Effective risk management and compliance adherence are paramount for the success of financial institutions and organizations. However, they often face significant challenges due to fraudulent activities, with ATM fraud, among others, emerging as a prevalent issue in today’s banking landscape. We proposed a novel profiling-based one-class classification (OCC) method to solve this problem. Then training phase of our approach employs the K-Means clustering algorithm to cluster non-fraudulent transactions exhibiting similar characteristics and patterns. A rule is generated from each cluster, thereby in a rule set comprising K rules, each consisting of conditions based on the lower and upper bounds on all features. This rule set is employed to identify fraudulent transactions presented in the test phase because ours is an OCC method. One distinctive feature of our approach is its interpretability and explainability, which is crucial for understanding the model’s predictions. Overall, our proposed approach demonstrates the best performance vis-à-vis that of various state-of-the-art OCC methods in terms of classification rate. Additionally, we provide sensitivity analysis by varying the number of conditions violated across the K rules.