[4K1-IS-2d] Machine learning

[4K1-IS-2d-05] Detecting Real vs. AI-Generated Arab Faces with Machine Learning Using CNN Models

Artificial Interlligance in Face detection

〇Leen N Khraim², Nasrullah M Khraim¹, Fatima H Alshamsi¹, Khalid AlFalasi¹ (1. Taqa Distribution Company, 2. Jordan University of Science & Technology)

Keywords:Artificial intelligence, Convolutional neural network, Generative Adversarial Network , Artificial Neural Network

This study addresses the challenges of distinguishing between real and AI-generated images in the context of Arab facial features, utilizing convolutional neural networks (CNNs). By creating a specialized dataset of 17,000 real and AI-generated images from the Arab League, the research fills a critical gap in datasets representing this demographic. The CNN model achieved a remarkable accuracy of 99.89%, significantly improving the reliability of image authenticity detection. The findings aim to mitigate biases in existing AI systems, support misinformation prevention, and enable culturally sensitive applications of facial recognition technologies. Using machine learning to detect real versus AI-generated Arab faces enhances cybersecurity in power control centers by preventing deepfake threats, ensuring personnel authenticity, and safeguarding critical systems against unauthorized access.