The emergence and progression of deepfake have caused widespread concern, as it can create misleading content that can potentially manipulate public opinion. People are finding it harder and harder to tell what is real and what is a deep fake because deep fakes may modify images and videos. The human eye is not always able to detect the results of deepfake. To address this issue, a deep learning-based method has been developed that can efficiently differentiate fake videos. The proposed method for detecting manipulated videos uses a pre-trained Res-Next convolutional neural network to extract features from frames of the input video. These features are subsequently utilized to train a long short-term memory (LSTM) network, which can classify whether the input video has been tampered with or not. To enhance the model's performance on real-time data, it has been trained with a balanced dataset. The Deepfake Detection Challenge dataset was used to train the model. The model that has been proposed has yielded promising results by accurately predicting the output. This model is expected to reduce the potential harm that deep fakes pose to society.

Deepfakes, Deep learning, Res-Next Convolutional Neural Network, LongShort-Term Memory (LSTM), Deepfake Detection Challenge Dataset.