Deepfake technology is becoming a big problem because it makes fake videos look very real. This creates confusion and reduces trust in digital media. To solve this problem, we designed a new system to detect deepfake videos. Our method combines two powerful models: LSTM (Long Short-Term Memory) and RESNEXT. RESNEXT helps in analyzing the visual features of each video frame (spatial features), while LSTM studies the sequence of frames to understand changes over time (temporal features). By combining both image and time-based analysis, our system can better detect whether a video is real or fake. We tested our model on different datasets and performed several experiments. The results show that our approach works well in identifying manipulated videos and distinguishing them from real ones. This research helps in fighting deepfake misinformation and supports maintaining trust and authenticity on digital media platforms.
S.P.Kakade et al. (2026) studied this question.