🛡️ AnomalyGuard AI

Behavioral Fraud Intelligence Platform

AI-powered behavioral anomaly detection system for identifying suspicious user activity using Machine Learning, FastAPI, Docker, and Streamlit.

---

🔥 Live Demo

🌐 Fraud Intelligence Dashboard

https://anomalyguard-console.streamlit.app/

⚡ FastAPI Swagger Docs

https://anomalyguard-ai.onrender.com/docs

---

📌 Project Overview

AnomalyGuard AI is a production-grade behavioral anomaly detection platform designed to detect suspicious user behavior patterns using machine learning.

The system analyzes:

• Typing speed
• Mouse movement behavior
• Click frequency
• Session duration
• Behavioral deviations

to generate:

• anomaly scores
• fraud risk levels
• risk intelligence insights

---

🏗️ System Architecture

                   ┌──────────────────────┐
                   │  Streamlit Dashboard │
                   │  Fraud Intelligence  │
                   └─────────┬────────────┘
                             │
                             ▼
                   ┌────────────────────┐
                   │ FastAPI Inference  │
                   │ Behavioral API     │
                   └─────────┬──────────┘
                             │
                             ▼
                   ┌────────────────────┐
                   │ ML Pipeline        │
                   │ Isolation Forest   │
                   └─────────┬──────────┘
                             │
                             ▼
                   ┌──────────────────────┐
                   │ Risk Scoring         │
                   │ Fraud Classification │
                   └──────────────────────┘

---

🧠 High-Level ML System Architecture

flowchart TD

    A[User Behavioral Data] --> B[Data Preprocessing]

    B --> C[Feature Engineering]

    C --> D[ML Pipeline]
    
    D --> E[Isolation Forest Model]

    E --> F[Anomaly Scoring Engine]

    F --> G[Risk Classification]

    G --> H[FastAPI Inference API]

    H --> I[Streamlit Fraud Dashboard]

    H --> J[Swagger API Docs]

    I --> K[Real-Time Monitoring]

    subgraph Deployment
        L[Docker Container]
        M[Render Cloud Platform]
    end

    H --> L
    I --> L

    L --> M

Loading

---

🚀 Tech Stack

Category	Technologies
Machine Learning	Scikit-learn, Isolation Forest
Backend API	FastAPI
Frontend Dashboard	Streamlit
Visualization	Plotly
Containerization	Docker
Deployment	Render
Language	Python
Data Processing	Pandas, NumPy

---

🧠 ML Pipeline

Behavioral Features

• Typing Speed
• Mouse Speed
• Click Rate
• Session Duration
• Typing Variance
• Click Pattern Difference

---

Anomaly Detection Model

IsolationForest(contamination=0.1)

---

Risk Scoring Logic

Score Range	Risk Level
Low	Legitimate behavior
Medium	Suspicious behavior
High	Potential fraud

---

📊 Dashboard Features

✅ Real-time fraud prediction
✅ Behavioral analytics visualization
✅ Risk meter gauge
✅ Threat intelligence panel
✅ Session anomaly monitoring
✅ Live API integration
✅ Interactive fraud simulator

---

⚡ API Example

POST /predict

Request

{
  "typing_speed": 220,
  "mouse_speed": 110,
  "click_rate": 6,
  "session_time": 350
}

---

Response

{
  "prediction": "Normal",
  "anomaly_score": 0.0489,
  "risk_level": "Low",
  "risk_score": 19
}

---

🐳 Docker Usage

Build Docker Image

docker build -t anomaly-app -f docker/Dockerfile .

---

Run Container

docker run -p 8000:8000 anomaly-app

---

🚀 Local Development

Clone Repository

git clone https://github.com/RajeshKumar3451/Anomaly-Detection-System.git
cd Anomaly-Detection-System

---

Install Dependencies

pip install -r requirements-api.txt

---

Run API

uvicorn api.main:app --reload

---

Run Dashboard

streamlit run dashboard/app.py

---

📂 Project Structure

Behavioral-Anomaly-Detection/
│
├── api/
│   └── main.py
│
├── dashboard/
│   └── app.py
│
├── docker/
│   └── Dockerfile
│
├── models/
│   └── pipeline.pkl
│
├── src/
│   ├── data/
│   ├── features/
│   └── models/
│
├── requirements-api.txt
├── requirements-dashboard.txt
└── README.md

---

📈 Future Improvements

• LSTM Autoencoder
• Real-time WebSocket streaming
• Kafka integration
• JWT authentication
• Database logging
• MLflow monitoring
• Drift detection
• Kubernetes deployment

---

🏆 Key Achievements

✅ Production ML deployment
✅ Real-time fraud intelligence
✅ Dockerized AI infrastructure
✅ Cloud-hosted inference API
✅ Interactive security dashboard
✅ End-to-end MLOps workflow

---

📬 Contact

Author

Rajesh Kumar

GitHub

https://github.com/RajeshKumar3451/Anomaly-Detection-System

---

⭐ If you found this project useful

Give it a star on GitHub ⭐