Agentic AI Security: Designing and Protecting Autonomous LLM Agents with Advanced Threat Models, Prompt Engineering, and Memory Safeguards - Tapa blanda

Sinopsis

In the rapidly evolving landscape of artificial intelligence, autonomous large language model (LLM) agents are redefining how systems reason, act, and interact with the world. These agents go beyond answering queries—they execute complex workflows, leverage external tools, and maintain persistent memory to achieve goals. However, with this transformative power comes unprecedented security challenges. Agentic AI Security: Designing and Protecting Autonomous LLM Agents with Advanced Threat Models, Prompt Engineering, and Memory Safeguards is your essential guide to building and securing these next-generation AI systems.
This comprehensive book provides AI engineers, security architects, DevSecOps professionals, and responsible AI practitioners with a robust framework to safeguard autonomous LLM agents. Across eight expertly crafted chapters, you’ll explore how to mitigate risks like prompt injection, memory poisoning, feedback loop attacks, and self-modifying agent behaviors. Learn to design secure agent architectures, implement layered defenses, and align with emerging compliance standards to ensure your systems are both powerful and trustworthy.
Inside, you’ll discover how to:

1. Develop agent-specific threat models using STRIDE and other frameworks tailored for autonomous systems.
2. Engineer schema-bound prompts and gated tool orchestration to prevent intent drift and unauthorized actions.
3. Implement memory integrity checks, anomaly detection, and write controls to secure agent recall and persistence.
4. Embed safety critics, intent modeling, and policy enforcement within the agent’s reasoning loop for real-time protection.
5. Conduct red teaming, adversarial testing, and continuous threat simulation to proactively harden agent deployments.
6. Navigate compliance with NIST AI RMF, OWASP GenAI Top 10, and the EU AI Act for enterprise-grade, auditable AI systems.

"Sinopsis" puede pertenecer a otra edición de este libro.