A Machine Learning-based Approach for Detecting Smishing Attacks at End-user Level

Smishing attacks, which target people through fraudulent text messages, are becoming one of the most prevalent social engineering attacks around the world. Personal information such as passwords, account information, or credit card numbers is sought by these messages and falling victim to one of these attacks could result in identity theft, financial loss, and unauthorized access to personal information. This research explores the pressing need for strong preventative measures against smishing attacks by examining the creation of new preventative measures while assessing the accuracy and flexibility of the current protection systems. Additionally, the viability of adding smishing detection to social media messaging platforms is explored, and the ability to identify smishing attempts made in various languages is investigated. As part of the research process, datasets are gathered, and various machine learning models, including SNN, CNN, and RNN are trained and evaluated. The integration of the Translator API into a mobile application to enhance the system’s capabilities by enabling multilingual detection is also considered. The highest test accuracy was achieved by the CNN model, and its efficacy in smishing detection was demonstrated. The robustness of the RNN model and its ability to generalize more effectively to new data were observed. The user-friendly elements of the mobile application, such as the floating bubble indicator, improve the user experience and promote preventive measures. A significant contribution to the field of smishing prevention is made by this study through the offering of an approachable way to identify and stop smishing attacks, ensuring a secure online environment.