Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{188219,
        author = {Vaishnavi Chirawande and Aditi Gade and Shweta Ahire and Sharvari Jadhav},
        title = {ShieldPhish : A Multi-Layered Phishing Detection System},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {7},
        pages = {1437-1444},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=188219},
        abstract = {Phishing is a continuing and advancing cyber threat that targets users across several means of communication. This report presents the design and implementation of ShieldPhish, a complete, multi-layered system designed to detect phishing attempts in invented URLs, emails, and SMS communication. ShieldPhish employs various feature extraction methods, including lexical, semantic (with transformer-based embeddings), and contextual analysis, in conjunction with strong machine learning models: a stacked ensemble of XGBoost and Logistic Regression for URLs and SMS, and a custom Multi-Input Neural Network for emails. ShieldPhish integrates real-time threat intelligence feeds as an initial defense, and it employs SHAP and LIME to provide explainable predictions to help users trust and understand the system. We discuss approaches and considerations for data acquisition, pre-processing, the architecture of the detection models, threat intelligence integration, explainability approaches, and finally, the deployment of the system across several interfaces: a Python package with CLIs, a Streamlit web app, and a browser extension for Chrome with proactive interception of navigation. Challenges encountered during development, such as heterogeneous data, complexity of feature engineering, instability of models, and complexity of environments/packaging, are discussed alongside proposed solutions. Performance evaluation shows high performance across the data modalities to suggest ShieldPhish as a protective system against contemporary phishing threats.},
        keywords = {Phishing Detection, Cybersecurity, Machine Learning, Ensemble Learning, Neural Networks, Transformer Embeddings, Explainable AI (XAI), SHAP, LIME, Threat Intelligence, Feature Extraction, URL Analysis, Email Security, SMS Filtering, Browser Extension, Streamlit Web App},
        month = {December},
        }