SENSE: Shadow Exposure & eNterprise Surveillance for AI

SENSE (Shadow Exposure & eNterprise Surveillance for AI) is an advanced AI Security Platform designed to monitor, control, and secure AI adoption within the enterprise. It features a transparent AI Gateway that intercepts and inspects LLM traffic, enforcing granular security policies for Data Loss Prevention (DLP) and Prompt Injection protection.

Note: This project was presented at BlackHat MEA 2025.

---

🚀 Key Features

🛡️ Next-Gen AI Gateway

• Transparent Proxy: Routes traffic to OpenAI, Anthropic, or Local LLMs (Ollama) seamlessly.
• Policy Enforcement: Blocks malicious requests (e.g., Prompt Injection) and prevents sensitive data leaks (DLP) in real-time.
• Dynamic Configuration: Manage routes and backends directly from the UI without restarts.

📊 Advanced Visualization & Dashboard

• Live World Map: 3D visualization of global AI traffic and threat origins.
• Executive Dashboard: High-level metrics on AI adoption, risk posture, and policy violations.
• Thirol Timeline: Interactive timeline of blocked threats and anomalies.
• Network Intelligence: Visual graph of internal services communicating with external AI APIs.

🧠 Intelligent Policy Management

• Visual Policy Editor: Create, edit, and toggle security rules via the Dashboard.
• Granular Control: Define rules based on JSON keys (messages, prompt), Regex patterns, or Keywords.
• AI Insights: Automated analysis of blocked requests to identify patterns and emerging threats.

🔍 Deep Visibility & Detection

• Shadow AI Detection: Passive sniffing (libpcap) to detect unauthorized "Shadow AI" API calls.
• Real-time Activity Logs: Detailed audit trail of every intercepted request and violation.
• Process Monitoring: Identify specific processes (e.g., ollama, python scripts) generating AI traffic.

---

🛠 Prerequisites

• Go: Version 1.23+
• Node.js: Version 18+ (for Frontend)
• libpcap: Required for network capture (brew install libpcap on macOS).

---

📦 Installation & Setup

1. Clone Repository

git clone https://github.com/Faux16/sense-ai.git
cd sense-ai

2. Install Dependencies

# Backend
go mod tidy

# Frontend
cd internal/ui/frontend
npm install
cd ../../..

3. Build Backend

go build -o sense cmd/sense/main.go

---

💻 Usage

Quick Start (Recommended)

We provide a helper script to start both the Gateway and Monitoring Engine with the necessary permissions:

./scripts/start.sh

Note: This script uses sudo to enable network packet capture.

Access the Dashboard

Open your browser to: http://localhost:5173/ui/

Using the Gateway

Direct your AI Client (e.g., LangChain, Python openai lib) to the Gateway:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8081/v1", # Point to SENSE Gateway
    api_key="unused" 
)

response = client.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[{"role": "user", "content": "Hello!"}]
)

---

📂 Project Structure

cmd/sense/      # Main entry point (Engine + API)
internal/
  ├── gateway/  # AI Gateway Proxy implementation
  ├── policy/   # Policy Engine (DLP/Injection checks)
  ├── api/      # REST API configuration endpoints
  └── ui/       # React Frontend (Dashboard & Editors)
scripts/        # Startup and testing scripts
policies.yaml   # Default security policies
gateway.yaml    # Routing configuration

---

📄 License

MIT License.