Development of Enhanced Video Manipulation Detection Using Hybrid CNN-LSTM Method for Object Forgery

The rise of sophisticated video manipulation techniques, such as deepfakes and object-based forgeries, has heightened the need for robust and efficient forgery detection systems. This study proposes a novel EfficientNet-LSTM hybrid model that combines EfficientNet for spatial feature extraction and LSTM networks for temporal analysis, enabling the detection of both spatial and temporal inconsistencies in video sequences. Evaluated on the VTD dataset, this proposed model achieves 85% accuracy, outperforming traditional CNN-based methods in terms of precision, robustness, and computational efficiency. By addressing the limitations of existing approaches, this work provides a scalable and reliable solution for real-time video forgery detection, advancing the field of multimedia forensics and network security.