I'm releasing the full source code of SENTINEL — an AI security platform. Not a "lite version" or "community edition" — everything.
What is it?
SENTINEL is a security platform for LLMs, AI agents, and multimodal systems:
- Defense — 97 detection engines (<10ms latency)
- Strike — Red team platform (39,000+ attack payloads)
Think of it as a firewall + pentest suite, but for AI.
What's included?
| Component | Details |
|---|---|
| 97 Detection Engines | From regex to Topological Data Analysis |
| Strange Math™ | 17 engines using advanced mathematics |
| Hive Intelligence | Threat Hunter, Post-Quantum Crypto |
| Strike Red Team | 39K+ payloads, AI-powered recon |
| Gateway | Production-ready Go HTTP server |
The "Strange Math" Engines
While most AI security tools use pattern matching, we went a different way:
Topological Data Analysis (TDA)
from gudhi import RipsComplex
rips = RipsComplex(points=embedding, max_edge_length=2.0)
simplex_tree = rips.create_simplex_tree(max_dimension=2)
persistence = simplex_tree.persistence()
Idea: Attacks create topological anomalies in embedding space.
Sheaf Theory
Coherence verification across multi-turn conversations.
Hyperbolic Geometry
Poincaré ball embeddings for attack clustering.
Quick Start
bash
git clone https://github.com/DmitrL-dev/AISecurity.git
cd AISecurity/sentinel-community
pip install -e .
Benchmarks
- Recall: 85.1%
- Precision: 84.4%
- F1 Score: 84.7%
- Latency: <10ms
Why Open Source?
AI security needs transparency
Threats evolve too fast for one team
It's Christmas 🎄
Links
GitHub: https://github.com/DmitrL-dev/AISecurity
HuggingFace: https://huggingface.co/datasets/Chgdz/sentinel-jailbreak-detection
Happy to answer questions! 🚀
Top comments (0)