A Study on Deepfake Detection Methods

The continued development of deepfake Generative Adversarial Networks (GANs) have generated much controversy within the multiple fields including politics, entertainment, and healthcare. Deepfakes cause the generation of highly convincing fake news to spread with the aim of influencing public opinions, and given that the pictures are near realistic and the sound quality is excellent, it gets hard to differentiate between original and fake content. The release of apps for creating deepfakes has greatly increased the need to find ways to mitigate them because the consequences are criminal on the levels of fake news and privacy infringement. For instance, deepfake technology is capable of manipulating political messages, deceiving the electorate, and inaccurately portraying diseases with likelyhoods of wrong diagnosis and subsequent wrong treatment. This present paper presents a comprehensive survey and analysis of the existing deepfake detection methods including the traditional machine learning techniques, state of the art deep learning methods, and the recent approach based on blockchain technology. Unlike some previous literature reviews that may only include new literature from the last few years, this research synthesizes results from more recent findings to explicate system limitations and recent developments. The presented research is designed to provide methodological approaches and findings to the discussion on deepfake detection to facilitate further empirical work and practical developments beneficial for the crucial field. For deepfake technology lies on the path of digital media’s continuous growth, therefore critical detection methods must be enhanced and united to maintain the desired information’s purity and shield society against potential risks.