Deepfake technology, which leverages deep learning to synthesize realistic yet fraudulent videos, has emerged as a major threat to media authenticity, enabling the spread of misinformation and identity manipulation. This paper presents an overview of cutting-edge DL methodologies for detecting deepfakes in videofiles. The study incorporates enhancement methods such as CLAHE [26] and green channel conversion to improve detection accuracy. Additionally, we analyze the impact of various activation functions on model performance. Finally, the research underscores the noteworthiness of developing robust, generalizable-methodology capable of adapting for the growing sophistication of deepfake technologies across various platforms and applications.