Version 1
: Received: 29 September 2023 / Approved: 9 October 2023 / Online: 9 October 2023 (16:47:31 CEST)
Version 2
: Received: 27 October 2023 / Approved: 30 October 2023 / Online: 30 October 2023 (16:14:36 CET)
Martin-Rodriguez, F.; Garcia-Mojon, R.; Fernandez-Barciela, M. Detection of AI-Created Images Using Pixel-Wise Feature Extraction and Convolutional Neural Networks. Sensors2023, 23, 9037.
Martin-Rodriguez, F.; Garcia-Mojon, R.; Fernandez-Barciela, M. Detection of AI-Created Images Using Pixel-Wise Feature Extraction and Convolutional Neural Networks. Sensors 2023, 23, 9037.
Martin-Rodriguez, F.; Garcia-Mojon, R.; Fernandez-Barciela, M. Detection of AI-Created Images Using Pixel-Wise Feature Extraction and Convolutional Neural Networks. Sensors2023, 23, 9037.
Martin-Rodriguez, F.; Garcia-Mojon, R.; Fernandez-Barciela, M. Detection of AI-Created Images Using Pixel-Wise Feature Extraction and Convolutional Neural Networks. Sensors 2023, 23, 9037.
Abstract
Generative AI has gained enormous interest nowadays due to new applications like chatGPT, DALL E, Stable Difussion and Deep Fake. Particularly DALL E, Stable Difussion and others (Adobe Firefly, ImagineArt...) are able to create images from a text prompt and are also able to recreate real photographs. Due to this fact, intense research has arisen to create new image forensics applications able to distinguish between real captured images and videos and artificial ones. Detecting forgeries made with Deep Fake is one of the most researched issues. This paper is about another kind of forgery detection. The purpose of this research aims to detect photo realistic AI created images versus real photos coming from a physical camera. For this purpose, techniques that perform a pixel level feature extraction are used. First one is Photo Response Non-Uniformity (PRNU). PRNU is a special noise due to imperfections on the camera sensor that is used for source camera identification. The underlying idea is that AI images will have a different PRNU pattern. Second one is Error level analysis (ELA). This is other type of feature extraction traditionally used for detecting image editions. In fact, ELA is being used nowadays by photographers to detect manually AI created images. Both kinds of features are used to train Convolutional Neural Networks to differentiate between AI images and real photographs. Good results are obtained achieving accuracy rates over 95%. Both extraction methods are carefully assessed by computing precision/recall and F1-score measurements.
Keywords
Artificial intelligence; AI images; photographs, PRNU; ELA; CCN; deep learning
Subject
Engineering, Electrical and Electronic Engineering
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Received:
30 October 2023
Commenter:
Fernando Martin-Rodriguez
Commenter's Conflict of Interests:
Author
Comment:
Full article has been reviewed. Some English language issues have been corrected. Some new references have been added. Discussion section has been restructured. New tests have been done. Conclusions and future lines are integrated in the discussion part.
Outstanding additions to paper: - Customizing a pre-trained AlexNet structure to compare it with the original CNN. Results are not so good. - Creating a new (smaller) dataset for testing. NO IMAGE IN COMMON WITH TRAINING DATASET. Besides, AI images have been created using different tools. Results are only a bit worse. What's more, this test results demonstrate that errors are generally of "false negative" type (AI image being recognized as real photo). This suggests an easy method for integrating the both kinds of feature extraction getting a more robust system, this possibility is documented in the new discussion section.
Some new figures have been added, others modified.
Commenter: Fernando Martin-Rodriguez
Commenter's Conflict of Interests: Author
Outstanding additions to paper:
- Customizing a pre-trained AlexNet structure to compare it with the original CNN. Results are not so good.
- Creating a new (smaller) dataset for testing. NO IMAGE IN COMMON WITH TRAINING DATASET. Besides, AI images have been created using different tools. Results are only a bit worse. What's more, this test results demonstrate that errors are generally of "false negative" type (AI image being recognized as real photo). This suggests an easy method for integrating the both kinds of feature extraction getting a more robust system, this possibility is documented in the new discussion section.
Some new figures have been added, others modified.