ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time Correction

TL;DR

ICON is a novel defense framework that detects and mitigates indirect prompt injection attacks in LLM agents by analyzing latent space signatures, significantly improving security without sacrificing task performance.

Contribution

We introduce ICON, a new inference-time correction method that detects IPI attacks via latent space analysis and performs targeted mitigation, enhancing security while maintaining task continuity.

Findings

Achieves 0.4% attack success rate, comparable to commercial detectors.

Over 50% improvement in task utility over existing defenses.

Demonstrates robustness across different models and modalities.

Abstract

Large Language Model (LLM) agents are susceptible to Indirect Prompt Injection (IPI) attacks, where malicious instructions in retrieved content hijack the agent's execution. Existing defenses typically rely on strict filtering or refusal mechanisms, which suffer from a critical limitation: over-refusal, prematurely terminating valid agentic workflows. We propose ICON, a probing-to-mitigation framework that neutralizes attacks while preserving task continuity. Our key insight is that IPI attacks leave distinct over-focusing signatures in the latent space. We introduce a Latent Space Trace Prober to detect attacks based on high intensity scores. Subsequently, a Mitigating Rectifier performs surgical attention steering that selectively manipulate adversarial query key dependencies while amplifying task relevant elements to restore the LLM's functional trajectory. Extensive evaluations on multiple backbones show that ICON achieves a competitive 0.4% ASR, matching commercial grade detectors, while yielding a over 50% task utility gain. Furthermore, ICON demonstrates robust Out of Distribution(OOD) generalization and extends effectively to multi-modal agents, establishing a superior balance between security and efficiency.