This repository contains the Exploratory Data Analysis (EDA) on a dataset regarding Social Engineering / Phishing Website Attacks. The analysis aims to uncover patterns and behaviors indicative of phishing attacks through detailed feature exploration.
Social engineering attacks, particularly phishing, rely heavily on deceptive URLs, abnormal request routing, and specific web traffic patterns to trick users. This project explores a curated dataset of website features (taking values -1 for Phishing, 0 for Suspicious, and 1 for Legitimate) to visualize key indicators that can be used by machine learning models to classify these attacks.
- Phising_Testing_Dataset.csv: The raw testing dataset containing various website features.
- eda.py: A Python script designed to automatically generate high-quality EDA visualizations.
- EDA_Reports/figures/: A directory containing the visual outputs of the EDA process.
The dataset is perfectly clean with zero missing values across all features, ensuring it is ready for machine learning algorithms without requiring imputation.
An overview showing the prevalence of Phishing (-1), Suspicious (0), and Legitimate (1) values across the different features.
Using SSLfinal_State as a historically strong proxy for target labels, we can visualize the anticipated balance of legitimate vs. phishing samples within the test set.
A correlation matrix uncovering the linear relationships between different web features.
A 2-dimensional projection using Principal Component Analysis to visualize natural clustering in the dataset.
Features with the highest variance, demonstrating which attributes hold the most informational weight for predictive modeling.
To reproduce these figures, simply run the Python script:
python eda.pyThis will automatically generate the EDA_Reports/figures folder and populate it with the charts.