Real-Time Deepfake Detection using a Hybrid MobileNet-LSTM Model For Image and Video Analysis
S Ambika, Yerramsetty Harini, B R Sumanth · 2025
As deepfake technology has become more advanced, there is an urgent need for strengthening detection mechanisms that can help counter misinformation and online fraud. Deepfake media is manufactured using deep learning techniques. It has recently become increasingly sophisticated and the boundary between genuine and tampered content is often hard to see. This paper introduces a hybrid MobileNet-LSTM model, which is made for spotting deepfakes in both image and video.The deepfake identification system uses MobileNet to efficiently sample spatial features and LSTM to gather temporal dependencies, thereby improving the detection rate of tampered content. Meanwhile, this method is validated on standard datasets and that the accuracy is better, in terms of computational efficiency and that it is more robust than other types of manipulation.