WHAZAZA/AI-cyber-range
⚔️ Build, break, and secure Large Language Models with our automated OWASP Top 10 cyber range for hands-on AI security training and research.
What's novel
⚔️ Build, break, and secure Large Language Models with our automated OWASP Top 10 cyber range for hands-on AI security training and research.
Code Analysis
10 files read · 3 roundsA CTF-style educational platform with 10 Dockerized FastAPI microservices that simulate OWASP LLM Top 10 vulnerabilities using hardcoded keyword-matching 'toy LLM' functions to leak flags when specific trigger phrases are entered.
Strengths
The labctl.py controller is well-designed with good UX (Rich console, interactive menus, port checking, signal handling, and proper Docker cleanup). The concept of gamifying LLM security education with difficulty tiers and flag pools is a reasonable pedagogical approach.
Weaknesses
The core 'vulnerability simulation' is trivially simple keyword matching with zero real depth, and ~90% of code is duplicated across all 10 labs with no shared base. The README is catastrophically broken (every link points to the same zip file), there are zero tests, and the project provides no actual security value beyond reading the source code.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
13
Contributors
3
Files
38
Active weeks
3
Repository
Language
Python
Stars
2
Forks
0
License
—