A Machine Learning Framework for Enhancing Security of Transaction in Saudi Banks Based on User Behavior

Nowadays, with most of the world operating remotely, online banking is very popular in Saudi Arabia. However, fraudsters often set up fake websites or apps to obtain bank account information, which they use to scam and steal money. They may create fake transactions or manipulate genuine ones to transfer funds to another account they own. This widespread problem requires solutions from banks to reduce the incidence of bank fraud. Our proposed system aims to tackle this issue by analyzing user behavior, identifying unusual behavior, and alerting users to stop the process if necessary. In this research, we applied two different models with two alternative datasets: one is real dataset, while the second simulated dataset. This research evaluated the performance of two different models: first is hybrid neuro-fuzzy model based on combination of deep neural networks and fuzzy logic algorithm (DNF), while the second is deep neural network (DNN) model. The result of our experiment showed that the DNN model achieved the highest accuracy by reaching 99.95%, while the DNF model is faster which seems to be more acceptable in real-time transactions.