Agentic Security

Agentic Security: Attack and Defend AI Agents is a three-day, hands-on course for cybersecurity professionals who need to understand, attack, and defend the autonomous AI systems now operating inside enterprise environments. Every agentic system that perceives, reasons, plans, and acts is a new attack surface. This course teaches you to exploit it and protect it.

Agentic Security Benefits

• In this course you will:

  1. Understand
  • Trace the AI architecture stack — ML, DNNs, transformers, LLMs, GenAI models, agentic systems — and identify the attack surface at each layer
  • Master agentic AI design patterns: Cognitive Loop, Planner-Executor-Verifier, multi-agent orchestration, and tool/API integration via MCP
  • Map the threat landscape: OWASP ML Top 10, OWASP LLM Top 10, NIST Adversarial ML Taxonomy, and MITRE ATLAS
  2. Build
  • Construct anomaly detection and deep learning malware classification models on real cybersecurity datasets
  • Deploy RAG pipelines integrating AlienVault OTX threat intelligence with chunk provenance validation
  • Implement multi-agent SecOps workflows using LangChain, CrewAI, or AutoGen with Apache Kafka for agent communication
  3. Attack
  • Execute all five prompt injection variants: direct, indirect, chained, multi-language, and refusal suppression
  • Conduct training data poisoning, model extraction, token inference side-channel attacks, hallucination exploits, and payload splitting
  • Perform AI-assisted memory forensic analysis using Volatility 3 to detect process hollowing, DLL injection, and advanced persistence
  4. Defend & Govern
  • Build autonomous threat detection and response workflows with human-on-the-loop oversight checkpoints
  • Apply NIST AI RMF AI 600-1, OWASP LLM Governance Checklist, and Zero Trust principles to agentic AI deployments

  Prerequisites

  2+ years cybersecurity experience; basic Python; Docker familiarity; comfort with Linux command line; understanding of common attack vectors and defensive frameworks.

  Who Should Attend:

  • Security Operations & Defensive Roles
  • Security Architecture & Engineering
  • AI / ML & Emerging Tech Roles
  • DevOps, Platform & Automation Roles
  • Governance, Risk & Compliance (GRC)
  • Leadership & Strategy Roles
  • Red Team & Offensive Security 

Agentic Security AI Training Outline

Learning Objectives

Module 1: AI Architecture & Agentic Foundations

• Trace the development of AI from Turing's test to modern agentic systems
• Demystify ML, deep neural networks, transformers, and LLMs
• Master agentic AI design patterns: Cognitive Loop, Planner-Executor-Verifier, multi-agent orchestration
• Identify the AI Security Ecosystem attack surface across compute, data, model, and agent pipeline layers

Module 2: Generative AI for SecOps and Risk Management

• Deploy RAG pipelines integrating live threat intelligence with chunk provenance validation
• Build AI-powered security operations workflows including incident reporting chatbots
• Establish a strong foundation in AI security risk management (CIA Triad, CVE, GenAI-specific risks, DLP)
• Apply adaptive authentication and data protection patterns to AI system deployments

Module 3: Hacking AI Agents – Adversarial Techniques

• Identify OWASP ML Security Top Ten and OWASP LLM Top Ten risks

• Execute the full prompt injection taxonomy: direct, indirect, chained, multi-language, refusal suppression
• Master jailbreaking (DAN), prompt leaking, and agent hijacking via crafted inputs
• Apply MITRE ATLAS and NIST AML taxonomy; execute AI Red Teaming methodology
• Understand GenAI social engineering, deepfake attacks, and the AI offensive toolkit

Module 4: Exploiting the AI Attack Surface

• Conduct training data poisoning, model extraction, and membership inference attacks
• Execute token inference side-channel attacks, hallucination exploits, and payload splitting
• Perform AI-assisted memory forensics using Volatility 3 to detect advanced threats
• Map all attacks to the NIST AI 100-2 taxonomy and MITRE ATLAS matrix

Module 5: Defending with Agents-Autonomous SecOps

• Build autonomous multi-agent threat detection and response workflows with human-on-the-loop oversight
• Integrate AI-based IDS, SOAR playbooks, and threat intelligence into agentic SecOps pipelines
• Deploy multi-agent systems using LangChain/CrewAI with Kafka and Redis/Celery for agent infrastructure
• Augment SIEM and SOAR with GenAI: NLP threat queries, playbook generation, AI-assisted triage

Module 6: AI Governance & Zero Trust for Agents

• Apply NIST AI RMF AI 600-1, OWASP LLM Governance Checklist, and regulatory frameworks to AI agent deployments
• Implement Zero Trust patterns for generative AI and agentic systems
• Deploy a role-aligned AI security agent with signed audit logging as the Zero Trust exit criterion
• Understand quantum computing implications and advanced persistent AI threats for future readiness