SpectraNet: FFT-assisted Deep Learning Classifier for Deepfake Face Detection

TL;DR

SpectraNet introduces a lightweight deep learning model utilizing FFT-based features and advanced training techniques to improve deepfake face detection accuracy, stability, and accessibility for non-experts.

Contribution

The paper presents a novel, generalizable deepfake detection framework combining EfficientNet-B6 with transformation techniques and robust preprocessing, with minimal reliance on Fourier features.

Findings

High accuracy and stability in deepfake detection

Effective handling of class imbalance through oversampling and preprocessing

FFT-based features had minimal impact on performance

Abstract

Detecting deepfake images is crucial in combating misinformation. We present a lightweight, generalizable binary classification model based on EfficientNet-B6, fine-tuned with transformation techniques to address severe class imbalances. By leveraging robust preprocessing, oversampling, and optimization strategies, our model achieves high accuracy, stability, and generalization. While incorporating Fourier transform-based phase and amplitude features showed minimal impact, our proposed framework helps non-experts to effectively identify deepfake images, making significant strides toward accessible and reliable deepfake detection.