Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2609.22510

Defusing Explosive Prompts: Understanding and Preventing Trigger-Based Prompt Injections in LLM Agents

Justin Szczepaniak·Elad Feldman·Naum Viner·Ben Nassi
View original source ↗Read source document ↗Export metadata ↓

Abstract

As LLM applications integrate with external tools, they are increasingly exposed to indirect prompt injection (IPI), where adversarial instructions are embedded in retrieved content. Conventional IPIs fire on contact: the moment an agent ingests the content, it carries out the instruction. We introduce the explosive prompt, a conditional payload that stays dormant until an attacker-chosen trigger is met, in effect a training-free, inference-time backdoor planted in a single piece of retrieved content. This temporal separation reaches where ordinary IPI cannot. On frontier models that refuse the bare imperative almost entirely, rephrasing the same goal as a dormant conditional drives real, state-changing tool execution against a live agent backend (a paired mean of 16.5% vs. 2.4% for the imperative, reaching 34.2% on a proprietary model). In trials on nine production agents (OpenAI Codex, Google Gemini CLI, Anthropic Claude Code CLI, Cursor CLI, GitHub Copilot, Devin AI CLI, Amazon Kiro CLI, Qwen Code, Google Assistant; n=30 each), explosive prompts succeed in 43-83% of cases versus at most 3% for an imperative baseline, and they slip past deployed defenses: off-the-shelf injection classifiers are miscalibrated on them, and a preference-optimized model that closes imperative injection entirely still executes 11.8% of explosive prompts, every one at the trigger turn. The durable defensive lever is ingestion-time detection of the conditional structure, once detectors are trained on explosive-prompt data, which no prior benchmark supplied and our generator does. Retraining cuts live tool-execution attack success from an undefended 34.3% to 7.5-8.1% for the encoder baselines. Our detector, DeFuse, reaches 3.0% at a calibrated 5% false-positive budget with the best detection quality of any method tested (AUC 0.9994) and 25x lower latency, though it needs length-aware thresholds.

Explore related subjects

cs.CRcs.LG

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Justin Szczepaniak, Elad Feldman, Naum Viner, Ben Nassi. 2026-09-18. Defusing Explosive Prompts: Understanding and Preventing Trigger-Based Prompt Injections in LLM Agents. https://arxiv.org/abs/2609.22510

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

UPPRESSO: Untraceable and Unlinkable Privacy-PREserving Single Sign-On Services

Single sign-on (SSO) allows a user to maintain only the credential for an identity provider (IdP) to log into multiple relying parties (RPs). However, SSO introduces privacy threats, as (a) a curious IdP could track a user's all visits to RPs, and (b) colluding RPs could learn a user's online profile by linking her identities across these RPs. This paper presents a privacypreserving SSO scheme, called UPPRESSO, to protect an honest user's online profile against (a) an honest-but-curious IdP and (b) malicious RPs colluding with other users. UPPRESSO proposes an identity-transformation approach to generate untraceable ephemeral pseudo-identities for an RP and a user from which the target RP derives a permanent account for the user, while the transformations also provide unlinkability. This approach protects the identities of the user and the target RPs in a login flow, while working compatibly with widely-deployed SSO protocols and providing services accessed from a commercial-off-the-shelf browser without plug-ins or extensions. We built a prototype of UPPRESSO on top of MITREid Connect, an open-source SSO system. The extensive evaluations show that it fulfills the security and privacy requirements of SSO with reasonable overheads.

cs.CR↗
arXiv2026

Attack Tree Distance: a practical examination of tree difference measurement within cyber security

Attack trees are a popular threat modeling method. In practice, there is often a need to compare attack tree models produced by human experts, based on both the structure of the tree and the meaning of the node labels. In this work, we investigate the problem of comparing attack trees and measuring their similarity. We define five different measures for measuring the distance between two attack trees: Label Distance (LD), Tree Edit Distance (TED), Radical Distance (RD), Multiset Distance (MSD) and Weighted Sum Distance (WSD). We further propose a repeatable method of both theoretical and experimental attack tree distance measures validation. Our theoretical validation consists of a series of basic transformations to evaluate the behavior of distance measures with respect to specific types of transformations that may appear between two attack trees. To experimentally validate our distance measures, we designed and executed a human study ($n=39$) to collect a dataset of attack trees to be used for evaluation and comparison of the measures. From our theoretical and experimental results, we find that applying semantic similarity as a means of comparing node labels is a valid approach. Further, we find four of the five attack tree distance measures are valid approaches in certain, varying circumstances. Our results suggest that these methods can already be used to identify similar real-world attack trees. Overall, this work lays the groundwork for improved threat model analysis, validation of AI-generated attack trees, and future research into threat similarity measurement in cybersecurity.

cs.CR↗
arXiv2026

Probabilistic Modeling of Jailbreak on Multimodal LLMs: From Quantification to Application

Recently, Multimodal Large Language Models (MLLMs) have demonstrated their superior ability in understanding multimodal content. However, they remain vulnerable to jailbreak attacks, which exploit weaknesses in their safety alignment to generate harmful responses. Previous studies categorize jailbreaks as successful or failed based on whether responses contain malicious content. However, given the stochastic nature of MLLM responses, this binary classification of an input's ability to jailbreak MLLMs is inappropriate. Derived from this viewpoint, we introduce jailbreak probability to quantify the jailbreak potential of an input, which represents the likelihood that MLLMs generated a malicious response when prompted with this input. We approximate this probability through multiple queries to MLLMs. After modeling the relationship between input hidden states and their corresponding jailbreak probability using Jailbreak Probability Prediction Network (JPPN), we use continuous jailbreak probability for optimization. Specifically, we propose Jailbreak-Probability-based Attack (JPA) that optimizes adversarial perturbations on input image to maximize jailbreak probability, and further enhance it as Multimodal JPA (MJPA) by including monotonic text rephrasing. To counteract attacks, we also propose Jailbreak-Probability-based Finetuning (JPF), which minimizes jailbreak probability through MLLM parameter updates. Extensive experiments show that (1) (M)JPA yields significant improvements when attacking a wide range of models under both white and black box settings. (2) JPF vastly reduces jailbreaks by at most over 60\%. Both of the above results demonstrate the significance of introducing jailbreak probability to make nuanced distinctions among input jailbreak abilities.

cs.CR↗