Image quality assessment based fake face detection

The tremendous growth of data in social media and other platforms has raised an interesting question of authenticity. It has led to an active research area named Digital Forensics. Especially, the face manipulation has become a major issue in character assassination. The image forgery tools are impr...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2023-03, Vol.82 (6), p.8691-8708
Main Authors: S, Kiruthika, V, Masilamani
Format: Article
Language:English
Subjects:
1209: Recent Advances on Social Media Analytics and Multimedia Systems: Issues and Challenges
Computer Communication Networks
Computer Science
Data Structures and Information Theory
Deception
Face recognition
Feature extraction
Forensic computing
Forgery
Frequency domain analysis
Image manipulation
Image quality
Multimedia Information Systems
Quality assessment
Special Purpose and Application-Based Systems
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The tremendous growth of data in social media and other platforms has raised an interesting question of authenticity. It has led to an active research area named Digital Forensics. Especially, the face manipulation has become a major issue in character assassination. The image forgery tools are improving everyday thereby posing a challenge in detection systems. The current detection systems provide deep learning based solutions which do not bring the reliability and also have chance to fail when different forgery tool is developed to synthesise or edit the face image. Therefore, an efficient system is required which gives explainability along with efficacy. In this paper, we propose a novel method to detect the forged faces using Image Quality Assessment(IQA) based features. As far as we know IQA has not been used for detecting AI generated images. Despite the visual appearance being same for original and fake images, most of the discriminative information will be available in the frequency domain of those images. With that intuition we have extracted image quality based features from frequency domain and also spatial domain. The proposed method has achieved the highest accuracy of 99% when different types of experiments were performed on standard datasets. The generalisation and explainability of the proposed model have also been discussed.
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-021-11493-9