This paper presents a multimodal signal processing pipeline for detecting synthetic media (deepfakes). By analyzing characteristic artifacts in both image and audio domains—specifically through Canny edge density, FFT frequency analysis, and MFCC spectral feature extraction—the system provides a lightweight, deterministic alternative to deep learning models. This work was conducted as part of the B.Tech Final Phase project in Cyber Security at Jain Deemed-to-be-University, Bengaluru.
perumalla et al. (Sun,) studied this question.