Mastering AI & Cybersecurity: From Foundations to Pro

00

Orientation & Lab Setup

Pre-Course

Install and verify course software; connect to the course lab
Set up Google Colab, Hugging Face, and GitHub accounts

01

AI & LLM Foundations

AI landscape: ML (supervised / unsupervised), neural networks, deep learning, generative & agentic AI
Model infrastructure: CPU vs GPU, VRAM, parameters, quantisation; where to run (laptop, Colab, HF Spaces, Vast.ai) and cost traps
Hands-onRun your first model with Ollama
ML for security: decision trees & Random Forest; intrusion detection on CIC-IDS2017
Metrics: confusion matrix, precision/recall, false positives; small neural network for packet classification; when ML beats an LLM
Hands-onIDS notebook
LLM internals: tokens & tokenisation (BPE, WordPiece), embeddings, attention, next-token prediction
Sampling (temperature, top-k, top-p); context window and overflow strategies
Training lifecycle: dataset curation, pre-training, fine-tuning, RLHF
Current LLM/SLM families; choosing a model for your hardware
Hands-onTokeniser, embedding visualiser & sampling experiments in LM Studio

02

Running & Building with LLMs

Serving locally: Ollama in depth (CLI, Modelfile, system prompts, custom models, publishing), LM Studio, llama.cpp
Model formats: GGUF, GPTQ, AWQ — what to download for which hardware
Hands-onBuild the "AppSec Guru" custom model
LLM APIs: REST generate / chat, OpenAI-compatible APIs, vLLM, Python clients
Front-ends: Streamlit, Open WebUI; exposing a model safely
Hands-onStreamlit chat app on top of AppSec Guru
Prompt engineering: anatomy (persona, task, context, format, tone, constraints); zero / one / few-shot; chain-of-thought; tree-of-thought; system vs user prompts
Hands-onWrite, compare & refine prompts for security tasks with a local model

03

AI-Assisted Development & Code Security

AI coding assistants: Cursor, GitHub Copilot, Continue with local models, Claude Code
Building full applications with AI; secure prompts for code generation; shift-left; human-in-the-loop
Hands-onGenerate a banking app with AI
Auditing AI-generated code: hard-coded secrets, XSS, SQL injection, root containers, vulnerable images & dependencies
Trivy, Semgrep & Gitleaks with LLM triage; local vs cloud models for code review; agent Skills (Skills.md) for repeatable scans
Hands-onBuild & Break — find and fix the flaws in the generated app
AI in CI/CD: a GitHub Actions pipeline with AI-assisted security gates
Reducing false positives; reporting; enterprise options
Hands-onSkills-based scanner and a CI pipeline that fails on findings

04

AI for Security

Threat modelling: STRIDE with AI; architecture diagrams as code (Mermaid); generating security requirements
Threat-modelling AI systems with MITRE ATLAS; where AI fits in the secure SDLC
Hands-onThreat model a web application and an LLM application
DAST with AI: test plans mapped to OWASP / CWE; agentic DAST (Strix, Claude Code); custom scripts & skills; Burp + LLM; rules of engagement
Hands-onAI-driven DAST against a vulnerable banking application
Mobile application analysis with JADX / MobSF and an LLM
Hands-onAnalyse an APK with a local model

05

AI Risks & Governance

Deepfakes: how they are made, detected, and used in fraud (ethics first)
Shadow AI: risks, detection, and controls
Hands-onDetect manipulated media; map Shadow AI in a sample organisation
Model provenance; malicious model weights (pickle); safetensors; model scanning; AI-SBOM
Regulation & frameworks: EU AI Act, NIST AI RMF, ISO/IEC 42001
Hands-onScan and verify a downloaded model; build an AI-BOM

06

Attacking & Defending LLM Applications

OWASP Top 10 for LLM 2025

Prompt injection, jailbreaks & system-prompt leakage (LLM01, LLM07): direct injection, DAN, many-shot, adversarial suffixes
Injection vs jailbreak vs leakage — same payload, different target
Defences: input validation, intent classification, prompt hardening
Hands-onGuessing-game levels and the SecureBank chatbot
Indirect injection & improper output handling (LLM01, LLM05, LLM06): hidden payloads in documents, URLs & retrieved content
XSS, SQL, command & Markdown injection through model output; excessive agency
Hands-onPoisoned résumé attack on an HR portal; output-handling labs
Data leakage & guardrails (LLM02, LLM09): Microsoft Presidio, NeMo Guardrails rails, LlamaGuard
Bias, fairness, misinformation, overreliance & hallucination
Hands-onAdd guardrails to the SecureBank chatbot
Model-level risks (LLM03, LLM04, LLM08, LLM10): unbounded consumption (Model DoS), rate limiting, model theft & extraction
Data and model poisoning; vector & embedding weaknesses
OWASP Top 10 for LLM 2025 — complete walkthrough and mapping exercise
Hands-onRate-limit gateway; extraction and poisoning demonstrations

07

LLM Red-Teaming Tooling

NVIDIA Garak: probes, detectors, reports & risk levels; Prompt Security Fuzzer
Testing a local model and reading the results
Hands-onScan a local model with both tools
Promptfoo red-team: plugins, strategies & reports
Testing black-box AI apps: capturing the API with Burp; writing REST configurations; reporting; red-team checks in CI
Hands-onRed-team SecureBank and write a short report

08

Agentic AI & MCP Security

Agent architecture: model, orchestrator, tools, memory; frameworks (n8n, CrewAI, LangGraph)
OWASP Top 10 for Agentic Applications: goal hijacking, tool misuse, excessive agency, insecure plugin / tool design
Hands-onBuild and connect a banking agent
Attacking agents: memory & context poisoning; insecure inter-agent communication; rogue agents and kill-switch design
Hands-onPentest the banking agent; add a kill switch and tool restrictions
MCP: architecture & design patterns; building custom MCP tools; MCP in AppSec workflows
Tool poisoning, context injection, secret exposure, scope creep, weak authentication / authorisation, shadow MCP servers
Data masking & DLP; AI Security Posture Management (AI-SPM); securing agentic workflows end-to-end
Hands-onBuild an MCP server, attack it, then harden it

★

Capstone Project

Final Session

Teams threat-model, red-team, and harden an AI application end-to-end
Present findings and fixes to the cohort

Mastering AI &
Cybersecurity

Bridge the Gap Between AI & Cybersecurity

Prerequisites

Built for Security Professionals

Everything You Need to Get Started

Start Date

Schedule

Duration

Mode

Timing

Certification

8 Modules of Intensive Training

Orientation & Lab Setup

AI & LLM Foundations

Running & Building with LLMs

AI-Assisted Development & Code Security

AI for Security

AI Risks & Governance

Attacking & Defending LLM Applications

LLM Red-Teaming Tooling

Agentic AI & MCP Security

Capstone Project

Skills That Set You Apart

Hands-on Practical Exercises

MCP Tool Development Proficiency

Real-World Red Team Techniques

AI-Augmented DAST / SAST Skills

Build & Serve Local LLMs (Ollama, vLLM, LM Studio)

Industry-Recognised Completion Certificate

OWASP LLM & Agentic AI Expertise

50 Hours of Live Instructor-Led Training

Ready to Master AI Security?