Fingerprints of authenticity: Exploring AI solutions for detecting deep fakes

Abstract

The human hand is one of the most extraordinary parts of the human body, possessing unparalleled dexterity and versatility. When trying to replicate the human hand for robotics, engineers, and scientists are puzzled by the complexity of the human hands. In the book, Design of Artificial Hands, the authors explain that “[t]he human hand is capable of performing complex and useful tasks using an effective integration of mechanisms, sensors, actuators, and control functions, and at the same time is also a cognitive instrument, allowing humans to develop a superior brain by interacting with the surrounding environment” (Balasubramanian and Santos, 2014). • While attempts to emulate the human hand in robotics continue, artificial intelligence (AI) confronts its own challenges in replicating hands, particularly in generating realistic images. Despite advancements in artificial intelligence technology, models will often struggle to accurately depict human hands, leading to the inclusion of extra hands or hands that look inhumane or unnatural. An easy way to distinguish between an AI-generated image and a genuine photograph is to look at the hands. If an extra finger is present, it is likely an AI-generated image rather than a genuine picture. • In today's day and age, where the authenticity of images and photos is increasingly difficult to ascertain, the need for automated methods to detect if images are genuine is vital. To address this challenge, the development of computer vision models has emerged as a promising approach. Specifically, by creating a computer vision model capable of detecting the presence of hands and incorporating a classifier to discern their authenticity, we can easily and significantly enhance the image verification process. • To create a model like this, one must understand how the images are created in the first place. To do that, one must understand the basics of Machine Learning and how it can be used.