Human life increasingly depends on smartphones, not only for social connections and entertainment but also for accessing essential and convenient services with just a few taps. However, attackers exploit these dependencies, particularly through smishing—a form of phishing that uses text messages to deceive users. Detecting smishing is challenging due to the minimal information shared by attackers and the limited availability of real datasets. In our proposed model, we utilized two different datasets: an email dataset, which provides rich feature sets, and an SMS dataset, which has fewer but highly relevant features, such as tiny URLs suited to short messaging systems. After cleaning and processing these datasets, we initially ran various machine learning models with default settings. In our optimized framework, we further refined the datasets to enhance feature richness, tuned hyperparameters, and then applied machine learning algorithms, resulting in significant improvements shown through accuracy metrics and confusion matrices. Our proposed model demonstrates superior performance, ensuring higher efficiency and a unified solution for multiple types of threat sources.