Phishing Website Detection With Semantic Features Based on Machine Learning Classifiers

The phishing attack is one of the main cybersecurity threats in web and spear phishing. Phishing websites continue to be a problem. One contributions our study was working extracting URL & Domain Identity feature, Abnormal Features, HTML JavaScript Features as semantic features detect websites, which makes process classification using those features, more controllable effective. current used machine learning model algorithms comparisons were made. We have 16 models adopted with 10 that represent most effective for detection webpages extracted from two datasets. GradientBoostingClassifier RandomForestClassifier had best accuracy based on comparison results (i.e., about 97%). In contrast, GaussianNB stochastic gradient descent (SGD) classifier lowest results; 84% 81% respectively, other classifiers.