Detecting Deepfakes: A Novel Framework Employing XceptionNet-Based Convolutional Neural Networks
Bibliographic record
Abstract
Social networking sites have become primary sources of information for web users, making the rapid dissemination of deepfakes a cause for concern. Deepfakes are digitally manipulated images or videos that contain the computer-generated face of another person. Advancements in hardware and computational technologies have made the creation of deepfakes increasingly accessible, even to individuals without technical expertise. The potential harm posed by deepfakes necessitates urgent efforts to improve the detection of these manipulated media. Deep learning (DL) models have experienced rapid growth, enabling the synthesis and generation of hyper-realistic videos, often referred to as "deepfakes." DL algorithms can now create faces, swap faces between 2 individuals in video, and modify facial expressions, gender, also other features. These video manipulation techniques have applications in numerous fields, but deepfakes specifically exploit DL to synthesize and alter images in a manner that makes it difficult to discern between fake and genuine media. In this study, we present novel deepfake detection framework using DL and pre-trained XceptionNet model depends upon deep CNNs (Convolutional Neural Networks). We employ facial landmark recognition to extract information related to several facial characteristics from videos. This data is then used to facilitate the deep learning model's differentiation between genuine and deepfake videos. Features extracted from videos are utilized to train CNN concurrently. Our deepfake detection system is built on a multi-input Xception Neural Network model, which leverages CNNs. The system is trained using the Dessa Dataset and subset of Deepfake Detection Challenge Dataset. Proposed model demonstrates strong performance, achieving 96% classification accuracy and an AUC of 0.97, offering a promising solution for detecting deepfake videos.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".