Project Logbook Sem Project

Project Title: AI Based Document Fraud Detection Web System Using Deep Learning

Sr.No	Contents	Date
1	Project Group Formation	08/01/2026 – 15/01/2026
2	Project Topic Finalization	19/01/2026 – 29/01/2026
3	Identified and analyzed the functional and non-functional requirements of the proposed system	29/01/2026 – 31/01/2026
4	Designed the overall system architecture, workflow, and module interactions	07/02/2026 – 18/02/2026
5	Implementation Phase – I	19/02/2026 – 28/02/2026
6	Developed the web interface and integrated core system functionalities.	01/03/2026 – 09/03/2026
7	Implementation Phase – II	10/03/2026 – 27/03/2026
8	Implementation Phase – III	28/03/2026 – 01/04/2026
9	Conducted functional, performance, and validation testing of the complete system	05/04/2026 – 27/04/2026
10	Evaluated model performance and analyzed the obtained results using relevant metrics	28/04/2026 – 09/05/2026
11	Prepared the project documentation and summarized the conclusions and future scope	09/05/2026 – 11/05/2026

---

AI-Based Document Fraud Detection Web System Using Deep Learning

Overview

The AI-Based Document Fraud Detection Web System is a deep learning-powered web application designed to automatically detect fraudulent and tampered documents. The system analyzes uploaded document images and classifies them as Genuine or Fraudulent using a Convolutional Neural Network (CNN) model trained on document datasets.

The project aims to reduce manual verification efforts and improve the accuracy, speed, and reliability of document authentication in educational institutions, recruitment agencies, financial organizations, and government sectors.

---

Features

• Document Image Upload
• Deep Learning-Based Fraud Detection
• Real-Time Prediction
• Genuine/Fake Classification
• Confidence Score Generation
• Responsive Web Interface
• Automated Verification Process
• Easy Deployment and Scalability

---

Project Architecture

User Uploads Document
          │
          ▼
 Image Preprocessing
          │
          ▼
 Feature Extraction
          │
          ▼
 CNN Deep Learning Model
          │
          ▼
 Fraud Detection Prediction
          │
          ▼
 Display Result on Web Interface

---

Technology Stack

Frontend

• HTML5
• CSS3
• JavaScrip

Backend

• Python
• Flask

Deep Learning

• TensorFlow
• Keras
• OpenCV
• NumPy
• Pandas

Model Training

• CNN (Convolutional Neural Network)

---

Dataset

The model is trained using internship certificate and document image datasets containing:

• Genuine Documents
• Fraudulent Documents
• Tampered Certificates
• Modified Document Images

Dataset Structure:

dataset/
│
├── genuine/
│   ├── doc1.jpg
│   ├── doc2.jpg
│   └── ...
│
└── fake/
    ├── doc1.jpg
    ├── doc2.jpg
    └── ...

---

Deep Learning Workflow

1. Data Collection

Collect genuine and fraudulent document images.

2. Data Preprocessing

• Resize images
• Normalize pixel values
• Convert image format
• Remove noise

3. Model Training

Train a CNN model using TensorFlow/Keras.

4. Model Evaluation

Evaluate using:

• Accuracy
• Precision
• Recall
• F1 Score

5. Deployment

Deploy trained model using Flask Web Framework.

---

Project Structure

AI-Document-Fraud-Detection/
│
├── app.py
├── train.py
├── predict.py
├── model/
│   └── fraud_detection_model.h5
│
├── dataset/
│   ├── genuine/
│   └── fake/
│
├── static/
│   ├── css/
│   ├── js/
│   └── uploads/
│
├── templates/
│   ├── index.html
│   └── result.html
│
├── requirements.txt
├── README.md
└── notebook.ipynb

---

Installation

Clone Repository

git clone https://github.com/PLACEHOLDER/AI-Document-Fraud-Detection.git

cd AI-Document-Fraud-Detection

Install Dependencies

pip install -r requirements.txt

---

Model Training

Run:

python train.py

The trained model will be saved as:

fraud_detection_model.h5

---

Run Web Application

python app.py

Open browser:

http://127.0.0.1:5000

---

Prediction Process

1. Upload document image.
2. Image is preprocessed.
3. CNN model extracts features.
4. Model predicts fraud probability.
5. Result displayed on webpage.

Performance Metrics

Metric	Value
Accuracy	84.1%
Precision	-
Recall	-
F1 Score	0.8403

---

Applications

• Educational Certificate Verification
• Recruitment Document Validation
• Government Record Authentication
• Banking and Financial Institutions
• Insurance Claim Verification
• Corporate Background Verification

Team Members:

Pranav Chaudhari

Darshan Shinde

Harsh Mali

Pratik Nikwade

Authors

College: R C Patel Institute of Technology

Department: Computer Science and Engineering (Data Science)

Guide: Prof. Priyanka D. Lanjewar

Academic Year: 2025-2026

---

License

This project is developed for academic and research purposes only.