Chromatic Camouflage Benchmarking for Vision-Language Models Under Adversarial Visual Conditions

Authors

  • Marek Herrera Faculty of Engineering, Architecture and Information Technology, University of Queensland, Brisbane, Australia
  • Eunwoo Yeo Faculty of Engineering, Architecture and Information Technology, University of Queensland, Brisbane, Australia
  • Agatha Bruno Faculty of Engineering, Architecture and Information Technology, University of Queensland, Brisbane, Australia

Keywords:

Multimodal Learning, Adversarial Robustness, Chromatic Camouflage, Visual Perception, Benchmarking

Abstract

The rapid proliferation of vision-language models has fundamentally transformed artificial intelligence, enabling unprecedented capabilities in multimodal understanding, visual reasoning, and cross-domain zero-shot generalization. However, the robustness of these models under complex adversarial conditions remains a critical vulnerability, particularly when subjected to chromatic camouflage phenomena. Chromatic camouflage, wherein object boundaries and textures blend seamlessly with background environments due to subtle color shifts and isoluminant manipulations, poses severe challenges for existing visual encoders. This paper introduces a comprehensive benchmarking framework designed to evaluate vision-language models under synthesized and naturally occurring adversarial chromatic conditions. We systematically construct a massive dataset of chromatically camouflaged instances, spanning diverse environments and object categories, to rigorously test the perceptual limits of contemporary multimodal architectures. Through extensive empirical evaluation, we demonstrate that state-of-the-art vision-language models exhibit catastrophic performance degradation when exposed to even moderate levels of chromatic perturbation. Our findings indicate that these models disproportionately rely on high-frequency textural cues and stark color contrast, failing to leverage semantic contextual reasoning when low-level visual signals are obfuscated. Furthermore, we provide a detailed analysis of the architectural bottlenecks contributing to these vulnerabilities, offering novel insights into the cognitive disconnect between textual representation and visual perception. This benchmark serves as a foundational step toward developing more resilient multimodal systems capable of maintaining robust performance in visually deceptive and unconstrained real-world environments

References

1. Liu, Y., & Kwon, H. (2025). Efficient Depth Estimation for Unstable Stereo Camera Systems on AR Glasses. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 6252-6261).

2. Tu, P., Xie, X., Ai, G., Li, Y., Huang, Y., & Zheng, Y. (2023). FemtoDet: An object detection baseline for energy versus performance tradeoffs. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 13318-13327).

3. Dong, J., Zhang, C., Qu, X., Ma, Z., Koniusz, P., & Ong, Y. S. (2026). Robust superalignment: Weak-to-strong robustness generalization for vision-language models. Advances in Neural Information Processing Systems, 38, 18345-18377.

4. Zhao, H., Wang, Q., Zhan, G., Min, W., Zou, Y., & Cui, S. (2022). Need only one more point (NOOMP): Perspective adaptation crowd counting in complex scenes. IEEE Transactions on Multimedia, 25, 1414-1426.

5. Zhang, W. (2026). A 5-6 GHz PVT Robust Current Mode Passive Mixer for Direct Down-Conversion Receiver.

6. Tu, P., Huang, Y., Zheng, F., He, Z., Cao, L., & Shao, L. (2022, June). Guidedmix-net: Semi-supervised semantic segmentation by using labeled images as reference. In Proceedings of the AAAI conference on artificial intelligence (Vol. 36, No. 2, pp. 2379-2387).

7. Dong, J., Koniusz, P., Feng, L., Zhang, Y., Zhu, H., Liu, W., ... & Ong, Y. S. (2025). Robustifying zero-shot vision language models by subspaces alignment. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 21037-21047).

8. Zhang, W., Zhang, C., Gu, C., Kou, J., Yuan, H., Fang, X., ... & Fang, Y. (2024, October). Hallucination in Large Language Models: From Mechanistic Understanding to Novel Control Frameworks. In 2024 7th International Conference on Universal Village (UV) (pp. 1-36). IEEE.

9. Dong, J., Koniusz, P., Zhang, Y., Zhu, H., Liu, W., Qu, X., & Ong, Y. S. (2025, October). Improving zero-shot adversarial robustness in vision-language models by closed-form alignment of adversarial path simplices. In Forty-second International Conference on Machine Learning.

10. Kerbl, B., Kopanas, G., Leimkühler, T., & Drettakis, G. (2023). 3D Gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4), Article 139.

11. Zhang, J., Shi, Y., Ma, Y., Xu, L., Yu, J., & Wang, J. (2023, June). Ikol: Inverse kinematics optimization layer for 3d human pose and shape estimation via gauss-newton differentiation. In Proceedings of the AAAI conference on artificial intelligence (Vol. 37, No. 3, pp. 3454-3462).

12. Qi, Z., Yuan, Y., Ruan, X., Wang, S., Zhang, W., & Huang, Q. (2024, March). Bias-conflict sample synthesis and adversarial removal debias strategy for temporal sentence grounding in video. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 38, No. 5, pp. 4533-4541).

13. Guo, Zixin, Kai Zhao, and Luyan Zhang. "InstanceRSR: Real-World Super-Resolution via Instance-Aware Representation Alignment." ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2026.

14. Sun, W., Dong, X. M., Cui, B., & Tang, J. (2025, April). Attentive eraser: Unleashing diffusion model’s object removal potential via self-attention redirection guidance. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 19, pp. 20734-20742).

15. Lv, Q., Deng, X., Chen, G., Wang, M. Y., & Nie, L. (2024). Decision mamba: A multi-grained state space model with self-evolution regularization for offline rl. Advances in neural information processing systems, 37, 22827-22849.

16. Song, S., Tang, Y., & Qin, R. (2025). Synthetic Data Matters: Re-training with Geo-typical Synthetic Labels for Building Detection. IEEE Transactions on Geoscience and Remote Sensing.

17. Guo, Hanzhong, et al. "Real-time identity defenses against malicious personalization of diffusion models." arXiv preprint arXiv:2412.09844 (2024).

18. Peng, Q., Zheng, C., & Chen, C. (2023). Source-free domain adaptive human pose estimation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (pp. 4826-4836).

19. Qu, W., Wang, J., Gong, Y., Huang, X., & Xiao, L. (2025). An end-to-end r