Multi Model Deepfake Detection System Using Audio, Video and Image Feature Fusion

Deepfake technology has emerged as a major threat due to its ability to generate highly realistic fake multimedia content. Existing deepfake detection systems mainly focus on single modalities such as images or videos, which limits their effectiveness. This paper proposes a multimodal deepfake detection system that integrates audio, video, and image features to improve detection accuracy. The system extracts frames from videos and applies Convolutional Neural Networks (CNN) for visual feature extraction. Simultaneously, audio signals are processed using Mel-Frequency Cepstral Coefficients (MFCC) to extract speech features. These features are fused and passed to a classification model to determine whether the content is real or fake. Experimental results demonstrate that the proposed multimodal approach improves robustness and achieves higher accuracy compared to single-modal systems