An Efficient Artificial Intelligence Framework for Phishing Threat Detection in Online Platforms

The increasing prevalence of malicious Uniform Resource Locators (URLs) and fraudulent websites poses a significant threat to online security, with search engines inadvertently becoming vectors for these harmful entities. Traditional phishing detection methods, such as blacklists, whitelists, and static rule-based heuristics, are demonstrably inadequate against the rapid evolution of modern phishing strategies, often failing to detect zero-day threats and yielding high false-positive rates. This research proposes an advanced, adaptive phishing detection framework that addresses these critical shortcomings by integrating Natural Language Processing (NLP) techniques with a robust ensemble of powerful machine learning algorithms: Support Vector Machine (SVM), Random Forest (RF), and Decision Tree (DT). NLP is leveraged for sophisticated feature extraction, analyzing lexical, structural, and domain-based characteristics of URLs to capture the behavioral patterns associated with malicious attacks. The ensemble model capitalizes on the specific strengths of each classifier: SVM for efficient high-dimensional feature handling, RF for enhanced accuracy via ensemble decision-making, and DT for interpretability and feature importance analysis. This unified architecture enables the accurate and reliable classification of URLs as legitimate or phishing in real- time. Furthermore, the system incorporates Advanced Encryption Standard (AES) to secure sensitive user data, such as browsing history and URL-related information, both in storage and during transmission, ensuring data confidentiality even upon interception. By combining intelligent, adaptive URL classification with robust, privacy-preserving encryption, this dual-focus framework provides a comprehensive and resilient cybersecurity solution, significantly enhancing user protection against complex modern phishing threats while setting a new standard for data confidentiality.