Due to the rapid advancement in ecommerce and in information technology, the fraud correlated with this field is also increased. This environment deals large amount of data, so efficient data analysis method is needed to find the fraudulent activities. The main challenging issue in fraud detection is handling the imbalanced datasets. The genuine activities overlap the fraudulent, so it makes the complexity in the findings. Binary classification is the best choice to handle the fraud detection problem. In this paper, Cost Based Support Vector Machines (CS-SVM) is applied to find the fraudulent activities. CSSVM plays a vital role in handling the nonlinear and imbalanced data and it is well suited for the binary classification.