AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

Causal Detection of Multi-Step LLM Agent Attacks

Viraaji Mothukuri, Reza M. Parizi

ICML 2026 · Multi-agent systems and games › LLM agents · approach: Agent safety and attack benchmarks

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a causal approach to detect multi-step prompt injection attacks on LLM agents, where malicious intent only emerges across a sequence of individually legitimate actions.

Problem

Multi-step prompt injection attacks on LLM agents: malicious intent appears only after the workflow completes while each individual action looks legitimate, so action-level or content-pattern defenses (input sanitization, output validation, instruction hierarchy) cannot capture the sequential struct

Approach

Frames detection around the causal structure of the agent's action sequence rather than individual actions or content patterns; the abstract is truncated so the specific mechanism is not stated.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add watermark features (ablation)CausalTrace · Semi-real validation setF1 (%): 99.1% -> 99.2%Improved
add taint features (ablation)CausalTrace · Semi-real validation setF1 (%): 99.0% -> 99.2%Improved
add taint and watermark features on top of causal graph topology (ablation)CausalTrace · Semi-real validation setF1 (%): 99.1% -> 99.2%No clear effect
use causal graph features instead of action sequence featuresdetection classifier · Held-out real attacksF1 (%): 81.2% -> 95.3%Improved
use causal graph features instead of action frequency featuresdetection classifier · Held-out real attacksF1 (%): 74.3% -> 95.3%Improved
use causal graph features instead of domain-only featuresdetection classifier · Held-out real attacksF1 (%): 65.1% -> 95.3%Improved
evaluate out-of-distribution vs in-distributionCausalTrace · CausalBench vs OOD real benchmarksF1 (%): 99.4% -> 92.6%Hurt
perturb betaI and betaT causal parameters (ablation)CausalTrace · OOD benchmarksF1 (%): 92.6% -> 92.6%No clear effect
corrupt 30% of auxiliary causal edges (ablation)CausalTrace · OOD benchmarksF1: 0.926 -> 0.924No clear effect
corrupt 60% of auxiliary causal edges (ablation)CausalTrace · OOD benchmarksF1: 0.926 -> 0.875Hurt
corrupt 100% of auxiliary causal edges (ablation)CausalTrace · OOD benchmarksF1: 0.926 -> 0.394Hurt
add task-conditioned counterfactual (task alignment score) (ablation)CausalTrace · User-authorized cross-domain testfalse positive rate (%): 85.7% -> zero false positivesImproved

Experimental setup

Baselines: input sanitization output validation instruction hierarchy

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy