Hybrid-AI-Blockchain-Framework-for-Insider-Threat-Mitigation-in-Databases

A hybrid security framework that integrates Blockchain, AI, and Role-Based Access Control (RBAC) to detect and mitigate insider threats in salary databases. It ensures tamper-proof transactions, access control enforcement, and automated anomaly detection for sensitive salary operations.

---

📌 Table of Contents

• Overview
• Project Structure
• Setup Instructions
• Running the System
• API Endpoints
• Roles & Access Control
• ML Model Information
• Report
• Future Work
• Contributors

---

🧠 Overview

This system aims to prevent unauthorized access and malicious manipulation of salary data using a three-layered defense approach:

1. RBAC: Restricts salary updates to authorized roles.
2. Blockchain: Records every transaction immutably.
3. AI Model: Flags suspicious transactions using a trained neural network.

---

🗂️ Project Structure

📦 Hybrid-AI-Blockchain-Framework
├── app.py                           # Flask backend for ML prediction
├── server.js                        # Node.js backend handling RBAC + blockchain
├── blockchain.js                    # Blockchain transaction logic
├── role.js                          # RBAC implementation
├── insider-threat-detection-model.ipynb  # Model training notebook
├── insider_threat_model.h5          # Pre-trained Keras model
├── encoder.pkl                      # Role label encoder
├── scaler.pkl                       # Feature standard scaler
├── salary_exploitation_insider_threat.csv # Training dataset
├── save_encoder.py                  # Script to generate encoder/scaler
├── .env                             # Environment variables
├── Report_dbs.pdf                   # Technical report
├── README.md                        # This file
├── package.json / package-lock.json # Node.js dependencies

---

🔧 Setup Instructions

🧱 Clone the Repository

git clone https://github.com/Dona-jui/Hybrid-AI-Blockchain-Framework-for-Insider-Threat-Mitigation-in-Databases.git
cd Hybrid-AI-Blockchain-Framework-for-Insider-Threat-Mitigation-in-Databases

🐍 Set Up Python Environment

Install required Python packages:

pip install flask pandas scikit-learn tensorflow keras

🟢 Set Up Node.js Environment

Install required Node.js packages:

npm install

⚙️ Train ML Model (Optional)

You can retrain the model using the notebook:

jupyter notebook insider-thret-detection-model.ipynb

Or use the pre-trained model: insider_threat_model.h5 encoder.pkl scaler.pkl

🚀 Run the System

1️⃣ Start the Flask ML Backend

python app.py
Runs the prediction server on http://localhost:5001.

2️⃣ Start the Node.js Backend

node server.js
Runs the RBAC and Blockchain system on http://localhost:3000.

📡 API Endpoints

🔮 Flask AI Prediction POST /predict Payload Example:

{
  "user_id": 1,
  "amount": 3096,
  "role": "admin",
  "time_of_change": "Work Hours"
}

Response Example:

{
    "message": "No threat detected.",
    "probability": 2.8129074053140357e-05,
    "threat_detected": false
}

🔐 Node.js Backend POST /login : Authorization and JWT-based authentication

{
    "token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VybmFtZSI6IkRvbmEgUm95Iiwicm9sZSI6ImFkbWluIiwiaWF0IjoxNzQ1MjU4NzE3LCJleHAiOjE3NDUyNjIzMTd9.Tgz__2izpCSRcQ5LoqIPGNld5jN8PhlFTNcVyebxVzQ"
}

POST /transaaction : Fetches entire blockchain transaction log

{
  "user_id": 1,
  "amount": 3096,
  "role": "admin",
  "time_of_change": "Work Hours"
}

Response Example:

{
    "message": "Transaction added to blockchain and salary updated",
    "blockchain": [
        {
            "index": 0,
            "timestamp": 1745258270424,
            "data": {
                "message": "Genesis Block"
            },
            "previousHash": "0",
            "hash": "e6302ce2f816e9197549ef55075aa641bd92b68c7793edd289b2a6984343df53",
            "nonce": 0
        },
        {
            "index": 1,
            "timestamp": 1745258790313,
            "data": {
                "user_id": 1,
                "amount": 3096,
                "role": "admin",
                "time_of_change": "Work Hours"
            },
            "previousHash": "e6302ce2f816e9197549ef55075aa641bd92b68c7793edd289b2a6984343df53",
            "hash": "0066c5c90988671938528d4caa0af6e682c2bda752bb1307a757ff7d3e911dbc",
            "nonce": 21
        }
    ]
}

🔐 Roles & Access Control

RBAC rules defined in role.js

🧠 ML Model Info

• Training Accuracy: ~97%
• 
• Model serialized as insider_threat_model.h5 with supporting encoder.pkl and scaler.pkl.

📄 Report

• Full technical documentation and workflow included in: 📄 Report_dbs.pdf

🌱 Future Work

• Add Frontend UI (React + Tailwind)
• Deploy on Ethereum Testnet (e.g., Sepolia)
• Dockerize the project

👨‍💻 Contributors

• Dona Roy (Postgraduate Student in Computer Science at NIT Karnataka, Surathkal)

---