{"title":"A statistical analysis for deepfake videos forgery traces recognition followed by a fine-tuned InceptionResNetV2 detection technique","authors":"Sandhya MSc, Abhishek Kashyap PhD","doi":"10.1111/1556-4029.15665","DOIUrl":null,"url":null,"abstract":"<p>Deepfake videos are growing progressively more competent because of the rapid advancements in artificial intelligence and deep learning technology. This has led to substantial problems around propaganda, privacy, and security. This research provides an analytically novel method for detecting deepfake videos using temporal discrepancies of the various statistical features of video at the pixel level, followed by a deep learning algorithm. To detect minute aberrations typical of deepfake manipulations, this study focuses on both spatial information inside individual frames and temporal correlations between subsequent frames. This study primarily provides a novel Euclidean distance variation probability score value for directly commenting on the authenticity of a deepfake video. Next, fine-tuning of InceptionResNetV2 with the addition of a dense layer is trained FaceForensics++ for deepfake detection. The proposed fine-tuned model outperforms the existing techniques as its testing accuracy on unseen data outperforms the existing methods. The propsd method achieved an accuracy of 99.80% for FF++ dataset and 97.60% accuracy for CelebDF dataset.</p>","PeriodicalId":15743,"journal":{"name":"Journal of forensic sciences","volume":"70 1","pages":"349-368"},"PeriodicalIF":1.5000,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of forensic sciences","FirstCategoryId":"3","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1111/1556-4029.15665","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MEDICINE, LEGAL","Score":null,"Total":0}
引用次数: 0
Abstract
Deepfake videos are growing progressively more competent because of the rapid advancements in artificial intelligence and deep learning technology. This has led to substantial problems around propaganda, privacy, and security. This research provides an analytically novel method for detecting deepfake videos using temporal discrepancies of the various statistical features of video at the pixel level, followed by a deep learning algorithm. To detect minute aberrations typical of deepfake manipulations, this study focuses on both spatial information inside individual frames and temporal correlations between subsequent frames. This study primarily provides a novel Euclidean distance variation probability score value for directly commenting on the authenticity of a deepfake video. Next, fine-tuning of InceptionResNetV2 with the addition of a dense layer is trained FaceForensics++ for deepfake detection. The proposed fine-tuned model outperforms the existing techniques as its testing accuracy on unseen data outperforms the existing methods. The propsd method achieved an accuracy of 99.80% for FF++ dataset and 97.60% accuracy for CelebDF dataset.
期刊介绍:
The Journal of Forensic Sciences (JFS) is the official publication of the American Academy of Forensic Sciences (AAFS). It is devoted to the publication of original investigations, observations, scholarly inquiries and reviews in various branches of the forensic sciences. These include anthropology, criminalistics, digital and multimedia sciences, engineering and applied sciences, pathology/biology, psychiatry and behavioral science, jurisprudence, odontology, questioned documents, and toxicology. Similar submissions dealing with forensic aspects of other sciences and the social sciences are also accepted, as are submissions dealing with scientifically sound emerging science disciplines. The content and/or views expressed in the JFS are not necessarily those of the AAFS, the JFS Editorial Board, the organizations with which authors are affiliated, or the publisher of JFS. All manuscript submissions are double-blind peer-reviewed.