Hybrid Model For Multi-Feature SMS And URL Safety Classification Using Machine Learning

Abstract

SMS and phishing a serious cybersecurity thread in mobile user. This article recommends a hybrid machine learning model framework that combines both Support Vector Machines (SVM) for optimal hyperplane-based classification and Random Forest (RF) ensemble learning to improve Security across textual features that are extracted from TF-IDF and URL-based features generated from lexical and structural patterns. In addition, sentiment analysis provides important understanding into user wish and behaviour of text message. An integrating machine learning system for multi-task classification inclusion of SMS spam detection, malicious URL identification, and sentiment analysis is presented in this Hybrid model. A Hybrid model to increase accuracy and reliability communications that are believed to come from reliable sources, such as banking sectors, trusting delivery services, and government Authorities incorporates predictions from both classifiers associations Random Forest (RF) and Support Vector Machine (SVM). The SVMRF model outperforms compare the individual SVM obtained 82% accuracy, 80% F1-score and RF achieved 77.82% accuracy, 77% F1-score when tested on a standard messaging dataset, SVMRF model providing higher accuracy for real-time mobile security thread. The model suggested that real-world mobile communication services is established by classification accuracy, overall accuracy is obtained 86%, it indicates that reliable performs of multifeatured SMS and Malicious URL SVMRF classification model