[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-115142-en":3,"doc-seo-115142-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},115142,13056703020460,"Valentina","https://ap-avatar.wpscdn.com/avatar/be000253dac470eee5d?_k=1778207105932848923",8,"Research & Report","Hybrid Deep-Learning Model for Deepfake Detection in Video - SHORT COMMUNICATION - Transfer Learning Approach","Deepfake videos undermine the authenticity and trustworthiness of visual information, creating urgent demand for accurate detection and mitigation methods. A hybrid deep-learning model is presented to improve deepfake video detection using a transfer learning strategy. The approach analyzes videos for tampering evidence through a combination of spatial feature learning and temporal modeling, enhancing sensitivity to subtle manipulations that traditional pipelines may miss.","Natl. Acad. Sci. Lett. (May–June 2025) 48(3):337–341  \n[https://doi.org/10.1007/s40009-024-01480-7](https://doi.org/10.1007/s40009-024-01480-7)  \nSHORT COMMUNICATION  \nHybrid Deep‑Learning Model for Deepfake Detection in Video using Transfer Learning Approach  \nRaksha Pandey1 · Alok Kumar Singh Kushwaha1  \nReceived: 28 August 2023 / Revised: 27 July 2024 / Accepted: 17 September 2024 / Published online: 15 October 2024 © The Author(s), under exclusive licence to The National Academy of Sciences, India 2024  \nAbstract Deepfake videos have become a growing concern in the digital age, presenting a substantial risk to the genuineness and trustworthiness of visual material. As these sophisticated manipulations continue to proliferate, there is a pressing need for advanced tools and techniques to detect and combat them effectively. In this article, we introduce a novel hybrid deep-learning model designed to enhance the accuracy of deepfake video detection using a Transfer Learning approach. Unlike traditional approaches, our hybrid model utilizes smart computer learning to carefully analyze videos for any signs of tampering. It’s akin to having a digital detective to safeguard the truth of videos.  \nKeywords Deepfake · Face swap · Face manipulation · Face to face · Transfer learning  \nIntroduction  \nDeepfakes have become a major cause for concern due to their potential for misuse, encompassing activities such as disseminating misinformation, impersonating individuals and altering visual or auditory content for nefarious intents [1, 2] . Given the prevalence of advanced artificial intelligence and digital manipulation, detecting and mitigating deepfake media have emerged as imperative tasks. Figure 1  \n* Raksha Pandey [rakshasharma10@gmail.com](rakshasharma10@gmail.com)  \nAlok Kumar Singh Kushwaha  \n[alokkumarsingh.jk@gmail.com](alokkumarsingh.jk@gmail.com)  \n1 Guru Ghasidas Vishwavidyalaya, Koni, Bilaspur, Chhattisgarh 495009, India  \nillustrates an instance of deepfake sourced from the FaceForensic + + [3] and Celeb-DF[4] datasets.  \nTransfer learning stands out as a potent strategy within deep learning, allowing models to utilize insights acquired from one task and transfer them to another [5] . This approach is particularly valuable in addressing the multifaceted challenge of Deep fake detection. By transferring knowledge from pre-trained models trained on large datasets like the DFDC [6], we can equip our hybrid model with the capacity to detect subtle manipulations across both spatial and temporal dimensions [7] . Figure 2 illustrates the Proposed Model Architecture for Video Forgery Detection.  \nMethodology  \nThe proposed video forgery detection process leverages a hybrid architecture combining deep learning and temporal modeling techniques. The architecture, as illustrated in Fig. 2, consists of the following modules:  \ni. Input Video: The initial step involves obtaining the input video, which acts as the raw material for analysis. This video can be any content requiring examination for deepfake elements.  \nii. Frame Extraction: The process of frame extraction commences, separating the video into individual frames for further analysis.  \niii. Face Detection, Alignment, and Optical Flow Esti‑ mation: Following frame extraction, the next phase involves face detection and alignment [7] . This step is critical for focusing the analysis on facial features. Additionally, optical flow estimation [8] is performed on aligned face regions in consecutive frames, calculat-  \nFig. 1 Upper row shows the Real faces from FaceForensic + + and Celeb DF dataset and Lower row shows the Fake images of respective datasets  \n\n|  |  |  |  | Face Detection Alignment & Optical Flow Estimation |\n| --- | --- | --- | --- | --- |\n|  |  |  |  |  |\n\nFig. 2 Proposed Model Architecture for Video Forgery Detection  \ning motion vectors crucial for identifying subtle movements and changes.  \niv. Deep Feature Extraction using ResNet: The aligned face regions, enr","cbCairAwg2YeH7dn","https://ap.wps.com/l/cbCairAwg2YeH7dn","pdf",717642,1,5,"English","en",105,"# Introduction\n## Transfer learning for deepfake detection\n# Methodology\n## Input video and frame extraction\n## Face detection, alignment, and optical flow estimation\n## Deep feature extraction with ResNet\n## Temporal modeling with GRU\n## Classification for binary detection\n# Rationale for GRU\n## Temporal inconsistencies\n## Temporal artifacts","[{\"question\":\"What problem does the document address?\",\"answer\":\"The document addresses the risk posed by deepfake videos, which threaten the genuineness and trustworthiness of visual content through sophisticated manipulations.\"},{\"question\":\"What is the core idea of the proposed detection model?\",\"answer\":\"It introduces a hybrid deep-learning model that improves deepfake video detection by using transfer learning together with spatial feature extraction and temporal modeling.\"},{\"question\":\"How does the methodology detect deepfakes across video frames?\",\"answer\":\"The method extracts frames, detects and aligns faces, estimates optical flow, uses a pre-trained ResNet to extract deep features, and feeds them into a GRU to capture temporal dependencies before binary classification.\"}]","Hybrid Deep-Learning Model for Deepfake Detection in Video - SHORT COMMUNICATION - Transfer Learning Approach | PDF",1785447881,13,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"hybrid-deep-learning-model-for-deepfake-detection-in-video-short-communication-transfer-learning-approach","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/hybrid-deep-learning-model-for-deepfake-detection-in-video-short-communication-transfer-learning-approach/115142/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-30",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the document address?","Question",{"text":75,"@type":76},"The document addresses the risk posed by deepfake videos, which threaten the genuineness and trustworthiness of visual content through sophisticated manipulations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the core idea of the proposed detection model?",{"text":80,"@type":76},"It introduces a hybrid deep-learning model that improves deepfake video detection by using transfer learning together with spatial feature extraction and temporal modeling.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the methodology detect deepfakes across video frames?",{"text":84,"@type":76},"The method extracts frames, detects and aligns faces, estimates optical flow, uses a pre-trained ResNet to extract deep features, and feeds them into a GRU to capture temporal dependencies before binary classification.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":21,"slug":137},19,"General","general"]