A Review: Unveiling the Unreal: Deepfake Face Detection Using CNN, BiLSTM and Attention Mechanisms.

Deepfake technology has rapidly advanced with the development of Artificial Intelligence and Deep Learning techniques, creating serious challenges related to misinformation, digital security, privacy, and cybercrime. This review paper analyzed various AI-based deepfake face detection approaches using CNN, BiLSTM, Attention Mechanisms, Transformer models, and hybrid deep learning frameworks. The study observed that CNN models are highly effective for spatial feature extraction, while BiLSTM networks improve temporal sequence analysis in manipulated videos. Attention mechanisms further enhance detection performance by focusing on important manipulated facial regions. The literature review highlighted that hybrid architectures combining CNN, BiLSTM, and Attention techniques provide better accuracy and robustness compared to traditional deepfake detection systems. However, several challenges still exist, including poor cross-dataset generalization, high computational complexity, lack of explainability, and limited real-time deployment capability. The study concludes that integrating Dense-Switch CNN, BiLSTM, and Attention Mechanisms can significantly improve deepfake detection performance and support the development of more reliable and scalable detection systems. Future research should focus on developing lightweight, explainable, and real-time deepfake detection frameworks capable of handling diverse real-world manipulated media environments.

[1]              A. Alhudhaif et al., “Deepfake Face Detection Using Dense-Switch Convolutional Neural Network, BiLSTM, and Attention Mechanism,” IEEE Access, vol. 13, 2025, doi: 10.1109/ACCESS.2025.3570761. [2]              S. Khan et al., “Deepfake Detection: Analysing Model Generalisation Across Architectures, Datasets and Pre-Training Paradigms,” 2024. [3]               H. Omar et al., “Deepfake Video Detection Using Ensemble CNN and Self-Attention Mechanisms,” Neural Computing and Applications, vol. 36, 2024. [4]               M. Masud et al., “LW-DeepFakeNet: Lightweight CNN-LSTM Framework for Deepfake Video Detection,” Signal, Image and Video Processing, vol. 17, 2023. [5]               S. Tipper et al., “Hybrid CNN-LSTM Architectures for Deepfake Video Detection,” Applied Sciences, vol. 14, no. 2, 2024. [6]               M. Al-Adwan et al., “Hybrid CNN-RNN Deepfake Detection Using Particle Swarm Optimization,” Computers, vol. 13, no. 4, 2024. [7]               A. Al-Dulaimi et al., “Transfer Learning-Based CNN-LSTM Framework for Deepfake Image Detection,” Electronics, vol. 13, no. 9, 2024. [8]               V. Thing, “A Comparative Study of CNN and Transformer-Based Deepfake Detection Systems,” 2023. [9]               M. Hashmi et al., “AVTENet: Audio-Visual Transformer Ensemble Network for Deepfake Video Detection,” 2023. [10]            Y. Li et al., “Multi-Attentional Deepfake Detection Network for Face Manipulation Detection,” IEEE Transactions on Information Forensics and Security, vol. 18, 2023. [11]            S. Dasgupta et al., “Lightweight CNN with SE Block Attention for Efficient Deepfake Detection,” 2025.