It is increasingly common for financial institutions and government entities to adopt face verification programs, streamlining application processes. Despite the benefits, a growing concern is the exploitation of these systems by fraudsters who use fake ID images, resulting in substantial financial losses, commonly known as Identity Theft (IT) fraud. To address this challenge, we propose the integration of advanced deep learning techniques, specifically employing a Bayesian Convolutional Neural Network along with some preprocessing methods. This approach achieves a considerably higher success rate than industry practices. Its primary goal is to identify and flag fake images in real time during the application process, providing a proactive solution to prevent potential losses before they occur.