Overview
The objective of the project is to classify if an URL is phishing or not. This model repo contains the required encoders (for url,dom and tld), scaler (for digit_cnt and is_https) and the trained model (RandomForest Classifier).
Dataset & Attribution
This project uses the URL-Phish dataset. The dataset was obtained from Kaggle, where it is available as Phishing URL Detection (111K URLs, 22 Features).
The dataset is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0), which permits sharing, redistribution, and adaptation with appropriate credit.
Dataset citation
Dam Minh, Linh; Tran Cong, Hung (2025).
URL-Phish: A Feature-Engineered Dataset for Phishing Detection.
Mendeley Data, V1.
DOI: https://doi.org/10.17632/65z9twcx3r.1
Original data sources referenced by the dataset authors
PhishTank – Community-driven phishing URL repository
Research Organization Registry (ROR) dataset – Source of trusted benign domain URLs
Paper citation
Dam Minh Linh, Tran Cong Hung,
A feature-engineered dataset of benign and phishing URLs for machine learning and large language models evaluation,
Data in Brief,
Volume 63,
2025,
112162,
ISSN 2352-3409,
https://doi.org/10.1016/j.dib.2025.112162.
Modifications:
The following preprocessing was applied to the original dataset:
- Duplicate rows and null/missing values were checked for and removed, if present
- Feature scaling applied to selected numeric features
- TF-IDF encoding applied to selected URL/text-derived feature(s)
- Data split into train / validation / test sets
Feature usage:
The final model was trained using a selected subset
of the features; the remaining
features were excluded at training time via feature selection, not by removing them from
the stored datasets.
License
- Code: MIT License — see
LICENSE - Data: Raw and processed datasets are redistributed under Creative Commons Attribution 4.0 International (CC BY 4.0) license, consistent with the original dataset's license (see Dataset & Attribution above).
- Model & preprocessors: MIT License — trained artifacts are provided under the same terms as the codebase. Model trained on Creative Commons Attribution 4.0 International (CC BY 4.0) license data; see Dataset & Attribution section for details.