LEXICAL PATTERN INTELLIGENCE: A MACHINE LEARNING SYSTEM FOR PREEMPTIVE DETECTION OF MALICIOUS URLS

The rapid increase in malicious Uniform Resource Locators (URLs) presents serious cybersecurity risks, enabling phishing attacks, malware dissemination, and financial fraud. Traditional detection methods, such as blacklisting and rule-based systems, have notable limitations. Blacklists rely on maintaining databases of known malicious URLs, but they are often unable to keep up with the continuous emergence of new threats. Rule-based approaches, which depend on predefined heuristics, struggle to detect novel or dynamically generated malicious URLs that do not conform to established patterns. Both methods are inherently reactive and ineffective against zero-day exploits. To address these challenges, this study explores the use of machine learning for accurate detection and classification of malicious URLs. The approach begins with the creation of a comprehensive dataset containing both benign and malicious URLs. Features are then extracted by analyzing lexical properties, hostbased characteristics, and network-level data. These features form the basis for training machine learning models that can learn to identify patterns indicative of malicious activity. By leveraging data-driven algorithms, the system offers a proactive defense capable of detecting previously unseen threats with high accuracy, adapting to the evolving landscape of cyberattacks