Deepfake a synthetic multimedia is created easily using zero investment and very little time, pose a significant challenge. Generated fake content using deepfake can easily create havoc. People using deepfake can easily create high quality malicious content which looks real in one look and circulate to create misinformation. It is necessary for researchers to come up with classifier which can detect deepfake content. This study investigates the effectiveness of various efficient Convolutional Neural Network (CNN) architectures for deepfake content detection. We compare the performance of MobileNet, MobileNetV2, DenseNet, EfficientNet, EfficientNetV2 and XceptionNet on a comprehensive datasets. Finally, the evaluation of the performance of above models on different images are measured in terms of accuracy. After the analysis of the models, proposed study found the highest accuracy of XceptionNet i.e. 94%.