The proliferation of the Internet of Things (IoT) has led to an unprecedented integration of diverse sensors, driving innovation across numerous domains. However, the reliability and security of IoT networks are significantly challenged by the presence of faulty sensors. Traditional fault detection methods are inadequate to manage the scale and complexity of modern IoT environments. This paper addresses the challenge of identifying faulty sensors in large-scale IoT networks by proposing a hybrid fault detection model that integrates deep learning and distrust mechanisms. Tested on simulated Hanoi air pollution data, the model demonstrates high accuracy and effectiveness, surpassing traditional fault detection methods. This approach provides a scalable, efficient solution to enhance the reliability of IoT networks.