For sensitive images, e.g., military and medical images, traditional digital image watermarking methods are insufficient to realize effective protection. The original image is distorted due to traditional digital image watermarking methods embedding a watermark into it. We suggest a method for generating a zero-watermark image that is robust and uses discrete wavelet transform and singular value decomposition techniques. The Arnold transform is used in the zero-watermark generation and verification phases to enhance the security of the proposed method. The feature image of the original image is extracted after normalization and then frequency transformed. Various attacks were evaluated to determine the robustness of the proposed method, and the experimental results indicate that it exhibits good robustness against most attacks.