SecureSwipe

UI:Streamlit
Platform:Google Colab\

SecureSwipe Fraud Detection Project

Overview

This project aims to detect fraudulent transactions in the SecureSwipe dataset. The dataset contains both fraudulent and successful transactions, and the goal is to build a model that can accurately classify transactions as fraudulent or successful.

Requirements

• TensorFlow
• Keras
• NumPy
• Pandas
• Matplotlib
• Seaborn
• scikit-learn
• scipy
• imbalanced-learn
• tabulate
• plotly

Project Structure

• data/: Contains the dataset used for training and testing.
• src/: Contains the source code for data preprocessing, model training, and evaluation.
• models/: Contains saved trained models.
• results/: Contains evaluation results and visualizations.

Steps

1. Data Preprocessing

• Removed null, missing, and duplicate values.
• Scaled 'amt' and 'time' columns using RobustScaler to handle outliers.
• Detected and handled outliers using the Interquartile Range (IQR) method.
• Oversampled the data using Synthetic Minority Over-sampling Technique (SMOTE) due to class imbalance.

2. Model Training

• Utilized Logistic Regression as the classification model.
• Checked for multicollinearity using Variance Inflation Factor (VIF).
• Addressed multicollinearity by applying Ridge regularization.

3. Evaluation Metrics

• Evaluated the model using the following metrics:
  • Accuracy
  • Precision
  • Recall
  • Confusion Matrix

4. Privacy-Preserving Techniques

• Implemented Privacy-Preserving Aggregated Training with Exponential Mechanism (PATE) using Laplacian noise for security.

Usage

1. Clone the repository:

git clone https://github.com/your-username/SecureSwipe-Fraud-Detection.git

Navigate to the project directory:

cd SecureSwipe-Fraud-Detection

Install dependencies:

pip install -r requirements.txt