Title:Malicious URL Detection using optimized Hist Gradient Boosting Classifier based on grid search method

Abstract:Trusting the accuracy of data inputted on online platforms can be difficult due to the possibility of malicious websites gathering information for unlawful reasons. Analyzing each website individually becomes challenging with the presence of such malicious sites, making it hard to efficiently list all Uniform Resource Locators (URLs) on a blacklist. This ongoing challenge emphasizes the crucial need for strong security measures to safeguard against potential threats and unauthorized data collection. To detect the risk posed by malicious websites, it is proposed to utilize Machine Learning (ML)-based techniques. To this, we used several ML techniques such as Hist Gradient Boosting Classifier (HGBC), K-Nearest Neighbor (KNN), Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Multi-Layer Perceptron (MLP), Light Gradient Boosting Machine (LGBM), and Support Vector Machine (SVM) for detection of the benign and malicious website dataset. The dataset used contains 1781 records of malicious and benign website data with 13 features. First, we investigated missing value imputation on the dataset. Then, we normalized this data by scaling to a range of zero and one. Next, we utilized the Synthetic Minority Oversampling Technique (SMOTE) to balance the training data since the data set was unbalanced. After that, we applied ML algorithms to the balanced training set. Meanwhile, all algorithms were optimized based on grid search. Finally, the models were evaluated based on accuracy, precision, recall, F1 score, and the Area Under the Curve (AUC) metrics. The results demonstrated that the HGBC classifier has the best performance in terms of the mentioned metrics compared to the other classifiers.