Enhanced phishing detection using heuristic rules and extreme learning machine (ELM)

Through fake website URLs, attackers steal personal data by developing phishing schemes, which represent an emerging threat in cybersecurity. Traditional approaches, including blacklisting, provide inconsistent protection against phishing attacks because phishing methods continually change. This research develops an improved phishing website preclassification model that uses heuristic rules and the Extreme Learning Machine (ELM) to increase detection efficiency. The heuristic approach collects vital website characteristics, including URL structure, domain age, HTTPS usage, and web content integrity, to detect possible phishing activities. The ELM classifier processes these extracted features through its single-layer feedforward architecture, which operates at high speed to perform accurate and fast classifications. By integrating heuristic rules with ELM, phishing detection accuracy improves while both false-positive rates decrease, and real-time prevention requires minimal computational expenditure. This research demonstrates that the proposed approach surpasses traditional detection methods, making it both scalable and effective for improving online security.