arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2012.13339cs.CL
收藏

一种上下文感知的自然语言攻击生成方法

A Context Aware Approach for Generating Natural Language Attacks

International Institute of Information Technology, Hyderabad(国际信息技术研究所海得拉巴分校);Data Sciences and Analytics Center(数据科学与分析中心);Kohli Center on Intelligent Systems(科利智能系统中心)

Rishabh Maheshwary, Saket Maheshwary, Vikram Pudi

摘要:

我们研究的一个重要任务是在黑盒环境中攻击自然语言处理模型。我们提出了一种攻击策略,在文本分类和蕴涵任务上制作语义相似的对抗性示例。我们提出的攻击通过考虑原始单词及其周围上下文的信息来找到候选单词。它联合利用掩蔽语言建模和下一句预测来进行上下文理解。与先前文献中提出的攻击相比,我们能够生成高质量的对抗性示例,这些示例在成功率和单词扰动百分比方面都有显着提高。

英文摘要:

We study an important task of attacking natural language processing models in a black box setting. We propose an attack strategy that crafts semantically similar adversarial examples on text classification and entailment tasks. Our proposed attack finds candidate words by considering the information of both the original word and its surrounding context. It jointly leverages masked language modelling and next sentence prediction for context understanding. In comparison to attacks proposed in prior literature, we are able to generate high quality adversarial examples that do significantly better both in terms of success rate and word perturbation percentage.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑