Unmasking Digital Deceptions: A Comprehensive Survey of Synthetic Reality Analysis Across Multimedia Domains

• Published in: 2024 International Conference on Electronic Systems and Intelligent Computing (ICESIC) pp. 202 - 207
• Main Authors: Chaudhary, Sneha, Chaurasiya, Komal, Sriram, Suthir, S, Ravikumar, V, Nivethitha, M, Thangavel
• Format: Conference Proceeding
• Language: English
• Published: IEEE 22.11.2024
• Subjects: Accuracy; Analysis; Artificial Intelligence; Deep learning; Deepfake; Deepfakes; Hoaxes; Neural Networks; Protection; Real-time systems; Social networking (online); Surveys; Technological innovation; Video on demand; Web sites
• DOI: 10.1109/ICESIC61777.2024.10846333

The general accessibility of social networking sites like Facebook, Myspace and TikTok has transformed the way information is communicated, but it has also made it simpler to spread misleading information, thanks to cutting-edge technologies like deepfakes. The use of artificial intelligence in deepfake technology produces material that is remarkably life- like but fake, making it difficult to identify and remove fraudulent content. The evolution of deepfake technology is reviewed in this study, starting with early breakthroughs like the Video Rewrite Program and continuing with more current developments like the Synthesizing Obama and Face2Face projects. We provide a comprehensive analysis of the state-of-the-art detection strategies, namely Region-based CNNs (RCNNs), Conventional Neural Networks (CNNs), and hybrid approaches that combine various deep learning techniques. Though these methods have advanced, there are still issues with accuracy, integration complexity, and adaptability to new deepfake techniques. This paper suggests a novel approach that integrates image, video, and audio analysis into a single detection framework in order to overcome these difficulties. The new approach is directed towards improvement of detection accuracy and the lowering of the false alarm using advanced and deep learning engines along with the real time processing requirements. Early results indicate that the detection capabilities have significantly improved compared to existing solutions operating in real time environment while having less latency and higher recall and precision rates. This method is more effective for deepfake detection and at the same time helps in augmenting the comprehensiveness and efficiency of the system there by making it suitable for wider uses. The study emphasizes the presence of the need for the evolution of static approaches towards the effective defection of sophisticated false information and the justification of new possibilities in the fight against disinformation and protection of information space in general.