A multi-touch authentication method based on mobile touch sensor was proposed to solve the information security threats of smart phone. First, the original data sequences of finger sliding collected from a touch sensor were preprocessed through smooth denoising and normalizing the position and length. And then, the authentication feature sequences were extracted from the normalized first and second derivatives, and the direction of the preprocessed data sequences. Finally, the template matching method was adopted to judge the identity of user by using dynamic time warping algorithm to compare the registered temple feature sequences with the test feature sequences. The simulation results show that the proposed algorithm can reach 3.83% False Rejection Rate (FRR) and 2.07% False Acceptance Rate (FAR) in average for different users. Compared with Support Vector Distribution Estimation (SVDE) classifier with Radial Basis Function (RBF), the FRR and FAR of the proposed algorithm reduces 1.81% and 2.35% respectively. The performance analysis shows that the proposed algorithm can improve the accuracy of user authentication obviously.