Review of Deepfake Detection Techniques and Challenges
The proliferation of deepfake technology, powered by advanced generative models such as generative adversarial networks (GANs), presents challenges in digital media authenticity, public trust, and cybersecurity. We reviewed recent advancements in deepfake detection across multiple modalities, includ...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Proceeding |
| Language: | en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/50728/1/%282025%29%20Review%20of%20Deepfake%20Detection%20Techniques%20and%20Challenges.pdf http://ir.unimas.my/id/eprint/50728/ https://ieeexplore.ieee.org/document/11255645 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The proliferation of deepfake technology, powered by advanced generative models such as generative adversarial networks (GANs), presents challenges in digital media authenticity, public trust, and cybersecurity. We reviewed recent advancements in deepfake detection across multiple modalities, including image, video, and audio. Benchmark datasets, such as FaceForensics++, the deepfake detection challenge (DFDC), and Celeb-deepfake (Celeb-DF), have been used to develop diverse detection models. These models encompass approaches based on EfficientNet-driven transfer learning, convolutional neural network–long shortterm memory (CNN-LSTM) hybrids for temporal feature extraction, graph-based neural architectures, and ensemble methods that integrate deep learning with handcrafted features. Although certain models report detection accuracies as high as 99.99% on specific datasets, many exhibit limited generalizability across different benchmarks, particularly when confronted with compression artifacts. Additionally, real-time deployment remains constrained by substantial computational
demands. Emerging threats, including adversarial perturbations and diffusion-based synthetic media, necessitate the development of more resilient detection strategies. Proactive countermeasures such as blockchain-based timestamping, digital watermarking, and cryptographic hashing have been adopted to enhance media integrity. The results of the review underscore the need for lightweight, interpretable, and multimodal detection frameworks to generalize the models’ applicability across diverse domains, thereby supporting reliable and scalable media verification in increasingly complex digital environments. |
|---|