Attribution-guided Adversarial Code Prompt Generation for Code Completion Models
Large language models have made significant progress in code completion, which may further remodel future software development. However, these code completion models are found to be highly risky as they may introduce vulnerabilities unintentionally or be induced by a special input, i.e., adversarial code prompt. Prior studies mainly focus on the robustness of these models, but their security has not been fully analyzed.
Authors
- Guozhu Meng (ORCID: https://orcid.org/0000-0001-6388-2571)
- Shangqing Liu (ORCID: https://orcid.org/0000-0002-5598-4006)
- Xueyang Li (ORCID: https://orcid.org/0000-0001-8687-3926)
- Yang Liu (ORCID: https://orcid.org/0000-0001-7300-9215)
- Xiapu Luo (ORCID: https://orcid.org/0000-0002-9082-3208)
- Kai Chen (ORCID: https://orcid.org/0000-0002-5624-2987)
- Lu Xiang (ORCID: https://orcid.org/0000-0002-7702-7274)
- Kun Sun
Institutions
- Hong Kong Polytechnic University (HK)
- Nanyang Technological University (SG)
- Chinese Academy of Sciences (CN)
- Institute of Information Engineering (CN)
Publication Details
- Published
- 2024-10-18
- DOI
- https://doi.org/10.1145/3691620.3695517
- Citations
- 3
- Primary Topic
- Adversarial Robustness in Machine Learning
- Type
- article
- Field-Weighted Citation Impact
- 1.04