Phishing Website Detection

Classifying URLs as phishing or legitimate using the PhiUSIIL Phishing URL dataset.

Team

• Juno: data preparation, EDA, preprocessing
• Emmanuel: train/test split, baseline and final models, evaluation

Data source

PhiUSIIL Phishing URL Dataset, UCI Machine Learning Repository: https://archive.ics.uci.edu/dataset/967/phiusiil+phishing+url+dataset

Setup after cloning

Run all commands from the repo root. You need Python 3.10 or newer.

1. Clone the repo

git clone https://github.com/EmmanuelUrum/Phishing-Website-Detection.git
cd Phishing-Website-Detection

2. Create a virtual environment and install packages

Linux / macOS:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Windows (PowerShell):

python -m venv .venv
.venv\Scripts\Activate.ps1
pip install -r requirements.txt

Your prompt should now start with (.venv). Re-activate it every time you open a new terminal.

3. Turn on notebook output stripping (once per clone)

nbstripout --install

This keeps notebook outputs out of your commits. Without it, notebooks create huge diffs and merge conflicts.

4. Get the data

1. Download the PhiUSIIL Phishing URL dataset from the UCI ML Repository.
2. Put the CSV here: data/raw/PhiUSIIL_Phishing_URL_Dataset.csv

Raw data is not committed to git.

5. Build the clean dataset

python -m src.prepare_data

This writes data/processed/clean.parquet. If that file is already in the repo, you can skip steps 4 and 5.

6. Open the notebooks

jupyter notebook

Open the files in notebooks/ in order, 01 to 04.

Using the shared code (for models)

from src.split import load_split
from src.evaluate import evaluate

X_train, X_test, y_train, y_test = load_split()   # same split for everyone (random_state=42)
# ... train a model ...
metrics = evaluate(model, X_test, y_test, name="random_forest")

load_split() reads data/processed/clean.parquet relative to the current folder. Run Python from the repo root, or inside a notebook pass load_split("../data/processed/clean.parquet").

Add each model's metrics to reports/model_comparison.csv.

Troubleshooting

• ModuleNotFoundError: No module named 'src': run from the repo root and use python -m src.prepare_data, not python src/prepare_data.py.
• Missing .../PhiUSIIL_Phishing_URL_Dataset.csv: the CSV is not in data/raw/ or has a different name.
• Fewer than 235,795 rows loaded, or the label column shows 1.0/0.0: the CSV is incomplete. Download it again.
• nbstripout: command not found: activate the virtual environment first (step 2).

Project layout

• data/raw/ raw PhiUSIIL CSV
• data/processed/ cleaned dataset (clean.parquet)
• notebooks/ 01 EDA, 02 preprocessing, 03 baseline, 04 models
• src/ shared code: prepare_data.py, split.py, evaluate.py
• figures/ saved plots
• reports/ metrics table and written reports