AE-LSTM: A Hybrid Approach for Detecting Deepfake Videos in Digital Forensics

AE-LSTM: A Hybrid Approach for Detecting Deepfake Videos in Digital Forensics

연구 분야: Strategies

학회: International Conference on Recent Developments in Cyber Security

Deepfakes can have serious implications for security, privacy, and trust, as deepfake can be utilized for the purpose of spreading misinformation, fake news, and propaganda. Deepfakes which are created through deep-learning techniques have become threatful in recent times and pose a significant challenge to digital forensics. As a result, deepfake video detection is a significant area of research in digital forensics. In this paper, we proposed an autoencoder-LSTM-based solution for the detection of deepfake videos, in this method autoencoder helps to obtain a robust solution. The proposed method gives an accuracy of 81.73 on the Celeb-df dataset.