Research map / Adversarial robustness and security / Black-box and transfer attacks
CodeAttack: Code-Based Adversarial Attacks for Pre-trained Programming Language Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use CodeAttack instead of BERT-Attack | CodeT5 · C#-Java translation | Δdrop (CodeBLEU): 10.98 -> 12.27 | Helped |
| Use CodeAttack instead of BERT-Attack | CodeBERT · C#-Java translation | Δdrop (CodeBLEU): 12.36 -> 17.03 | Helped |
| Use CodeAttack instead of BERT-Attack | GraphCodeBERT · C#-Java translation | Δdrop (CodeBLEU): 30.26 -> 27.99 | Hurt |
| Use CodeAttack instead of BERT-Attack | CodeT5 · Java repair (small) | Δdrop (CodeBLEU): 8.43 -> 7.92 | Hurt |
| Use CodeAttack instead of BERT-Attack | CodeBERT · Java repair (small) | Δdrop (CodeBLEU): 9.38 -> 9.31 | Hurt |
| Use CodeAttack instead of BERT-Attack | GraphCodeBERT · Java repair (small) | Δdrop (CodeBLEU): 8.83 -> 10.19 | Helped |
| Use CodeAttack instead of BERT-Attack | CodeT5 · PHP summarization | Δdrop (BLEU): 8.70 -> 9.59 | Helped |
| Use CodeAttack instead of BERT-Attack | CodeBERT · PHP summarization | Δdrop (BLEU): 8.35 -> 8.87 | Helped |
| Use CodeAttack instead of BERT-Attack | RoBERTa · PHP summarization | Δdrop (BLEU): 7.71 -> 8.08 | Helped |
| Select high-influence vulnerable tokens rather than random tokens (ablation) | CodeAttack · C#-Java translation | Δdrop | Helped |
| Add token-level code constraints to vulnerable-token attacks with operator constraints (ablation) | CodeAttack · C#-Java translation | Δdrop | Helped |
| Restrict vulnerable-token substitutions to operators only (ablation) | CodeAttack · C#-Java translation | Attack success rate | Hurt |
Closest papers by meaning
- Black-Box Adversarial Attacks on LLM-Based Code Completion (ICML 2025 · similarity 0.87)
- Understanding and Enhancing the Transferability of Jailbreaking Attacks (ICLR 2025 · similarity 0.87)
- An LLM can Fool Itself: A Prompt-Based Adversarial Attack (ICLR 2024 · similarity 0.87)
- Chain of Attack: On the Robustness of Vision-Language Models Against Transfer-Based Adversarial Attacks (CVPR 2025 · similarity 0.86)
- MathAttack: Attacking Large Language Models towards Math Solving Ability (AAAI 2024 · similarity 0.86)
- Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation (ICLR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.