Deep Fakes in Healthcare: How Deep Learning Can Help to Detect Forgeries

With the increasing use of deep learning technology, there is a growing concern over creating deep fake images and videos that can potentially be used for fraud. In healthcare, manipulating medical images could lead to misdiagnosis and potentially life-threatening consequences. Therefore, the primary purpose of this study is to explore the use of deep learning algorithms to detect deep fake images by solving the problem of recognizing the handling of samples of cancer and other diseases. Therefore, this research proposes a framework that leverages state-of-the-art deep convolutional neural networks (CNN) and a large dataset of authentic and deep fake medical images to train a model capable of distinguishing between authentic and fake medical images. Specifically, the paper trained six CNN models, namely, ResNet101, ResNet50, DensNet121, DenseNet201, MobileNetV2, and MobileNet. These models had trained using 2000 samples over three classes: Untampered, False-Benign, and False-Malicious, and compared against several state-of-the-art deep fake detection models. The proposed model enhanced ResNet101 by adding more layers, achieving a training accuracy of 99%. The findings of this study show near-perfect accuracy in detecting instances of tumor injections and removals.