3D models unauthorized modification is an attractive research problem nowadays due to the widespread availability of technologies and designs on the internet. In this paper, we propose a blind fragile watermarking scheme in the spatial domain for 3D model authentication based on the curvature features of the 3D model and chaos sequence. First, we compute the curvature feature for every vertex of the 3D model, after that, the vertices are classified into 3 classes; peak, flat, and middle vertices using K-means clustering algorithm. We suggested two methods for embedding the watermark based on the least significant bit (LSB) substitution technique; Cluster Type-based Embedding (CTE) and Cluster Size-based Embedding (CSE). The proposed methods employs a Chaos sequence generator to generate a Chaos sequence that is used to generate the embedded watermark, where the tampering region can be verified and located by the Chaos sequencebased watermark check. Many assessment methods are employed to evaluate the proposed method with various unauthorized attacks like rotation, translation, scaling, cropping, and noise addition. The experiment results show an improvement of embedding imperceptibility as well as tempered regions detection compared to existing literature works.