengths. With a richer range of available strategies, CLARE is able to attack a victim model more efficiently with fewer edits. Extensive experiments and human evaluation demonstrate that CLARE outperforms the baselines in terms of attack success rate, textual similarity, fluency and grammaticality.
arXiv commentAccepted by NAACL 2021, long paper
References in corpus (12)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Universal Sentence Encoder
- Non-Autoregressive Neural Machine Translation
- BAE: BERT-based Adversarial Examples for Text Classification
- Generating Natural Adversarial Examples
- Deep Text Classification Can be Fooled
- Towards Crafting Text Adversarial Samples
- Reevaluating Adversarial Examples in Natural Language
- Weight Poisoning Attacks on Pre-trained Models
- Natural Language Adversarial Defense through Synonym Encoding
- POINTER: Constrained Progressive Text Generation via Insertion-based Generative Pre-training
- T3: Tree-Autoencoder Constrained Adversarial Text Generation for Targeted Attack
Cited by in corpus (6)
- BAE: BERT-based Adversarial Examples for Text Classification
- Speaker-Oriented Latent Structures for Dialogue-Based Relation Extraction
- Sometimes We Want Translationese
- Preserving Semantics in Textual Adversarial Attacks
- Improving Text Generation with Student-Forcing Optimal Transport
- Text Counterfactuals via Latent Optimization and Shapley-Guided Search