Enhancing Deepfake Detection with Explainable AI Through Neural-Symbolic RCNN-BiGRU Approach

Zainab Ali Abbood; Raghad Tariq Al-Hassan; Mahmoud Shuker Mahmoud; Atheel Sabih Shaker

doi:10.58564/IJSER.5.2.2026.369

Authors

Zainab Ali Abbood Computer Technology Engineering Department, Al-Mansour University College, Baghdad, Iraq
Raghad Tariq Al-Hassan Ministry of Higher Education and Scientific Research, Minister’s Office, Baghdad, Iraq
Mahmoud Shuker Mahmoud Computer Engineering - Computer Networks, Gilgamesh University, Baghdad, Iraq , and Cybersecurity Technology Engineering Department, Middle Technical University, Electrical Engineering Technical College, Baghdad, Iraq
Atheel Sabih Shaker Dept. of Computer Engineering Techniques, College of Technology Engineering Al-Iraqia Science University, Baghdad, Iraq

DOI:

https://doi.org/10.58564/IJSER.5.2.2026.369

Keywords:

Deepfake detection, Explainable Artificial Intelligence (XAI), Neural-symbolic reasoning, Deep learning, Multimedia

Abstract

The rapid advancement of deepfake generation techniques poses a significant threat to digital media authenticity, necessitating detection systems that are not only accurate but also explainable and robust across diverse content types. While deep learning models such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) have shown promising performance in detecting spatial and temporal inconsistencies, they often operate as black-box systems with limited interpretability and generalization capability. To address these challenges, this paper proposes a neural-symbolic deepfake detection framework that integrates Explainable Artificial Intelligence (XAI) with hybrid deep learning models. The proposed approach combines Region-based Convolutional Neural Networks (RCNN) and Bidirectional Gated Recurrent Units (Bi-GRU) for effective spatiotemporal feature extraction. These features are further processed through a propositional inference layer that incorporates symbolic reasoning based on logical rules reflecting natural facial behavior, including eye movement, lip synchronization, and facial consistency. The model is evaluated on benchmark datasets, including Celeb-DF, Deepfake TIMIT, and WLDR, demonstrating superior performance compared to baseline methods in terms of F1-score, Area Under the Curve (AUC), and Matthews Correlation Coefficient (MCC), along with improved true positive rates in ROC analysis. Furthermore, ablation studies confirm that the integration of symbolic reasoning enhances detection performance by enforcing logical consistency and providing interpretable decision-making. Overall, the results highlight the effectiveness of neural-symbolic reasoning as a robust and transparent framework for deepfake detection, contributing to the advancement of explainable and trustworthy AI systems in multimedia forensics.

References

[1] A. Raza, K. Munir, and M. Almutairi, “A novel deep learning approach for deepfake image detection,” Applied Sciences, vol. 12, no. 19, p. 9820, 2022. DOI: https://doi.org/10.3390/app12199820

[2] S. H. Silva, M. Bethany, A. M. Votto, I. H. Scarff, N. Beebe, and P. Najafirad, “Deepfake forensics analysis: An explainable hierarchical ensemble of weakly supervised models,” Forensic Science International: Synergy, vol. 4, p. 100217, 2022. DOI: https://doi.org/10.1016/j.fsisyn.2022.100217

[3] V.-N. Tran, S.-G. Kwon, S.-H. Lee, H.-S. Le, and K.-R. Kwon, “Generalization of forgery detection with meta deepfake detection model,” IEEE Access, vol. 11, pp. 535–546, 2022. DOI: https://doi.org/10.1109/ACCESS.2022.3232290

[4] J. Kang, S.-K. Ji, S. Lee, D. Jang, and J.-U. Hou, “Detection enhancement for various deepfake types based on residual noise and manipulation traces,” IEEE Access, vol. 10, pp. 69031–69040, 2022. DOI: https://doi.org/10.1109/ACCESS.2022.3185121

[5] R. Tolosana, S. Romero-Tapiador, R. Vera-Rodriguez, E. Gonzalez-Sosa, and J. Fierrez, “Deepfakes detection across generations: Analysis of facial regions, fusion, and performance evaluation,” Engineering Applications of Artificial Intelligence, vol. 110, p. 104673, 2022. DOI: https://doi.org/10.1016/j.engappai.2022.104673

[6] A. Ismail, M. Elpeltagy, M. S. Zaki, and K. Eldahshan, “An integrated spatiotemporal-based methodology for deepfake detection,” Neural Computing and Applications, vol. 34, no. 24, pp. 21777–21791, 2022. DOI: https://doi.org/10.1007/s00521-022-07633-3

[7] F. Dong, X. Zou, J. Wang, X. Liu, and Y. Cao, “Contrastive learning-based general deepfake detection with multi-scale RGB frequency clues,” Journal of King Saud University - Computer and Information Sciences, vol. 35, no. 4, pp. 90–99, 2023. DOI: https://doi.org/10.1016/j.jksuci.2023.03.005

[8] N. M. Alnaim et al., “DFFMD: A deepfake face mask dataset for infectious disease era with deepfake detection algorithms,” IEEE Access, vol. 11, pp. 16711–16722, 2023. DOI: https://doi.org/10.1109/ACCESS.2023.3246661

[9] Y. Patel et al., “An improved dense CNN architecture for deepfake image detection,” IEEE Access, vol. 11, pp. 22081–22095, 2023. DOI: https://doi.org/10.1109/ACCESS.2023.3251417

[10] A. Elhassan et al., “DFT-MF: Enhanced deepfake detection using mouth movement and transfer learning,” SoftwareX, vol. 19, p. 101115, 2022. DOI: https://doi.org/10.1016/j.softx.2022.101115

[11] S. Kolagati, T. Priyadharshini, and V. M. A. Rajam, “Exposing deepfakes using a deep multilayer perceptron–convolutional neural network model,” International Journal of Information Management Data Insights, vol. 2, no. 1, p. 100054, 2022. DOI: https://doi.org/10.1016/j.jjimei.2021.100054

[12] Y. Li, C. Zhang, H. Qi, and S. Lyu, “ADANI: Adaptive noise injection to improve adversarial robustness,” Computer Vision and Image Understanding, vol. 238, p. 103855, 2024. DOI: https://doi.org/10.1016/j.cviu.2023.103855

[13] Y. Wang, Q. Sun, D. Rong, and R. Geng, “Multi-domain awareness for compressed deepfake videos detection over social networks guided by common mechanisms between artifacts,” Computer Vision and Image Understanding, vol. 247, p. 104072, 2024. DOI: https://doi.org/10.1016/j.cviu.2024.104072

[14] J. Koo and D. Klabjan, “Improved classification based on deep belief networks,” in Proc. ICANN 2020, Springer, pp. 541–552. DOI: https://doi.org/10.1007/978-3-030-61609-0_43

[15] Gautam, V., Kaur, G., Malik, M., Pawar, A., Singh, A., Singh, K. K., ... & Abouhawwash, M. (2024). FFDL: feature fusion-based deep learning method utilizing federated learning for forged face detection. IEEE Access, 13, 5366-5379. DOI: https://doi.org/10.1109/ACCESS.2024.3523257

[16] Alrawahneh, A. A. M., Abdullah, S. N. A. S., Abdullah, S. N. H. S., Kamarudin, N. H., & Taylor, S. K. (2025). Video authentication detection using deep learning: a systematic literature review: A. AM. et al. Applied Intelligence, 55(4), 239. DOI: https://doi.org/10.1007/s10489-024-05997-8

[17] Shanmuganathan, C., Thamizharasi, M., Anish, T. P., & Sivasankari, K. (2024). Enhancing deepfake detection: Leveraging mesonet for video fraud identification. SN Computer Science, 5(3), 301. DOI: https://doi.org/10.1007/s42979-024-02629-3

[18] Rajeev, A., & Raviraj, P. (2024). Performance evaluation of deep learning models for detecting deep fakes. International Journal of Systematic Innovation, 8(1), 49-62.