一句话：揭示神经代码搜索模型对投毒攻击的脆弱性，呼吁关注其鲁棒性。
论文背景是：基于自然语言查询从开源软件仓库中搜索并复用代码片段，能显著提升编程效率，而基于深度学习的代码搜索方法近年日益流行。尽管训练准确率已有长足进步，这些模型的鲁棒性此前却很少受到关注。标题进一步指向神经代码搜索中的投毒漏洞（poisoning vulnerabilities）。因此，本文关注的核心问题是：神经代码搜索模型在恶意投毒场景下是否可靠，以及这类脆弱性为何值得重视。由于所给摘要仅提供背景与问题动机，未展开具体攻击构造、防御机制或实验评估细节，无法从摘要进一步概括方法与关键结果。
Searching and reusing code snippets from open-source software repositories based on natural-language queries can greatly improve programming productivity.Recently, deep-learning-based approaches have become increasingly popular for code search. Despite substantial progress in training accurate models of code search, the robustness of these models has received little attention so far.
DOI 原文 ·
@inproceedings{paperbot7081,
title = {You see what I want you to see: poisoning vulnerabilities in neural code search},
author = {Yao Wan and Shijie Zhang and Hongyu Zhang and Yulei Sui and Guandong Xu and Dezhong Yao and Hai Jin and Lichao Sun},
booktitle = {FSE},
year = {2022},
doi = {10.1145/3540250.3549153}
}