AI Security Fundamentals: Securing with AI vs. Securing AI

Part 1: AI Fundamentals

• AI, ML, and Deep Learning: How Machines Actually Learn
  • Understanding Artificial Intelligence, AI Capabilities and Functionalities, Machine Learning, and Deep Learning, including the three core learning paradigms: Supervised, Unsupervised, and Reinforcement Learning
  • Distinguishing Predictive AI from Generative AI, and understanding where each is applied in real-world systems
  • How AI Models are built and mapping the full AI system pipeline: data collection, preprocessing, model training, validation, deployment, monitoring, and retraining
• Generative AI, LLMs, RAG, and Agentic AI Explained
  • Breaking down how Large Language Models work: tokens, context windows, transformer architecture, system vs. user prompts and difference between foundational and fine-tuned models
  • Understanding Retrieval Augmented Generation (RAG) and Agentic AI: how models retrieve external knowledge, and how autonomous agents plan, reason, and use tools
  • Introduction to the Model Context Protocol (MCP)
• Learning Pathway
  • What Developers & Engineers need to learn to build and maintain AI/ML systems, covering architecture, training pipelines, and deployment practices
  • What Analysts and Cybersecurity End Users (Offensive, Defensive, and GRC professionals) need to learn to effectively and safely use AI in their day-to-day work

Part 2: Securing with AI: AI for Security Operations & Offensive Security

• Using AI for Security Operations
  • How AI helps SOC analysts process high volumes of log data faster and reduce alert fatigue
  • Applying AI to threat intelligence workflows: summarizing threat feeds and correlating Indicators of Compromise (IOCs) against frameworks like MITRE ATT&CK
  • AI use cases in vulnerability management and incident response
• AI-Driven Offensive Security
  • Leveraging AI to accelerate reconnaissance and scanning: recon script generation, and interpreting scanning and enumeration results
  • AI-assisted password attacks and phishing simulation: targeted wordlist generation and crafting realistic social engineering scenarios for authorized testing

Part 3: Securing AI — Attacking & Defending AI

• How AI Systems Get Attacked: Poisoning, Evasion, Prompt Injection, Jailbreaks
  • Understanding data and model poisoning, evasion attacks, and model and data theft
  • Prompt injection (direct and indirect) and jailbreaking techniques used against LLMs and agentic systems, OWASP ML/LLM Top 10 and MITRE ATLAS
• Defending AI: Guardrails, Data, Model & Application Security
  • Implementing input/output guardrails for LLMs, and securing AI data across its lifecycle through classification, leakage prevention, and de-identification
  • Strengthening model security through adversarial training and integrity validation, and extending application/API security practices to cover AI-specific risks
• Secure AI Operations
  • Monitoring AI systems in production for drift, bias, and abuse patterns such as high-volume querying or repeated jailbreak attempts
  • Building AI-specific incident response capability and maintaining supply chain transparency through model provenance and SBOM/AI-BOM practices