RTBAS: Defending LLM Agents Against Prompt Injection and Privacy Leakage
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913690332168192 |
|---|---|
| author | Zhong, Peter Yong Chen, Siyuan Wang, Ruiqi McCall, McKenna Titzer, Ben L. Miller, Heather Gibbons, Phillip B. |
| author_facet | Zhong, Peter Yong Chen, Siyuan Wang, Ruiqi McCall, McKenna Titzer, Ben L. Miller, Heather Gibbons, Phillip B. |
| contents | Tool-Based Agent Systems (TBAS) allow Language Models (LMs) to use external tools for tasks beyond their standalone capabilities, such as searching websites, booking flights, or making financial transactions. However, these tools greatly increase the risks of prompt injection attacks, where malicious content hijacks the LM agent to leak confidential data or trigger harmful actions. Existing defenses (OpenAI GPTs) require user confirmation before every tool call, placing onerous burdens on users. We introduce Robust TBAS (RTBAS), which automatically detects and executes tool calls that preserve integrity and confidentiality, requiring user confirmation only when these safeguards cannot be ensured. RTBAS adapts Information Flow Control to the unique challenges presented by TBAS. We present two novel dependency screeners, using LM-as-a-judge and attention-based saliency, to overcome these challenges. Experimental results on the AgentDojo Prompt Injection benchmark show RTBAS prevents all targeted attacks with only a 2% loss of task utility when under attack, and further tests confirm its ability to obtain near-oracle performance on detecting both subtle and direct privacy leaks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_08966 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | RTBAS: Defending LLM Agents Against Prompt Injection and Privacy Leakage Zhong, Peter Yong Chen, Siyuan Wang, Ruiqi McCall, McKenna Titzer, Ben L. Miller, Heather Gibbons, Phillip B. Cryptography and Security Artificial Intelligence Tool-Based Agent Systems (TBAS) allow Language Models (LMs) to use external tools for tasks beyond their standalone capabilities, such as searching websites, booking flights, or making financial transactions. However, these tools greatly increase the risks of prompt injection attacks, where malicious content hijacks the LM agent to leak confidential data or trigger harmful actions. Existing defenses (OpenAI GPTs) require user confirmation before every tool call, placing onerous burdens on users. We introduce Robust TBAS (RTBAS), which automatically detects and executes tool calls that preserve integrity and confidentiality, requiring user confirmation only when these safeguards cannot be ensured. RTBAS adapts Information Flow Control to the unique challenges presented by TBAS. We present two novel dependency screeners, using LM-as-a-judge and attention-based saliency, to overcome these challenges. Experimental results on the AgentDojo Prompt Injection benchmark show RTBAS prevents all targeted attacks with only a 2% loss of task utility when under attack, and further tests confirm its ability to obtain near-oracle performance on detecting both subtle and direct privacy leaks. |
| title | RTBAS: Defending LLM Agents Against Prompt Injection and Privacy Leakage |
| topic | Cryptography and Security Artificial Intelligence |
| url | https://arxiv.org/abs/2502.08966 |