This paper introduces a novel cross-modality perturbation synergy attack (CMPS) for person re-identification (ReID) systems. The main challenge in cross-modality ReID is effectively handling visual differences between different modalities, such as infrared and visible images. Existing attack methods focus on visible image characteristics, neglecting other modalities and their data distribution variations, which can undermine their effectiveness in cross-modal retrieval. The CMPS attack optimizes perturbations by leveraging gradients from diverse modality data, disrupting the discriminator and reinforcing modality differences. Experiments on RegDB and SYSU datasets demonstrate the effectiveness of the proposed method, showing that it can significantly reduce model accuracy in cross-modality ReID tasks. The CMPS attack is designed to capture shared features across modalities by using grayscale transformations and gradient information from multiple modalities. The method is evaluated against state-of-the-art approaches, showing superior performance in terms of attack effectiveness and transferability across different models. The results indicate that the CMPS attack is more effective in cross-modality scenarios compared to traditional single-modality attacks. The study contributes to the understanding of cross-modality ReID security and provides insights for improving the robustness of cross-modality ReID systems.This paper introduces a novel cross-modality perturbation synergy attack (CMPS) for person re-identification (ReID) systems. The main challenge in cross-modality ReID is effectively handling visual differences between different modalities, such as infrared and visible images. Existing attack methods focus on visible image characteristics, neglecting other modalities and their data distribution variations, which can undermine their effectiveness in cross-modal retrieval. The CMPS attack optimizes perturbations by leveraging gradients from diverse modality data, disrupting the discriminator and reinforcing modality differences. Experiments on RegDB and SYSU datasets demonstrate the effectiveness of the proposed method, showing that it can significantly reduce model accuracy in cross-modality ReID tasks. The CMPS attack is designed to capture shared features across modalities by using grayscale transformations and gradient information from multiple modalities. The method is evaluated against state-of-the-art approaches, showing superior performance in terms of attack effectiveness and transferability across different models. The results indicate that the CMPS attack is more effective in cross-modality scenarios compared to traditional single-modality attacks. The study contributes to the understanding of cross-modality ReID security and provides insights for improving the robustness of cross-modality ReID systems.