Back to AI News
arxiv

PIPETTE Turns Red-Team Discoveries Into Harder-to-Inject LLM Agents

Researchers introduce PIPETTE, a purple-teaming framework for indirect prompt injection in LLM agents. The paper shows that two state-of-the-art RL-based red-teaming methods, RL-Hammer and PISmith, suffer from mode collapse, so high attack success rates during red-teaming do not translate into robustness gains from adversarial training. PIPETTE adds a dictionary learning-based reward to broaden attack strategy coverage and an on-policy adversarial training pipeline tailored to prompt injection defense. On AgentDojo and ASB, defenders trained with PIPETTE reach near-zero attack success rates on static attacks and withstand strong adaptive attacks while keeping benign task utility close to an undefended baseline.

AI Brief

It offers practitioners a concrete method for converting prompt injection red-teaming output into measurable agent robustness rather than redundant attacks.

arXiv:2611.24362v1 Announce Type: cross Abstract: Large language model (LLM) agents routinely interact with untrusted external data, exposing them to indirect prompt injection (IPI) attacks that hijack agents into executing attacker-chosen tasks. To uncover these vulnerabilities at scale, automated red-teaming methods often leverage reinforcement learning (RL) to generate attacks that achieve high attack success rates (ASR). Model developers use the attacks discovered during internal red-teaming to harden agents through adversarial training before deployment, a collaborative offensive-defensive process known in cybersecurity as purple-teaming. However, we show that achieving high ASR during red-teaming does not guarantee robustness gains from subsequent adversarial training. Specifically, we find that two state-of-the-art RL-based red-teaming methods, RL-Hammer and PISmith, suffer from mode collapse, repeatedly discovering similar attack strategies whose additional successes add no additional robustness improvements to adversarial training. To address this gap, we introduce PIPETTE, a unified prompt injection purple-teaming framework that effectively converts red-teaming discoveries into defender robustness improvements. PIPETTE employs a dictionary learning-based reward that encourages broader coverage of attack strategies during red-teaming. It also integrates an on-policy adversarial training pipeline tailored to IPI defense, enabling agents to resist attacks while completing legitimate user tasks. Extensive evaluations across agentic benchmarks (AgentDojo and ASB) demonstrate that PIPETTE produces defenders that achieve near-zero attack success rates on static attacks and can withstand strong adaptive attacks, while maintaining benign task utility close to that of the undefended baseline.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!