DEEPFAKES - CONSEQUENCES AND STEPS TO BE TAKEN

Abstract

Deepfakes is one among many other rapidly advancing technologies in the tech industry. These are fake AI-generated videos or audio resembling a certain subject. The deep learning methods for producing deepfakes involve training generative neural network architectures like autoencoders and Generative Adversarial Networks (GAN). Since 2018, the number of deepfake online videos has doubled. It is a fascinating technology yet mostly misused technology. It was reported that 96% of deepfakes are pornographic videos mainly targeting women, especially female celebrities and even children. The present manuscript is a review paper on consequences of deepfakes in various domains which include digital impersonation, manipulation, blackmail, misinformation, misleading, public shaming, fabricating evidence, financial fraud etc. Deepfakes can be used to damage the reputation of celebrities, politicians and any public figures.  (Paper Modi)

 Initially, deepfakes were mostly created by experts but now many accessible applications are available to create deepfakes thereby drastically increasing the risk of it being misused. Although deep fakes have many drawbacks, they also have advantages such as hyper-personalization, better dubbing, cheaper video campaigns, creating films with lesser utilisation of time and labour, making dynamic images and videos of the deceased, improved video game characters etc. To combat deepfakes, certain laws and safety measures are being implemented.