Biometric Source Scanner Identification for Secure Authentication

This repository implements an end-to-end secure fingerprint-based authentication system using an R307 fingerprint scanner, Raspberry Pi, deep learning, and a custom cryptographic client-server protocol.

The project combines biometric verification, machine-learning-based fingerprint classification, and secure socket communication to demonstrate a robust multi-factor authentication workflow.

System Overview

The system operates in four major stages:

Fingerprint Acquisition

• Fingerprint images are captured from an R307 sensor connected to a Raspberry Pi via UART.

Model Training & Selection

• Extracted fingerprint images are used to train multiple classification models.
• The best-performing model is saved and used during authentication.

Secure Client-Server Communication

A socket-based protocol securely transmits fingerprint data from client to server using:

• Hash-based integrity checks
• Timestamp validation (replay attack prevention)
• Symmetric encryption

Biometric Authentication

The server verifies:

• Fingerprint similarity (feature-based matching)
• Biometric Scanner
• Cryptographic validation (Authentication using Hashed Username and Password)

Repository Structure

.
├── Image_Download.py
├── Binary_Biometric_model.ipynb
├── client.py
├── server.py
├── fingerprint_check.py
├── OnevAll-95%.h5
├── User.txt
├── details.txt
└── README.md

File Descriptions

Image_Download.py

Purpose:

• Captures and stores raw fingerprint images from the R307 fingerprint sensor connected to a Raspberry Pi 3 via UART.

Key Features:

• Uses pyfingerprint library

• Continuously listens for fingerprint input

• Saves images as .bmp files for dataset creation

• Supports organizing images by sensor or user

Binary_Biometric_model.ipynb

Purpose:

• End-to-end notebook for data preprocessing, model training, evaluation, and export.

Workflow:

• Loads fingerprint images

• Performs preprocessing and resizing

• Trains multiple binary classification models

• Evaluates performance metrics

• Saves the best model as a .h5 file

client.py

Purpose:

• Acts as the biometric client running on the Raspberry Pi.

Responsibilities:

• Captures fingerprint image from R307 sensor

• Encodes image using Base64

Generates:

• SHA-256 integrity hash

• Timestamps

• Encrypts payload using Fernet symmetric encryption

• Sends fingerprint securely to the server

• Displays authentication result via GUI popup

Security Highlights:

• Prevents replay attacks (timestamp checks)

• Encrypted socket communication

• Multi-step verification logic

server.py

Purpose:

• Acts as the authentication server responsible for biometric verification.

Responsibilities:

• Receives encrypted fingerprint payload

• Verifies hash integrity & timestamps

• Decodes fingerprint image

• Compares Fingerprint image with database for authentication

• Runs Fingerprint image through the trained model for Biometric Source Scanner Identification

Performs:

• Feature-based fingerprint matching

• Deep learning-based scanner classification

• Generates session keys

• Sends authentication decision back to client

Machine Learning:

• Loads trained .h5 model

• Predicts scanner validity

• Uses threshold-based binary classification

fingerprint_check.py

Purpose:

• Performs traditional fingerprint feature matching between two images.

Techniques Used:

• CLAHE contrast enhancement

• Skeletonization & thinning

• Harris corner detection

• ORB feature extraction

• Brute-Force Hamming matching

Outcome:

• Determines whether two fingerprints match based on a distance threshold

Authentication Flow

Fingerprint Capture (Client)
        ↓
Image Encryption + Hashing
        ↓
Secure Socket Transmission
        ↓
Server Verification
  ├─ Hash Validation
  ├─ Timestamp Check
  ├─ Feature Matching
  └─ DL Classification
        ↓
Authentication Decision
        ↓
Encrypted Response to Client

Tech Stack

• Hardware: Raspberry Pi 3, R307 Fingerprint Sensor

• Languages: Python

• Libraries:

  • OpenCV

  • TensorFlow / Keras

  • NumPy, PIL

  • PyFingerprint

  • Cryptography (Fernet)

  • Socket Programming

• Model: Binary CNN classifier (.h5)

• Security: SHA-256 hashing, symmetric encryption, timestamps

How to Run

Hardware Setup

• Connect R307 sensor to Raspberry Pi UART pins

• Enable UART on Raspberry Pi

Install Dependencies

pip install pyfingerprint opencv-python tensorflow cryptography pillow numpy

Capture Fingerprint Dataset

python Image_Download.py

Train the Model

Open and run:

Binary_Biometric_model.ipynb

Start the Server

python server.py

Run the Client

python client.py

Key Highlights

• Combines biometrics + cryptography + DL

• Protects against replay & tampering attacks

• Modular design (easy to extend)

• Suitable for academic research, patents, and real-world demos