Machine Learning-Driven Phishing Website Classification Using Extreme Learning Machine

Phishing efforts are on the rise, increasing the need for automatic, real-time detection. Similarly, existing conventional phishing detection methods such as decision trees (DT) and support vector machines (SVM) are computationally expensive, need extensive feature engineering, and have significant training cycles. These limits their adaptability to evolving phishing schemes. To address these issues, an Extreme Learning Machine (ELM) is used, which is simple to classify, has a quick training speed, high accuracy, and scalability. ELM detects phishing websites quickly since it requires fewer processing resources and can be deployed in real time. Quick detection is made possible through the use of primitive URL attributes, host information, and realistic content-based features. However, tests show that the proposed system has an accuracy of 98.5%, precision of 97.8%, and recall of 99.1% better than the existing system in both accuracy and computational cost. Moreover, it shows strength and reliability in phishing detection with a vastly lower number of false positives and false negatives than current strategies.