Malicious SMS detection using ensemble learning and SMOTE to improve mobile cybersecurity

The widespread use of cell phones, along with their constant internet connection, makes them vulnerable to malicious SMS attacks, including smishing and spam. Smishing involves attempts to steal personal information, while spam focuses on unwanted advertisements. Both pose cybersecurity threats, often requiring effective filtering techniques. Researchers have devised multiple methods for detecting malicious SMS, yet a notable gap remains in creating algorithms to reduce false positives, where normal messages are wrongly classified as malicious. The method employs ensemble learning to automatically identify malicious or legitimate messages. It combines Support Vector Machine and Random Forest models, compared with individual machine learning approaches for smishing detection. Feature extraction methods like Term Frequency (TF) and Term Frequency–Inverse Document Frequency (TF–IDF) are employed to derive features from the data. The imbalanced issue of the dataset is addressed by applying the Synthetic Minority Oversampling Technique (SMOTE). The results showed that the ensemble model outperformed the individual models, with an accuracy score of 99.58% when trained using TF–IDF on the balanced dataset. The proposed approach offers proactive defense against malicious SMS attacks, enhancing cybersecurity in the mobile communications sector.