The modern adoption of e-Commerce has accelerated the need for effective customer protection, as part of the roadmap to expanding e-Commerce in smart cities. Before fully adopting e-Commerce for smart city applications, there are a few challenges that need to be addressed; mainly security and fraud risks. Here, credit cards are the most popular target, as a fraudster does not need much to carry out a ‘card not present’ transaction. Many techniques are used in identifying sources of cyber fraud, yet their detection accuracy degrades substantially when they are applied to datasets with an extremely large number of features and training samples. The objective of this research is to propose a new model to accurately identify sources of fraud in e-Commerce based on features extraction. This research is needed for early detection of fraud in order to avoid further occurrences of these crimes. In attaining the objectives of this research, an intensive data analysis will be performed to extract fraudulent detection patterns, which assists in constructing an effective classifier to identify sources of fraud. To construct the classifier, a Support Vector Machine (SVM) will be used as the base of our proposed model, along with a new algorithm called the multi-verse optimizer for fraudulent features extraction. The proposed approach, the credit card fraud detection model, is based on multi-verse features extraction (CCFD-MVFEX). The significance of this research comes from its ability to optimize the accuracy of detecting sources of fraud in e-Commerce, as well as effectively training the fraud detection model on different sets of fraudulent features. This will leverage the confidence of e-Commerce users. The results illustrate that the produced features extraction model can identify fraudulent sources with high accuracy rates compared with other representative detection methods, using four online data sources.