MITRE ATLAS & AI threat modeling
Think like an attacker: map the steps of a real AI attack.
Time this week: 6–8 hours
By Sunday you can explain MITRE ATLAS tactics, put a real AI attack in order, and threat-model a simple AI app.
OWASP tells you WHAT can go wrong. MITRE ATLAS tells you HOW attackers do it, step by step, based on real case studies. Threat modeling is how you find problems before launch.
Start here: this week's labs
An AI attack as a chain of tactics (ATLAS style)
Key ideas
MITRE ATT&CK
A famous public map of hacker tactics and techniques for normal IT systems.
MITRE ATLAS
The same idea for AI systems: Adversarial Threat Landscape for Artificial-Intelligence Systems. It has tactics, techniques and real case studies.
Tactic vs technique
Tactic = the attacker's goal at that step (e.g. Initial Access). Technique = how they do it (e.g. prompt injection).
Threat modeling
Drawing the system, asking 'what can go wrong?' at each part, and deciding what to do about it.
Trust boundary
A line where data moves from less trusted to more trusted, like from the internet into your AI app. Attacks happen at boundaries.
STRIDE
A simple checklist: Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege.
Day-by-day plan
Tick each day when you finish it. Your ticks are saved in this browser.
Mini game: STRIDE for an AI chatbot
Practice outside this course (free)
The matrix and the case studies. Start with the case studies.MITRE ATT&CK
The original framework for normal IT attacks.OWASP Threat Modeling
Plain intro to threat modeling.
These links need the internet. Everything else on this page works offline.
Exercises & worksheet
Weekly quiz
Mark it done to fill your progress bar on the course home.