Akbarjon Olimov

AI Injection — security

Module by module

5 modules · 15 lessons · ~5 weeks

1. The attack surface
How an LLM app is put together and where it can be approached.
1. Anatomy of an LLM app
2. The threat map
3. Real incidents
2. Prompt injection
Direct and indirect injection: how the attack works.
1. Direct injection
2. Indirect injection
3. Tool abuse
3. Data leakage
Where secrets leak from the context and how to see it.
1. Context leakage
2. The secrets risk
3. Exfiltration paths
4. Defence layers
One barrier is not enough — defence appears when layers work together.
1. Input rules
2. The permission layer
3. Sandbox and monitoring
5. Testing and audit
Breaking your own app yourself — the cheapest form of defence.
1. Red-team basics
2. Evaluation exercises
3. The audit checklist

Mentor

Course instructor

Instructor of the “AI Injection — security” course. Full profile coming soon.