The rapid advancement and spread of deepfake videos have become a major concern due to their ability to convincingly mislead viewers, including trained professionals. This work aims to address the challenges of detecting deepfakes, especially in videos with low resolution and short durations—scenarios that often hinder detection accuracy. In this study, we developed a deep learning-based binary classification model using a Long Short-Term Memory (LSTM) architecture. The model was trained and evaluated on the FaceForensics++ dataset, which provides a diverse range of manipulated and authentic video samples. In this project LSTM-based model achieved an accuracy of 94.63%, demonstrating strong performance in identifying fake content under challenging conditions. The results highlight the effectiveness of leveraging temporal information in videos for improved detection. This research contributes to the growing field of deepfake detection by offering a reliable and automated approach, aiming to reduce the harmful effects of synthetic media on public discourse, trust, and societal integrity.