• Kanchan Warkar
• Dr. Rishikesh Rawat

The development of generative artificial intelligence has led to the advanced generation of synthetic videos and images with high visual quality. Nowadays, generative AI is widely used in deepfake technology to create fake videos, audio, and images. Deepfake technology poses a great threat to the authenticity of visual content, especially in live-stream environments where quick detection is crucial. Current methods for detecting deepfakes face various limitations and challenges, highlighting the need for more reliable and accurate solutions. To address these critical challenges, we propose a hybrid deep learning model using a feature extraction and optimization approach based on the Marine Predators' Algorithm (MPA). For feature extraction, we employ the Xception network V3. The MPA variant enhances the Xception network features for CR-CNN, along with key learning configurations of CNNs, aiming to extract effective discriminative spatial-temporal cues. Both weighted and hybrid learning strategies are utilized, using the optimized networks as base classifiers. Specifically, the MPA algorithm identifies optimal subsets of these optimized base networks for hybrid deep learning generation, balancing complexity and performance. When evaluated on several well-known synthetic video datasets, our approach outperforms existing methods and various deep learning models developed by other search strategies, showing statistical significance in video authenticity classification. Additionally, the proposed hybrid deep learning model demonstrates superior statistical performance compared to multiple search methods for solving optimization problems across various artificial landscapes with diverse geometrical structures.

Deepfake, Videos, Deep Learning, Optimization, MPA, Feature Extraction

"Hybrid Deep Learning Technique for Optimized Feature Detection in Deepfake Videos", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 9, page no.c694-c701, September-2025, Available :http://www.jetir.org/papers/JETIR2509276.pdf

"Hybrid Deep Learning Technique for Optimized Feature Detection in Deepfake Videos", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 9, page no. ppc694-c701, September-2025, Available at : http://www.jetir.org/papers/JETIR2509276.pdf