Abstract: While DCGAN as deep learning model utilizing spectrogram, allows for detection of deepfake audio, it is prone to overfitting which affects its ability to discriminate between real and fake ...