Deep Deception Detector: Exposing AI Generated Fake Video
Bibliographic record
Abstract
Deepfakes are artificially generated videos or images created to falsely portray someone as saying or doing things they never actually did. These manipulated media forms can lead to serious issues, especially when circulated on social media platforms. As a result, detecting deepfakes has become increasingly critical. This project builds on an existing detection method called FAMM, which targets identifying deepfakes, particularly in videos that have been compressed. The original FAMM approach analyzes facial movements by calculating distances and angles between key facial landmarks, such as the eyes, nose, and mouth, and observing how these points change over time. In this earlier method, GRU and SVM models were used to capture the temporal and static changes, and their outputs were combined to determine whether the video was authentic or fake. In our updated approach, we introduce more advanced techniques. We replace the GRU with a Transformer model, which offers improved capabilities in capturing time-based changes in facial movements. Additionally, we implement EfficientNet to extract more precise features from the face images. The data from both models are processed and then combined through a fusion strategy to reach a final classification of whether the video is real or fake. With these advancements, our system demonstrates improved accuracy in detecting deepfakes, even in low-quality or compressed videos. This project highlights how cutting-edge deep learning techniques can better address the spread of deepfakes on social media platforms.
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 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 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".