Survey Paper: Comparative Analysis of Detection Methods for Real vs. AI-Generated Images

The rapid evolution of AI-generated image technologies, including GANs and diffusion models, has blurred the lines between real and synthetic content. In this survey, we explore state-of-the-art methods for detecting fake images. Traditional forensic techniques, deep learning approaches, and advance...

Full description

Saved in:
Bibliographic Details
Published in2025 International Conference on Artificial Intelligence and Data Engineering (AIDE) pp. 659 - 664
Main Authors Salam, Ridha, J, Sahanna Rachel, Kumar B, Naveen, V, Ramachandra H, K, Sameeksha M, Madari, Vishwa
Format Conference Proceeding
LanguageEnglish
Published IEEE 06.02.2025
Subjects
AI-generated images
CNN
Deep learning
Diffusion models
Ensemble Learning
Explainable AI
Feature Analysis
Feature extraction
Image Authentication
LLM
Media
Real-time systems
Robustness
Scalability
Surveys
Transfer learning
Online AccessGet full text
DOI10.1109/AIDE64228.2025.10986774

Cover

More Information
Summary:The rapid evolution of AI-generated image technologies, including GANs and diffusion models, has blurred the lines between real and synthetic content. In this survey, we explore state-of-the-art methods for detecting fake images. Traditional forensic techniques, deep learning approaches, and advanced methods like Explainable AI (XAI) and blockchain-based verification are used. We evaluate these methods using performance metrics and derive insights into their strengths and limitations. A comparative analysis of existing research papers is provided, examining the approaches used, feature extraction techniques, classifiers employed, and datasets utilized. Additionally, it puts forth actional research directions like diffusion models, multimodal analysis, and expansion of dataset that can contribute towards the advancement of this field. Practical examples and visualizations make this survey both insightful and tutorial-like, offering valuable guidance for researchers and practitioners.
DOI:10.1109/AIDE64228.2025.10986774