Deepfake Detection in AV1 Compressed Videos with EfficientNet and Stacked Bi-LSTM Model

Ghadge, Aniket Suryakant (2024) Deepfake Detection in AV1 Compressed Videos with EfficientNet and Stacked Bi-LSTM Model. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

The rise of Deepfake video content online jeopardises the integrity of digital media as they are optimised to spread disinformation, sow doubt, and facilitate confusion. The following research investigates the effects of AV1 lossy compression on deepfake video detection, highlighting the need for models that adapt to different compression levels. To capture temporal inconsistencies in video frames, the author proposed an ensemble model that combines a three-layered bidirectional LSTM network with EfficientNet-B0 for feature extraction. This experiment tested the model on raw and AV1-compressed videos at 250 kbps and 1024 kbps bitrates using the FaceForensics++ dataset. The key experimental findings show that our model achieves over 90% accuracy in all formats, with the best performance in terms of accuracy, recall, and fewer false positives and negatives being seen in high-bitrate AV1 videos. On the other hand, low-bitrate compression adds complexity by hiding fake artefacts, which degrades model performance with a higher number of false positives. This study highlights the challenges and importance of adapting deepfake detection models to handle various levels of lossy compression effectively.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Hamill, David UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning Z Bibliography. Library Science. Information Resources > ZA Information resources > ZA4150 Computer Network Resources > The Internet > World Wide Web > Websites > Online social networks T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Telecommunications > The Internet > World Wide Web > Websites > Online social networks
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	18 Aug 2025 15:01
Last Modified:	18 Aug 2025 15:01
URI:	https://norma.ncirl.ie/id/eprint/8567

Actions (login required)

View Item