Security Engineering for Agentic AI Systems

Packt via Coursera

Go to class Write review

Details

Go to class

Provider

Coursera
Pricing

Paid Course
Languages

English
Certificate

Certificate Available
Effort

14 weeks, 1 hour a week
Sessions

Self-Paced
Level

Intermediate
Subtitles

English

Found in

Overview

Google, IBM & Meta Certificates – 40% Off

One Coursera Plus subscription covers most Professional Certificates on Coursera.

Unlock All Certificates

Dive into the world of agentic AI security and master the principles, tools, and strategies to secure autonomous AI systems. Learn how to defend against attacks, protect agent goals, and ensure the integrity of AI-driven environments. This course offers a comprehensive deep dive into securing these intelligent systems, from understanding their unique characteristics to managing autonomy risks. You'll start by exploring the new attack surfaces agentic systems introduce and why traditional cybersecurity methods fail to protect them. Through real-world examples and lab exercises, you’ll learn how to build a secure AI architecture, develop defensive engineering practices, and understand how human-agent interactions can create vulnerabilities. As you progress, you'll master the complexities of goal integrity, tool security, and the critical importance of securing agent communications. You'll also learn how to prevent misalignment, rogue agents, and manipulation within AI-driven systems. The course covers essential topics such as securing multi-agent systems, memory, and context integrity, with a focus on building a resilient and safe environment for agentic AI. In the final stages of the course, you will apply your knowledge through practical exercises that focus on integrating security across all layers of agentic AI systems. By the end, you’ll be prepared to tackle complex security issues, ensuring your agentic systems are robust, secure, and aligned with ethical and safety standards. This course is designed for cybersecurity professionals, AI engineers, and system architects who are responsible for securing agentic AI systems. It is ideal for those working in industries that implement autonomous AI agents, such as cloud platforms, cybersecurity, and AI development teams. A solid understanding of cybersecurity principles and basic AI concepts is recommended, though not mandatory. The course is structured to guide you through the essential concepts of agentic AI security, progressing from foundational principles to advanced techniques. Each module combines theoretical learning with hands-on labs to ensure practical application. You'll gain a deep understanding of the security risks unique to agentic AI and how to build resilient systems to defend against them. This course is based on Security Engineering for Agentic AI Systems, by Anand Rao Nednur. This course is licensed and distributed by Packt. All rights reserved. Packt is one of the world's most prolific publishers of cutting-edge technical content. For over two decades we've made it our mission to curate and publish the knowledge of only the very best technical experts. We focus on real-world courses that help our customers get the job done, with coverage that extends across a wide range of established and cutting-edge technical topics. If you're an individual or an organisation that embraces learning by doing, Packt is the perfect fit for you.

Syllabus

Introduction
- This module provides an overview of the importance of agentic AI security and offers guidance on how to effectively navigate the masterclass. Learners will gain foundational knowledge and practical tips for engaging with the course content.
Module 1 — The Agentic AI Security Universe
- This module explores the unique characteristics and security challenges of agentic AI systems, including their attack surfaces, risks from autonomy and delegation, and the limitations of traditional cybersecurity approaches. Learners will gain insights into secure AI architecture, risk management, and human-agent interactions through real-world case studies and hands-on labs.
Module 2 — Mastering Agent Goal Integrity
- This module explores the critical concepts of agent goal integrity in AI systems, focusing on the risks of goal drift, hijacking, and injection. It provides practical insights into securing agent intentions through defensive engineering, intent provenance, and monitoring techniques. Learners will gain skills to audit, detect, and prevent malicious or unintended behavior in autonomous AI agents.
Module 3 — Tool Security, Sandboxing & Execution Safety
- This module covers critical aspects of securing agentic AI systems, including tool interfaces, sandboxing, egress control, and safe orchestration. Learners will explore real-world security risks and gain practical skills in designing and monitoring secure toolchains. The module emphasizes both theoretical concepts and hands-on strategies for mitigating tool misuse and unauthorized delegation.
Module 4 — AI Identity, Access Controls & Credential Security
- This module explores the critical aspects of identity and access control for AI agents, including secure credential management, privilege delegation, and methods to prevent identity-based attacks. Learners will gain an understanding of zero-trust principles and how to implement secure identity models in dynamic environments. Practical labs and knowledge checks reinforce hands-on application of these concepts.
Module 5 — Agentic Supply Chain & Component Trust
- This module explores key security strategies for protecting agentic AI systems throughout their supply chain. Learners will gain an understanding of risks like dataset poisoning, typosquatting, and registry compromise, as well as practical methods for securing components, building AI Bills of Materials, and implementing zero-trust architectures.
Module 6 — Code Safety, RCE Defense & Execution Control
- This module covers critical security concepts in agentic systems, including code safety, remote code execution (RCE) defense, and execution control. Learners will explore techniques for detecting and mitigating security risks in AI-generated code, such as code hallucinations, prompt injection, and unsafe dependencies. The module also includes practical labs on securing code pipelines and execution environments.
Module 7 — Memory, Context & Knowledge Base Security
- This module explores critical aspects of memory and context security in agentic systems, including memory types, poisoning risks, embedding attacks, and secure design practices. Learners will gain an understanding of how to protect agent memory and ensure data integrity in AI systems.
Module 8 — Multi-Agent Comms, Protocols & Coordination Security
- This module covers the fundamentals of secure communication in multi-agent systems, focusing on common risks, attack vectors, and defensive strategies. Learners will gain a deep understanding of how to design and implement secure coordination mechanisms. The content includes practical examples of cryptographic techniques and secure protocol design.
Module 9 — Cascading Failures, Systemic Risk & Resilience
- This module explores the causes and consequences of cascading failures in agentic AI systems, focusing on systemic risks, error propagation, and strategies to build resilient architectures. Learners will examine real-world scenarios and mitigation techniques to enhance system reliability and fault tolerance.
Module 10 — Human-Agent Trust, Psychology & Manipulation Defense
- This module explores how human cognitive biases and psychological factors influence trust in AI systems, and teaches strategies to defend against manipulation through ethical design, secure interfaces, and human-centered interactions.
Module 11 — Rogue Agents, Misalignment & Behavioral Security
- This module explores the challenges of managing rogue agents and misaligned AI systems, focusing on detection strategies, security controls, and ethical safeguards. Learners will gain an understanding of how agents can behave unpredictably and the tools to monitor and mitigate such risks effectively.
Module 12 — Capstone: Build a Secure Agentic System
- This module covers the design, integration, and security hardening of agentic AI systems. Learners will gain hands-on skills in threat modeling, red-teaming, and secure architecture implementation. The content emphasizes practical approaches to building resilient and production-ready agent-based systems.
Course Conclusion
- This module offers a reflective summary of the course content, emphasizing key takeaways and encouraging learners to continue their journey in agentic AI security. It provides a final perspective on the skills and knowledge gained throughout the course.