Deepfake detection has become a critical area of research with the growing prevalence of sophisticated face manipulation technologies, which pose severe ethical and security challenges. In this study, we propose an advanced deepfake detection system leveraging the EfficientNetB0 model, a state-of-the-art convolutional neural network (CNN) architecture, to address the challenges of efficiency and accuracy in identifying manipulated media. Our system utilizes video frame extraction and comprehensive data augmentation techniques to preprocess inputs, ensuring enhanced generalization on limited training data. EfficientNetB0, pre-trained on the ImageNet dataset, serves as the backbone for feature extraction, employing its highly efficient architecture with depth wise separable convolutions. Evaluated on the Celeb-DF dataset, the proposed system demonstrates high accuracy and robustness in detecting deepfake content while maintaining computational efficiency, making it suitable for real-world applications. Experimental results validate the effectiveness of this approach, highlighting its potential to contribute significantly to mitigating the adverse impacts of deepfakes.