135 matches found
Manipulating-Web-Agents
Manipulation d'agents web autonomes basés sur LLM par attaques par injection indirecte de prompts Avec l'omniprésence des applications intégrant des LLM, l'étude des risques de sécurité potentiels est cruciale. L'une des vulnérabilités les plus importantes de ces systèmes est leur sensibilité aux...
ActGuard
ActGuard ActGuard est une défense de pré-exécution par audit d'actions contre l'injection indirecte de prompts dans les agents LLM utilisant des outils. Ce dépôt contient l'implémentation finale d'ActGuard et le runtime basé sur AgentDojo nécessaire pour l'évaluer. La version publiée contient...
IPI-exposure-signal
Signal d'exposition IPI Ceci est le dépôt de code de notre article : Your Agentic LLMs Secretly Encode Latent Signals of Indirect Prompt-Injection Exposure. Version ArXiv et lien vers l'article : https://arxiv.org/abs/2608.02657 Ce dépôt implémente le pipeline de sondage des signaux latents...
llm-security
Nouveau : Démonstration d'attaques par injection indirecte sur Bing Chat Compromettre les LLM à l'aide de l'injection indirecte de prompts « ... un modèle de langage est une machine bizarre Turing-complète exécutant des programmes écrits en langage naturel ; lorsque vous faites de la récupération...
ROPE
ROPE : Application de la politique d'origine routée Code source de notre article : ROPE : Application de la politique d'origine routée contre l'injection indirecte d'invites par Xinhang Ma, Chaowei Xiao, William Yeoh, Ning Zhang, Yevgeniy Vorobeychik Résumé L'injection indirecte d'invites IPI...
prompt-injection-email-samples
Exemples d'e-mails d'injection de prompt Un petit ensemble de tests ouvert d'e-mails permettant de vérifier si vos contrôles de sécurité des e-mails et vos assistants de boîte aux lettres IA gèrent l'injection de prompt indirecte : des instructions cachées dans un e-mail qui tentent de détourner ...
Can CaMeLs Talk? Securing Multi-Agent Systems against Indirect Prompt Injection Attacks
Indirect prompt injection attacks - malicious instructions embedded in content processed by large language models - remain a major obstacle to safely deploying tool-using agents. CaMeL Debenedetti et al., 2025 mitigates this threat for an individual agent by separating trusted control flow from...
Who Is Your Agent Serving? Provider-Side Indirect Prompt Injection in Proactive Agents
Proactive personal agents increasingly decide what to recommend, how to personalize advice, and what follow-up assistance to offer, creating a new user-decision attack surface for provider-side indirect prompt injection. We show that an external provider need not access private user context,...
APEX: Active Protection at Execution Boundaries for LLM Agents
Indirect prompt injection IPI hides adversarial instructions in content that large language model LLM agents read at runtime. As agents compose heterogeneous capability units, including Tools, MCP servers, and Skills, the carriers of injection multiply, and defenses built to recognize attack...
Securing Computer-Use Agents against Branch Steering Attacks
Modern Computer Use Agents CUAs directly interact with graphical user interfaces and execute third-party web tools, exposing them to indirect prompt injection across every rendered page and tool response. While the Dual-LLM pattern is the primary system-level architecture offering formal security...
CVE-2026-96561
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
EUVD-2026-90465
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
CVE-2026-96561 AI Engine <= 3.8.0 - Unauthenticated Stored Cross-Site Scripting via 'model_' Parameter → PHP Error-Log Injection → Advisor Indirect Prompt Injection
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
CVE-2026-96561
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin is vulnerable to Stored Cross-Site Scripting (XSS) in versions up to and including 3.8.0 . The vulnerability stems from a chain of missing input neutralization and output escaping starting at the /mwai-ui/v1/chats/submit REST en...
CVE-2026-96561 AI Engine <= 3.8.0 - Unauthenticated Stored Cross-Site Scripting via 'model_' Parameter → PHP Error-Log Injection → Advisor Indirect Prompt Injection
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
CVE-2026-96561: Improper Neutralization of Input During Web Page Generation
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
PT-2026-103667
The AI Engine – The Chatbot, AI Framework & MCP for WordPress plugin for WordPress is vulnerable to Stored Cross-Site Scripting in versions up to, and including, 3.8.0 This is due to a chain of missing input neutralization and output escaping across the /mwai-ui/v1/chats/submit REST endpoint, the...
From A2A Attacks to Envelope-Layer Defense: Red-Teaming Evaluation of LLM Agents and a Three-Layer Isomorphic Attack-Defense Model
Agent interaction protocols such as ACP and A2A have moved LLM-based agents toward multi-agent collaboration, introducing new security threats. A task sent by a remote peer over A2A is treated as a legitimate request, providing a natural channel for indirect prompt injection. Existing agent...
Pikit: A Composable Toolkit for Indirect Prompt Injection Research and Evaluation
Indirect prompt injection embeds malicious instructions within external content retrieved by LLM-based agents, altering target behavior without user authorization. We introduce pikit, a research toolkit designed to systematically evaluate these threats across three core dimensions: attacks 13...
