Multi-Agent Systems Security: Design Patterns for Building Auditable, Compliant, and Attack-Resistant AI Agents in Production - Couverture souple

Corwin, Cade

Couverture souple

ISBN 13 : 9798175020138

Editeur : Independently published, 2026

Synopsis

Your AI agents don't just answer anymore. They plan, call tools, hold credentials, and act — across your infrastructure, your data, and increasingly, across each other. When one gets it wrong, the damage isn't a bad sentence. It's an executed action: a deleted database, a leaked credential, a compromised production system.

Most security teams are defending this new reality with tools built for the old one.

Multi-Agent Systems Security is a complete, incident-anchored guide to closing that gap — built entirely around real, disclosed, dated breaches, not hypothetical scenarios. Every one of the ten risk categories in this book is grounded in an actual event: the zero-click Microsoft Copilot exploit that redefined prompt injection, the coding assistant weaponized through its own legitimate tools, the AI agent that deleted a production database during an explicit freeze, and the state-sponsored campaign that ran 80% of a real espionage operation autonomously before being caught.

Inside, you'll find:

• The complete OWASP Top 10 for Agentic Applications — each category with a real incident, a working attack walkthrough, concrete design patterns, and production-ready code
• Full architecture patterns for identity management, sandboxing, dependency tracking (AIBOM), and observability built specifically for autonomous agents — not adapted from human-user security
• Direct compliance mapping to the EU AI Act (including the 2026 regulatory delay most current guidance hasn't caught up to), ISO/IEC 42001, NIST AI RMF, SOC 2, and DORA — turn one set of controls into evidence for every framework your auditors actually ask about
• A complete red-teaming methodology, incident response playbook, and organizational maturity model for operating agentic systems safely at scale
• Real, working Python implementations throughout — policy enforcement layers, credential brokers, kill switches, and correlation logic you can adapt directly

This isn't a book about building multi-agent systems — plenty of good ones exist for that. This is the book for what happens after you've built one: how it gets attacked, how it fails, and exactly what stands between a wrong decision and a real incident.

If you're responsible for an AI agent that can take real action in the real world, this book tells you — with receipts — what's actually at stake, and exactly what to build first.

Scroll up and get your copy today.

This book is provided for informational and educational purposes. It does not constitute legal advice; readers should independently verify current regulatory deadlines before relying on it for compliance decisions.

Détails de l'édition

• Éditeur: Independently published
• Date d'édition: 2026
• Langue: anglais
• ISBN 13: 9798175020138
• Reliure: Taschenbuch
• Nombre de pages: 198
• Coordonnées du fabricant: Printed by Amazon on behalf of the author; https://www.amazon.de/hz/contact-us; ; c/o Amazon Media EU S.à.r.l., 38 Avenue John F. Kennedy; Luxembourg; L-1855; Luxembourg