Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AI-Powered Deepfake Detection Using CNN and Vision Transformer Architectures

Sifatullah Sheikh Urmi, Kirtonia Nuzath Tabassum Arthi, Md Al-Imran

TL;DR

This work addresses deepfake detection by comparing CNN and Vision Transformer architectures on a large real-vs-fake face dataset. It combines a preprocessing-and-augmentation pipeline with four models—DFCNET, MobileNetV3, ResNet50, and VFDNET—to assess generalization and efficiency. The results show VFDNET achieving the highest accuracy (~$99.13\%$) and strong robustness, while MobileNetV3 offers a compelling efficiency-performance trade-off; ResNet50 lags behind, and DFCNET shows moderate performance. The findings demonstrate the viability of transformer-based and lightweight architectures for reliable, scalable deepfake detection in practical settings.

Abstract

The increasing use of artificial intelligence generated deepfakes creates major challenges in maintaining digital authenticity. Four AI-based models, consisting of three CNNs and one Vision Transformer, were evaluated using large face image datasets. Data preprocessing and augmentation techniques improved model performance across different scenarios. VFDNET demonstrated superior accuracy with MobileNetV3, showing efficient performance, thereby demonstrating AI's capabilities for dependable deepfake detection.

AI-Powered Deepfake Detection Using CNN and Vision Transformer Architectures

TL;DR

This work addresses deepfake detection by comparing CNN and Vision Transformer architectures on a large real-vs-fake face dataset. It combines a preprocessing-and-augmentation pipeline with four models—DFCNET, MobileNetV3, ResNet50, and VFDNET—to assess generalization and efficiency. The results show VFDNET achieving the highest accuracy (~) and strong robustness, while MobileNetV3 offers a compelling efficiency-performance trade-off; ResNet50 lags behind, and DFCNET shows moderate performance. The findings demonstrate the viability of transformer-based and lightweight architectures for reliable, scalable deepfake detection in practical settings.

Abstract

The increasing use of artificial intelligence generated deepfakes creates major challenges in maintaining digital authenticity. Four AI-based models, consisting of three CNNs and one Vision Transformer, were evaluated using large face image datasets. Data preprocessing and augmentation techniques improved model performance across different scenarios. VFDNET demonstrated superior accuracy with MobileNetV3, showing efficient performance, thereby demonstrating AI's capabilities for dependable deepfake detection.
Paper Structure (27 sections, 13 equations, 6 figures, 3 tables)

This paper contains 27 sections, 13 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Literature review
  3. Methodology
  4. Preprocessing Pipeline for Enhanced Model Generalization
  5. Fundamental CNN Layers
  6. Pooling Layer
  7. Convolutional Layer
  8. Fully Connected (FC) or Dense Layer
  9. Flatten Layer
  10. Commonly used Activation Functions
  11. Rectified Linear Unit
  12. Sigmoid Activation Function
  13. Framework for the detection of fake images
  14. Deepfake Convolutional Network (DFCNET)
  15. Vision Fake Detection Network (VFDNET)
  16. ...and 12 more sections

Figures (6)

  • Figure 1: Schematic Workflow of the Proposed Method
  • Figure 2: Fully Connected (FC) layer with two layers.
  • Figure 3: Architecture of the Vision Fake Detection Network (VFDNET)
  • Figure 4: Training and validation accuracy and loss curves for all employed models.
  • Figure 5: Class-wise Prediction Performance via Confusion Matrices of Proposed Models
  • ...and 1 more figures