journal-of-information-and-optimization-sciences Journal of Information and Optimization Sciences 2169-0103 0252-2667 Taru Publications 10.47974/JIOS-2357 A multi-level trusted security framework for high-fidelity GAN video deepfake detection in real-time systems Kumar Sanjay Department of Basic Sciences and Humanities, Pranveer Singh Institute of Technology, Kanpur, Uttar Pradesh, 209305, India Awasthi Deepak Department of Mechanical Engineering, Pranveer Singh Institute of Technology, Kanpur, Uttar Pradesh, 209305, India Pandey Ravi Krishan Department of Computer Science & Engineering, Faculty of Engineering and Technology, University of Lucknow, Lucknow, Uttar Pradesh, 226007, India Uttam Arun Kumar Department of Computer Applications, PSIT College of Higher Education, Kanpur, Uttar Pradesh, India, 209305 Gupta Shashi Kant Department of Computer Science and Applications, ITM University, Gwalior, Madhya Pradesh, 475001, India 47 8 3063 3072 31 08 2026

High-fidelity GANs enable realistic deepfakes, increasing risks of misinformation and fraud. This paper proposes a multi-level trusted framework combining spatio-temporal forensics, hybrid classification, and cryptographic validation for real-time detection. Results show improved accuracy, robustness, and low latency, offering a secure, scalable solution for applications like digital forensics, authentication, surveillance, and social media verification.

 Deepfake detection High-fidelity GAN Video forensics Real-time detection Trusted security framework Multi-level authentication Spatio-temporal features Integrity verification Hybrid deep learning Cryptographic validation Decision fusion Adversarial robustness access open retracted no