logo
科言猫学术研究的AI总结
language
手机端手机端
  • Phone Feature微信小程序
简体中文简体中文
notice消息通知
博客博客
delete
arrow
返回

Hybrid backdoor attacks for deep code models

delete2025-12-08
delete0
PRE
AI
T
Tongcheng Geng *
DOI：10.1371/journal.pone.0338083delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
深度代码模型面临后门攻击带来的安全漏洞。先前的方法主要依赖单一触发机制，导致隐蔽性有限且易受防御策略影响。本文提出了一种新型混合后门攻击方法，将函数签名特征和死代码插入作为触发器。我们的方法利用这些触发器的互补特性，创造出协同效应，在增强攻击有效性的同时保持隐蔽性。实验表明，在极低的污染率下，我们的混合方法实现了高攻击成功率（ASR），显著优于单一触发机制的方法。该混合方法有效规避了基于频谱特征的防御，同时保持对正常功能的最低影响。我们的发现强调了针对深度代码模型中复杂混合后门攻击的专用防御机制亟待开发的迫切性。
Keyword:
hybrid backdoor attack
deep code models
function signature
dead code insertion
stealth attack

期刊

PLoS One 封面图
PLoS One
IF:
2.6
论文数:
2.6W
被引数:
81.6W

机构

暂无机构信息
引用论文

引用论文

A review of backdoor attacks and defenses in code large language models: Implications for security measures
errINFORMATION AND SOFTWARE TECHNOLOGY
IF4.3
err2025-03-01
err0
PREAI
errQu, Yubin; Huang, Song; Nie, Peng
err分享
err收藏
CodeT5: Identifier-aware Unified Pre-trained Encoder-Decoder Models for Code Understanding and GenerationCodeT5: 用于代码理解和生成的标识符感知的统一预训练编码器-解码器模型
errProceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
IF0
err2021-01-01
err0
errOAAI
errYue Wang; Weishi Wang; Shafiq Joty; Steven C.H. Hoi
err分享
err收藏
Examining Zero-Shot Vulnerability Repair with Large Language Models
err2023 IEEE Symposium on Security and Privacy (SP)
IF0
err2023-05-01
err0
errOAAI
errHammond Pearce; Benjamin Tan; Baleegh Ahmad; Ramesh Karri; Brendan Dolan-Gavitt
err分享
err收藏
LateBA: Latent Backdoor Attack on Deep Bug Search via Infrequent Execution Codes
errProceedings of the 15th Asia-Pacific Symposium on Internetware
IF0
err2024-07-24
err0
PREAI
errYi,Xiaoyu; Li,Gaolei; Huang,Wenkai; Lin,Xi; Li,Jianhua; Liu,Yuchen
err分享
err收藏
Detection of backdoor attacks using targeted universal adversarial perturbations for deep neural networks
errJOURNAL OF SYSTEMS AND SOFTWARE
IF4.1
err2024-01-01
err2
PREAI
errQu, Yubin; Huang, Song; Chen, Xiang; Wang, Xingya; Yao, Yongming
err分享
err收藏
Deep learning type inference
errProceedings of the 2018 26th ACM Joint Meeting on European Software Engineering Conference and Symposium on the Foundations of Software Engineering
IF0
err2018-10-26
err0
errOAAI
errVincent J. Hellendoorn; Christian Bird; Earl T. Barr; Miltiadis Allamanis
err分享
err收藏
TrojanPuzzle: Covertly Poisoning Code-Suggestion ModelsTrojanPuzzle：隐蔽地污染代码建议模型
err2024 IEEE Symposium on Security and Privacy (SP)
IF0
err2024-05-19
err0
PREAI
errHojjat Aghakhani; Wei Dai; Andre Manoel; Xavier Fernandes; Anant Kharkar; Christopher Kruegel; Giovanni Vigna; David Evans; Ben Zorn; Robert Sim
err分享
err收藏
Code completion with statistical language models
errProceedings of the 35th ACM SIGPLAN Conference on Programming Language Design and Implementation
IF0
err2014-06-09
err0
PREAI
errVeselin Raychev; Martin Vechev; Eran Yahav
err分享
err收藏
CodeBERT: A Pre-Trained Model for Programming and Natural LanguagesCodeBERT: 编程和自然语言的预训练模型
errFindings of the Association for Computational Linguistics: EMNLP 2020
IF0
err2020-01-01
err0
errOAAI
errZhangyin Feng; Daya Guo; Duyu Tang; Nan Duan; Xiaocheng Feng; Ming Gong; Linjun Shou; Bing Qin; Ting Liu; Daxin Jiang; Ming Zhou
err分享
err收藏
A survey on robustness attacks for deep code models关于深度代码模型鲁棒性攻击的调查
errAutomated Software Engineering
IF0
err2024-08-09
err0
PREAI
errYubin Qu; Song Huang; Yongming Yao
err分享
err收藏
学者 查看更多内容