Guardian-AI: A novel deep learning based deepfake detection model in images
The rapid advancement of deepfake technology has introduced significant challenges and opportunities across various domains. This article proposes a robust deepfake detection pipeline utilising a combination of attention mechanisms, pre-trained Vision Transformers (ViTs), and Long Short-Term Memory...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-07-01
|
| Series: | Alexandria Engineering Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1110016825005927 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1839626708910604288 |
|---|---|
| author | Hadeel Alsolai Khalid Mahmood Asma Alshuhail Achraf Ben Miled Mohammed Alqahtani Abdulrhman Alshareef Fouad Shoie Alallah Bandar M. Alghamdi |
| author_facet | Hadeel Alsolai Khalid Mahmood Asma Alshuhail Achraf Ben Miled Mohammed Alqahtani Abdulrhman Alshareef Fouad Shoie Alallah Bandar M. Alghamdi |
| author_sort | Hadeel Alsolai |
| collection | DOAJ |
| description | The rapid advancement of deepfake technology has introduced significant challenges and opportunities across various domains. This article proposes a robust deepfake detection pipeline utilising a combination of attention mechanisms, pre-trained Vision Transformers (ViTs), and Long Short-Term Memory (LSTM) networks. The initial phase of the pipeline involves preparing photos and videos, potentially using optional facial detection to enhance accuracy. Vision Transformers derive features by capturing the global dependencies of input data. Long short-term memory (LSTM) networks address inter-frame temporal dependencies, whereas multi-head and traditional attention processes focus on essential elements. Ultimately, fully connected layers are employed for classification within the ensemble architecture, which consolidates the outcomes of several models. To ensure generalisability, assessment and regularisation approaches are employed to train the model using labelled datasets. Given the escalating threat of deepfakes, the findings indicate that the pipeline can consistently distinguish between genuine and fabricated information. |
| format | Article |
| id | doaj-art-fd85944edab844588a05e46039aa6f40 |
| institution | Matheson Library |
| issn | 1110-0168 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Alexandria Engineering Journal |
| spelling | doaj-art-fd85944edab844588a05e46039aa6f402025-07-17T04:44:06ZengElsevierAlexandria Engineering Journal1110-01682025-07-01126507514Guardian-AI: A novel deep learning based deepfake detection model in imagesHadeel Alsolai0Khalid Mahmood1Asma Alshuhail2Achraf Ben Miled3Mohammed Alqahtani4Abdulrhman Alshareef5Fouad Shoie Alallah6Bandar M. Alghamdi7Department of Information Systems, College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University PO Box 84428, Riyadh 11671, Saudi ArabiaDepartment of Information Systems, Applied College at Mahayil, King Khalid University, Saudi ArabiaDepartment of Information Systems, College of Computer Sciences & Information Technology, King Faisal University, Saudi ArabiaDepartment of Computer Science, College of Science, Northern Border University Arar 73213, Saudi Arabia; Corresponding author.Department of Information System and Cyber Security, College of Computing and Information Technology, University of Bisha, Bisha 61922, Saudi ArabiaDepartment of Information Systems, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi ArabiaDepartment of Information Systems, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi ArabiaDepartment of Information Technology, Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah 21589, Saudi ArabiaThe rapid advancement of deepfake technology has introduced significant challenges and opportunities across various domains. This article proposes a robust deepfake detection pipeline utilising a combination of attention mechanisms, pre-trained Vision Transformers (ViTs), and Long Short-Term Memory (LSTM) networks. The initial phase of the pipeline involves preparing photos and videos, potentially using optional facial detection to enhance accuracy. Vision Transformers derive features by capturing the global dependencies of input data. Long short-term memory (LSTM) networks address inter-frame temporal dependencies, whereas multi-head and traditional attention processes focus on essential elements. Ultimately, fully connected layers are employed for classification within the ensemble architecture, which consolidates the outcomes of several models. To ensure generalisability, assessment and regularisation approaches are employed to train the model using labelled datasets. Given the escalating threat of deepfakes, the findings indicate that the pipeline can consistently distinguish between genuine and fabricated information.http://www.sciencedirect.com/science/article/pii/S1110016825005927Deep learningDeepfakeCDDBViT transformerAttention mechanism |
| spellingShingle | Hadeel Alsolai Khalid Mahmood Asma Alshuhail Achraf Ben Miled Mohammed Alqahtani Abdulrhman Alshareef Fouad Shoie Alallah Bandar M. Alghamdi Guardian-AI: A novel deep learning based deepfake detection model in images Alexandria Engineering Journal Deep learning Deepfake CDDB ViT transformer Attention mechanism |
| title | Guardian-AI: A novel deep learning based deepfake detection model in images |
| title_full | Guardian-AI: A novel deep learning based deepfake detection model in images |
| title_fullStr | Guardian-AI: A novel deep learning based deepfake detection model in images |
| title_full_unstemmed | Guardian-AI: A novel deep learning based deepfake detection model in images |
| title_short | Guardian-AI: A novel deep learning based deepfake detection model in images |
| title_sort | guardian ai a novel deep learning based deepfake detection model in images |
| topic | Deep learning Deepfake CDDB ViT transformer Attention mechanism |
| url | http://www.sciencedirect.com/science/article/pii/S1110016825005927 |
| work_keys_str_mv | AT hadeelalsolai guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT khalidmahmood guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT asmaalshuhail guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT achrafbenmiled guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT mohammedalqahtani guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT abdulrhmanalshareef guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT fouadshoiealallah guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages AT bandarmalghamdi guardianaianoveldeeplearningbaseddeepfakedetectionmodelinimages |