InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915218227986432 |
|---|---|
| author | Li, Hao Liu, Xiaogeng |
| author_facet | Li, Hao Liu, Xiaogeng |
| contents | Prompt injection attacks pose a critical threat to large language models (LLMs), enabling goal hijacking and data leakage. Prompt guard models, though effective in defense, suffer from over-defense -- falsely flagging benign inputs as malicious due to trigger word bias. To address this issue, we introduce NotInject, an evaluation dataset that systematically measures over-defense across various prompt guard models. NotInject contains 339 benign samples enriched with trigger words common in prompt injection attacks, enabling fine-grained evaluation. Our results show that state-of-the-art models suffer from over-defense issues, with accuracy dropping close to random guessing levels (60%). To mitigate this, we propose InjecGuard, a novel prompt guard model that incorporates a new training strategy, Mitigating Over-defense for Free (MOF), which significantly reduces the bias on trigger words. InjecGuard demonstrates state-of-the-art performance on diverse benchmarks including NotInject, surpassing the existing best model by 30.8%, offering a robust and open-source solution for detecting prompt injection attacks. The code and datasets are released at https://github.com/leolee99/InjecGuard. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_22770 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models Li, Hao Liu, Xiaogeng Computation and Language Artificial Intelligence Cryptography and Security Prompt injection attacks pose a critical threat to large language models (LLMs), enabling goal hijacking and data leakage. Prompt guard models, though effective in defense, suffer from over-defense -- falsely flagging benign inputs as malicious due to trigger word bias. To address this issue, we introduce NotInject, an evaluation dataset that systematically measures over-defense across various prompt guard models. NotInject contains 339 benign samples enriched with trigger words common in prompt injection attacks, enabling fine-grained evaluation. Our results show that state-of-the-art models suffer from over-defense issues, with accuracy dropping close to random guessing levels (60%). To mitigate this, we propose InjecGuard, a novel prompt guard model that incorporates a new training strategy, Mitigating Over-defense for Free (MOF), which significantly reduces the bias on trigger words. InjecGuard demonstrates state-of-the-art performance on diverse benchmarks including NotInject, surpassing the existing best model by 30.8%, offering a robust and open-source solution for detecting prompt injection attacks. The code and datasets are released at https://github.com/leolee99/InjecGuard. |
| title | InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models |
| topic | Computation and Language Artificial Intelligence Cryptography and Security |
| url | https://arxiv.org/abs/2410.22770 |