🧠 217 Detection Engines — Strange Math™ Protection
SENTINEL Brain is the detection core of the SENTINEL AI Security Platform. It analyzes every prompt and response in real-time using 217 specialized engines.
| Category | Engines | Detection |
|---|---|---|
| 🎯 Injection | 30+ | Prompt injection, jailbreak, Policy Puppetry |
| 🤖 Agentic | 25+ | RAG poisoning, tool hijacking, memory attacks |
| 📐 Mathematical | 15+ | TDA, Sheaf Coherence, Chaos Theory |
| 🔒 Privacy | 10+ | PII detection, data leakage prevention |
| 📦 Supply Chain | 5+ | Pickle security, serialization attacks |
pip install sentinel-llm-securityfrom sentinel import scan
result = scan("Ignore previous instructions and...")
print(result.is_safe) # False
print(result.risk_score) # 0.95
print(result.detections) # ['prompt_injection', 'policy_puppetry']What makes SENTINEL different from keyword matching?
| Standard Approach | SENTINEL Strange Math™ |
|---|---|
| Keyword matching | Topological Data Analysis |
| Regex patterns | Sheaf Coherence Theory |
| Simple ML classifiers | Hyperbolic Geometry |
| Static rules | Optimal Transport |
| — | Chaos Theory |
from sentinel.brain.engines import TDAAnalyzer
analyzer = TDAAnalyzer()
result = analyzer.analyze(prompt)
# Uses persistent homology to detect semantic anomalies| Engine Category | Precision | Recall | F1 | P50 | P99 |
|---|---|---|---|---|---|
| Injection (Tier 1) | 97% | 94% | 95.5% | 3ms | 12ms |
| Jailbreak (Tier 2) | 95% | 91% | 93% | 8ms | 25ms |
| RAG Poisoning | 92% | 89% | 90.5% | 15ms | 45ms |
| TDA Analyzer (Tier 3) | 89% | 96% | 92.4% | 45ms | 120ms |
| Combined Pipeline | 94% | 93% | 93.5% | 18ms | 85ms |
Tested on SENTINEL Strike payloads + internal validation set.
src/brain/
├── engines/ # 217 detection engines
│ ├── injection/ # Prompt injection detection
│ ├── jailbreak/ # Jailbreak prevention
│ ├── agentic/ # RAG/Agent/Tool protection
│ ├── mathematical/ # TDA, Sheaf, Chaos engines
│ └── privacy/ # PII, secrets detection
├── core/ # Engine orchestration
├── config/ # Engine configuration
├── api/ # REST/gRPC endpoints
└── integrations/ # OpenAI, Anthropic, LangChain
Built-in support for:
- LLM Providers: OpenAI, Anthropic, Google, Ollama, HuggingFace
- Frameworks: LangChain, LlamaIndex, Haystack
- Vector Stores: Pinecone, Weaviate, Qdrant, Chroma
- Orchestration: FastAPI, Flask, Django
- SENTINEL Shield — C DMZ proxy
- SENTINEL Strike — Red team platform
- SENTINEL Framework — Python SDK
SENTINEL Brain
217 Reasons You're Protected