A deepfake is a computer-generated image or video that appears to be real but is a fabricated representation created to make an individual appear to be saying or doing something that did not occur. Deepfakes generate misleading or deceptive information by manipulating and superimposing faces onto pre-existing footage using artificial intelligence. This paper introduces a novel approach for deepfake detection through a combination of EfficientNet and Recurrent Neural Networks (RNNs). This method enhances detection efficiency by leveraging the hierarchical features acquired by EfficientNet and employing RNNs, specifically Long Short-Term Memory (LSTM) networks, to capture temporal dependencies. Application of this approach to the Celeb-DF dataset resulted in an accuracy of 99.98%.