Analysis of the Effect of Social Media Compression on the Accuracy of Deepfake Detection Using MobileNetV3 and Compressed Data Augmentation

Erika Ramadhani; M. Fahrul Ramadhan

doi:10.29040/ijcis.v7i3.279

Analysis of the Effect of Social Media Compression on the Accuracy of Deepfake Detection Using MobileNetV3 and Compressed Data Augmentation

Erika Ramadhani, M. Fahrul Ramadhan

Abstract

Deepfakes are synthetic audio-visual content that is artificially engineered that is increasingly difficult to distinguish from real content and is mostly circulated through social media platforms. The problem studied in this study is the decline in the accuracy of the deepfake detection model when the test content has gone through a series of social media compression processes, such as H.264 re-encoding and resolution degradation, which removes some of the forensic traces of the pixels that are the basis for detection. The objectives of this study were to measure the magnitude of the decline in detection accuracy in the lightweight architecture of MobileNetV3, to find the compression threshold where the decline began to be significant, and to test the augmentation of compressed data as a mitigation strategy. The method used was a quantitative experiment on the Celeb-DF v2 dataset with three levels of compression (raw, c23, and c40) simulated using FFmpeg; The MobileNetV3-Small model was trained on raw data and then tested at all three levels of compression, then compared to training scenarios using combined data of all three levels of compression. Test results on 5,180 face images of standard test data showed a decrease in accuracy from 70.23% in raw data to 69.54% in c23 and 58.11% in c40, with the sharpest decrease occurring between medium compression and high compression and mainly due to a drop in recall from 89.79% to 61.24%. Training scenarios with compressed data augmentation were shown to increase AUC at all levels of compression, from 59.51% to 64.29% at c40, although accompanied by a slight decrease in accuracy on raw data. The contribution of this study is empirical evidence of the compression resistance of lightweight architectures that are rarely studied as well as practical mitigation recommendations for the development of deepfake detection systems on mobile devices.

Full Text:

PDF

References

R. Ramanaharan, D. B. Guruge, and J. I. Agbinya, “DeepFake video detection: Insights into model generalisation — A Systematic review,” Data Inf. Manag., vol. 9, no. 4, p. 100099, Dec. 2025, doi: 10.1016/j.dim.2025.100099.

A. K. Singh and R. K. Banyal, “A Comprehensive Review of DL-Based Methods for DeepFake Image, Audio, and Video Detection,” Int. J. Image Graph., Mar. 2026, doi: 10.1142/S0219467827501002.

A. Montibeller, D. Shullani, D. Baracchi, A. Piva, and G. Boato, “Bridging the Gap: A Framework for Real-World Video Deepfake Detection via Social Network Compression Emulation,” in Proceedings of the 1st on Deepfake Forensics Workshop: Detection, Attribution, Recognition, and Adversarial Challenges in the Era of AI-Generated Media, New York, NY, USA: ACM, Oct. 2025, pp. 29–36. doi: 10.1145/3746265.3759670.

S. A. Khan and D.-T. Dang-Nguyen, “Deepfake Detection: Analyzing Model Generalization Across Architectures, Datasets, and Pre-Training Paradigms,” IEEE Access, vol. 12, pp. 1880–1908, 2024, doi: 10.1109/ACCESS.2023.3348450.

A. Howard et al., “Searching for MobileNetV3,” in 2019 IEEE/CVF International Conference on Computer Vision (ICCV), IEEE, Oct. 2019, pp. 1314–1324. doi: 10.1109/ICCV.2019.00140.

A. M. Lagsoun et al., “Compressed deepfake detection via GA-LASSO selection of deep features and machine learning models,” Sci. Rep., vol. 16, no. 1, p. 19708, Mar. 2026, doi: 10.1038/s41598-025-34733-6.

C.-Y. Lai et al., “UMCL: Unimodal-generated Multimodal Contrastive Learning for Cross-compression-rate Deepfake Detection,” Int. J. Comput. Vis., vol. 134, no. 1, p. 40, Jan. 2026, doi: 10.1007/s11263-025-02606-0.

C. Tan, Y. Zhao, S. Wei, G. Gu, P. Liu, and Y. Wei, “Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 5, pp. 5052–5060, Mar. 2024, doi: 10.1609/aaai.v38i5.28310.

M. Qiao, R. Tian, Y. Zhai, and C. Li, “SFCL: A Spatial-Frequency Collaborative Learning Framework for Generalizable Deepfake Detection,” in 2025 IEEE International Joint Conference on Biometrics (IJCB), IEEE, Sep. 2025, pp. 1–10. doi: 10.1109/IJCB65343.2025.11410680.

D. Nayak, B. S. Prasad Mishra, and S. Champati Rai, “Lightweight and Efficient Deepfake Detection Using Transfer-Learned MobileNetV2,” in 2025 2nd International Conference on Circuits, Power and Intelligent Systems (CCPIS), IEEE, Sep. 2025, pp. 1–6. doi: 10.1109/CCPIS65231.2025.11234268.

T. Harshitha, T. Simhadri, T. Jaswanthi, N. Rayvanth, and R. P. Singh, “Deepfake Image Detection Using Light-Weight Attention Integrated MobileNetV3 Model,” 2025, pp. 130–142. doi: 10.1007/978-3-031-84543-7_11.

A. Kumar, “HybridNet: Advancing Deepfake Detection Through Residual, SE, and Depthwise Convolutions,” IEEE Access, vol. 13, pp. 207541–207552, 2025, doi: 10.1109/ACCESS.2025.3640106.

Z. Xia, T. Qiao, M. Xu, X. Wu, L. Han, and Y. Chen, “Deepfake Video Detection Based on MesoNet with Preprocessing Module,” Symmetry (Basel)., vol. 14, no. 5, p. 939, May 2022, doi: 10.3390/sym14050939.

Y. Li, X. Yang, P. Sun, H. Qi, and S. Lyu, “Celeb-DF (v2): a new dataset for DeepFake Forensics,” arXiv preprint arXiv:1909.12962, 2019, [Online]. Available: https://www.researchgate.net/profile/Yuezun-Li/publication/336147158_Celeb-DF_A_New_Dataset_for_DeepFake_Forensics/links/5e1bec72a6fdcc28376e4548/Celeb-DF-A-New-Dataset-for-DeepFake-Forensics.pdf

A. Rossler, D. Cozzolino, L. Verdoliva, and ..., “Faceforensics++: Learning to detect manipulated facial images,” Proceedings of the …, 2019, [Online]. Available: http://openaccess.thecvf.com/content_ICCV_2019/html/Rossler_FaceForensics_Learning_to_Detect_Manipulated_Facial_Images_ICCV_2019_paper.html

S. D, R. S, S. Ravi, V. M, and P. M. P. U, “A Lightweight CNN for Efficient Deepfake Detection of Low-resolution Images in Frequency Domain,” in 2024 Second International Conference on Emerging Trends in Information Technology and Engineering (ICETITE), IEEE, Feb. 2024, pp. 1–6. doi: 10.1109/ic-ETITE58242.2024.10493406.

DOI: https://doi.org/10.29040/ijcis.v7i3.279