Recently, the phishing attack is one of the critical threats against Organizations, internet users, service providers, cloud computing, and many other fields in daily life. In the phishing attack, the intruder attempts to defraud the users and leak or steal the credential information, including personal information such as bank account, passwords, etc., by sending a fooled email or SMS to redirect the user to an untrusted website. Various methods have been proposed in terms of filtering and detect different types of phishing attacks; however, the researchers and security information experts are still studying to find a solution to assure the internet security from phishing and other attacks. Viewing SMS phishing messages are mostly short text and become a relatively low number associated with legitimate messages, new features for quick writing, and oversampling technique for imbalanced data utilized to SMS phishing detection. In this research, a novel framework of the SMS phishing detection presented. The proposed method combines feature extraction, oversampling, optimization algorithm for feature selection and classification. For the feature extraction and classification, the Support vector machine is implemented. In addition, the Adaptive Synthetic Sampling Approach method used to be an oversampling method. Then, the Binary Gray Wolf Optimizer Algorithm (BGWO) is applied to analyze the extracted features and select the optimal sequence of all the features. Experimental results show that the BGWO approach enhances the accuracy of SMS phishing detection system. The proposed method in this thesis achieves the best accuracy with 99.25% by using only an average of 87.4 of features. The results demonstrate that the proposed method has a promising performance in detecting the SMS phishing messages.