Decoding Deepfake Detection: Harnessing the Strengths of Traditional Machine Learning for Superior Accuracy
P. Yogendra Prasad, Maddula Lohitha, Gandikota Sairam Roopak, Dommaraju Sai Harshini, Kandipati Sumanth · Advances in computer science research · 2024
This comprehensive study delves into the dynamic landscape of deep learning applications, focusing on the burgeoning realm of deep fakes.Deep learning has seamlessly integrated into fields like natural language processing, machine learning, and computer vision, giving rise to innovative applications.However, the surge in deep fakes, sophisticatedly manipulated videos/images, has become a pressing concern.The nefarious applications of this technology, such as fake news, celebrity impersonations, financial scams, and revenge porn, pose significant threats in the digital realm.Particularly, public figures like celebrities and politicians are highly susceptible to the Deep fake detection challenge.This research systematically assesses both the production and detection aspects of deep fakes, employing diverse deep learning algorithms, including In-ceptionResnetV2, VGG19, CNN, and Xception.The evaluation, conducted on a Kaggle deep fake dataset, highlights Xception as the most accurate among the algorithms studied.As malicious uses of deep fakes escalate, the imperative for robust detection mechanisms intensifies to safeguard against potential societal consequences.