The popularity of the worldwide web and the easy availability of the digital content has brought the security issues to the forefront. NURBS is widely used in computer-aided geometry design and computer graphics for its strong representative properties. In this paper, we present a blind watermarking algorithm for NURBS surfaces. First the points sampled from the NURBS surface are watermarked in DCT domain, then the watermarked NURBS surface is obtained by fitting the watermarked points iteratively. In watermark detection stage, the sign correlation detector is used, which is a blind detector and accordingly the original NURBS surface is not required for detection. The experimental results show that the algorithm preserves the shape of the NURBS surface. The proposed algorithm is robust against attacks such as knot insertion, knot removal, reparameterization, order elevation, order reduction, additive white Gaussian noise, rotation, translation, scaling, multi-watermark attacks, etc.