SQL Injection Detection – ML-Based Web Application

Overview

A Flask-based machine-learning web application that classifies SQL payloads as SQL Injection or Benign using a TF-IDF vectorizer and Logistic Regression model. The app provides a stylish Bootstrap UI, real-time probability scoring, and an adjustable threshold slider for fine-grained detection.

Technologies Used

• Python 3
• Flask
• scikit-learn
• pandas
• Bootstrap 5
• HTML + CSS

---

Directory Structure

SQL-Injection-ML-Detector/
├── app.py
├── SQL_Dataset.csv
├── Payloads.txt
├── requirements.txt
└── README.md

---

Requirements

• Python 3.9+
• pip installed
• Any IDE (VS Code recommended)
• Dataset file: SQL_Dataset.csv
• Columns required:
  • Query (payload)
  • Label (0 or 1)

---

Dependencies

• flask
• pandas
• scikit-learn

---

How It Works

Model Workflow

• Loads dataset
• Vectorizes text using TF-IDF (1–2 n-grams)
• Trains Logistic Regression classifier
• Predicts SQL Injection probability for user inputs

Application Flow

• User enters a payload in the UI
• App vectorizes input → ML model
• Model returns:
  • SQL Injection (if probability ≥ threshold)
  • Benign (otherwise)
• Threshold controlled by slider(0.50–0.99)

UI Features

• Dark theme
• Modern card layout
• Resizable text area
• Gradient badges for SQLi/Benign
• Smooth user experience

---

Setup Instructions

1. Clone the repository:

git clone https://github.com/rahul07890-dev/SQL-Injection-Detector-ML.git

2. Install the dependencies:

pip install -r requirements.txt

3. Ensure dataset is in the same directory as app.py
4. Start the Flask server:

python app.py

5. Open the application:

http://localhost:5000

---

Screenshots

• Home Page

• Prediction Result Page

  • Benign payload

    

  • Malicious Payload

    

---

Video Demonstration

• Video Demo

SQLi.mp4

---

Features

• Machine Learning–powered SQL Injection detection
• Real-time probability scoring
• Adjustable threshold slider
• Clean modern Bootstrap UI
• Fully offline and local
• Easy to customize and extend

---

Improvements

• Save trained model using joblib
• Add REST API endpoint (/predict)
• Add Docker support
• Deploy on Render / Railway
• Add logs for prediction activity
• Expand dataset with more SQL payloads

---

Contribution

• Pull requests are welcome.
• For major changes, open an issue first.

---

Acknowledgments

• scikit-learn documentation
• Bootstrap 5
• Open-source SQL injection datasets