How phishing-detection-system/detection-chrome-extension works
Main Components
a. Data Collection & Storage:
• Main_dataset.csv: the raw dataset acquired from Kaggle.
• Phishing Websites Features.docx: documentation on domain-specific features.
b. Feature Engineering Module:
• featureExtraction.py and feature_extraction.py: scripts that transform raw data into engineered features.
• phishing_feature_engg.csv: the resulting feature dataset.
c. Model Development & Evaluation Module:
• Detecting_phishing_websites.ipynb: the notebook that orchestrates model training, testing, and hyperparameter tuning.
• Visualization_Lux_Phishing_Sites_Detection.ipynb: notebook for data visualization using Lux.
• phishing_classifier.pkl: the saved final XGBoost model.
d. API/Integration & Extension Component:
• api.py: could be used for serving the model or interfacing with external systems.
• Extension folder (manifest.json, popup.html, popup.js, style.css): contains files for a browser extension that can integrate with the classifier to warn users about phishing websites.
e. Environment & Dependencies:
• requirements.txt: lists necessary Python packages.
• .gitignore, __pycache__: standard development files which are less important from an architectural view.
Out Relationships and Data Flow
• The output (phishing_feature_engg.csv) is then used by the notebook and/or API scripts for training several ML models.
• After evaluation and tuning (as indicated in the Detecting_phishing_websites.ipynb), the best model (XGBoost) is saved as phishing_classifier.pkl.
• In parallel, the visualization notebook uses the feature set to generate insights and correlation plots.
• The API layer (api.py) and Extension directory indicate a future (or partial) integration where a browser extension may communicate with the back-end model to analyze URLs in real time.
Architectural Patterns and Design Principles
• Clearly separate data pre-processing/feature engineering from the model training phase.
• Use a modular design: distinct components for data preparation, training, evaluation, and extension integration.
• That the extension is intended to serve as a client interface – potentially a separate microservice component (accessing the classification model via the API).
Relevant Technologies and Frameworks
• Python and Jupyter Notebooks are used for experimentations and model training.
• Libraries such as scikit-learn, numpy, pandas, xgboost and visualization tools (Lux, matplotlib, seaborn) play a role in processing and evaluating data.
• The pickle module is used for persisting the trained model.
• The Extension folder indicates web technologies (HTML, CSS, JavaScript) which may integrate with a browser for real-time detection.