Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in Education
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918208204701696 |
|---|---|
| author | Yi, Xin Li, Yue Shi, Dongsheng Wang, Linlin Wang, Xiaoling He, Liang |
| author_facet | Yi, Xin Li, Yue Shi, Dongsheng Wang, Linlin Wang, Xiaoling He, Liang |
| contents | Large Language Models (LLMs) are increasingly integrated into educational applications. However, they remain vulnerable to jailbreak and fine-tuning attacks, which can compromise safety alignment and lead to harmful outputs. Existing studies mainly focus on general safety evaluations, with limited attention to the unique safety requirements of educational scenarios. To address this gap, we construct EduHarm, a benchmark containing safe-unsafe instruction pairs across five representative educational scenarios, enabling systematic safety evaluation of educational LLMs. Furthermore, we propose a three-stage shield framework (TSSF) for educational LLMs that simultaneously mitigates both jailbreak and fine-tuning attacks. First, safety-aware attention realignment redirects attention toward critical unsafe tokens, thereby restoring the harmfulness feature that discriminates between unsafe and safe inputs. Second, layer-wise safety judgment identifies harmfulness features by aggregating safety cues across multiple layers to detect unsafe instructions. Finally, defense-driven dual routing separates safe and unsafe queries, ensuring normal processing for benign inputs and guarded responses for harmful ones. Extensive experiments across eight jailbreak attack strategies demonstrate that TSSF effectively strengthens safety while preventing over-refusal of benign queries. Evaluations on three fine-tuning attack datasets further show that it consistently achieves robust defense against harmful queries while maintaining preserving utility gains from benign fine-tuning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_14423 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in Education Yi, Xin Li, Yue Shi, Dongsheng Wang, Linlin Wang, Xiaoling He, Liang Computation and Language Large Language Models (LLMs) are increasingly integrated into educational applications. However, they remain vulnerable to jailbreak and fine-tuning attacks, which can compromise safety alignment and lead to harmful outputs. Existing studies mainly focus on general safety evaluations, with limited attention to the unique safety requirements of educational scenarios. To address this gap, we construct EduHarm, a benchmark containing safe-unsafe instruction pairs across five representative educational scenarios, enabling systematic safety evaluation of educational LLMs. Furthermore, we propose a three-stage shield framework (TSSF) for educational LLMs that simultaneously mitigates both jailbreak and fine-tuning attacks. First, safety-aware attention realignment redirects attention toward critical unsafe tokens, thereby restoring the harmfulness feature that discriminates between unsafe and safe inputs. Second, layer-wise safety judgment identifies harmfulness features by aggregating safety cues across multiple layers to detect unsafe instructions. Finally, defense-driven dual routing separates safe and unsafe queries, ensuring normal processing for benign inputs and guarded responses for harmful ones. Extensive experiments across eight jailbreak attack strategies demonstrate that TSSF effectively strengthens safety while preventing over-refusal of benign queries. Evaluations on three fine-tuning attack datasets further show that it consistently achieves robust defense against harmful queries while maintaining preserving utility gains from benign fine-tuning. |
| title | Unified Defense for Large Language Models against Jailbreak and Fine-Tuning Attacks in Education |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2511.14423 |