A Lightweight CNN for Efficient Deepfake Detection of Low-resolution Images in Frequency Domain

Deepfake detection in frequency domain is extensively explored for two reasons –artifacts generated by the up-sampling layer of Generative Adversarial Networks are more prominent in the frequency domain and some image processing tasks are simpler and efficient in this domain. Traditional convolutional neural networks outperform the machine learning models for their ability to detect deepfakes of low-resolution images also. But these networks involve numerous trainable parameters resulting in huge computation cost, time and high memory requirements. This paper presents a lightweight Convolutional Neural Network (CNN) architecture that is optimized for efficient deepfake detection of low-resolution images in the frequency domain. The proposed model achieves almost the same accuracy as that of the traditional network but with the parameter requirement reduced by 92%. This renders the model suitable for use in memory-constrained and power-constrained environments such as mobile devices and other embedded systems.