CLaRE 🕵️‍♀️

This project introduces CLaRE, a deepfake detection method that builds on CLIP by integrating Latent Reconstruction Error (LaRE) [1] with Context Optimization (CoOp) [2] or Conditional Context Optimization (CoCoOp) [3] for adaptive prompt learning. While CLIP with learnable prompts offers strong generalization, it struggles with realistic forgeries from modern generative tools. CLaRE addresses this gap and is evaluated on the DF40 dataset across multiple training regimes. It shows competitive performance on GAN-generated images and surpasses prompt-tuned CLIP on diffusion-based fakes.

The code for our study is available at: link

Note: The poster does not represent the final camera-ready version.

@inproceedings{10.1145/3733813.3764366,
author = {Thakur, Udit and Khan, Mohammad Hafeez and Changlani, Meher and Dotsinski, Asen and Mahadevan, Aswin Krishna and Kechagias, Ioannis and Najdenkoska, Ivona},
title = {CLaRE: CLIP with Latent Reconstruction Errors for Generated Face Detection},
year = {2025},
isbn = {9798400719028},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3733813.3764366},
doi = {10.1145/3733813.3764366},
booktitle = {Proceedings of the 1st ACM Workshop on Deepfake, Deception, and Disinformation Security},
pages = {5–15},
numpages = {11},
keywords = {Deepfake detection, Diffusion models, CLIP, Reconstruction error, Prompt tuning},
series = {3D-Sec '25}
}

References

1. [1] Y. Luo, J. Du, K. Yan, and S. Ding, “LaRE^2: Latent Reconstruction Error Based Method for Diffusion-Generated Image Detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 17006–17015.
2. [2] K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Learning to Prompt for Vision-Language Models,” International Journal of Computer Vision, vol. 130, no. 9, pp. 2337–2348, Jul. 2022, doi: 10.1007/s11263-022-01653-1.
3. [3] K. Zhou, J. Yang, C. C. Loy, and Z. Liu, “Conditional prompt learning for vision-language models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16816–16825.