AI and Big Data-Enabled Deep Reinforcement Learning Model for Mobile Smishing Threat Detection

SMS phishing (smishing) is a form of SMS-based phishing scam that occurs when an attacker attempts to acquire a user's sensitive data, leading to monetary damages for individuals. At present, smartphones are very flexible and are frequently used by users due to the extensive array of features they offer. Still, determining the legitimacy of mobile smishing attacks poses an individual issue compared to the computer phishing, in response to the differing structures of each platform. In addition, accurate detection of mobile smishing attacks is of major concern, as less work has been carried out in this domain. Multiple anti-phishing techniques for mobile devices have been presented a long way; yet, however, a comprehensive solution remains challenging. To address this issue, our research presents a hybrid deep reinforcement learning model that employs CNNLSTM trained on an SMS spam collecting dataset, integrated with Q-learning through hyperparameter tuning. The hybrid DRL model attains an accuracy of 98.1% without class weighting, and a remarkable 99% with class weighting.