This study is a deep learning approach to classify images as either forged (fake) or pristine (original, not manipulated). The purpose of this study is to create a convolutional neural network that can recognize different types of manipulations on an image with an optimal accuracy. Some of these manipulations include plain white/black patches, filtered patches, and GAN-generated images. The CNN was created during my Data Science Internship at Factogram, a startup dedicated to the spread of fact-based information via social media. I was kindly given the chance to spearhead this project under the guidance of Factogram's founder, Tooraj Helmi: https://toorajhelmi.github.io/home