The homomorphic encryption algorithm Paillier is a cryptographic method that can be applied in areas such as privacy-preserving computation and federated learning. However, due to the inherent complexity and computational overhead of the algorithm, it may affect real-time requirements when dealing with a large volume of encrypted data. In this paper, we propose a method to improve the efficiency of the encryption algorithm by combining GPU parallel processing technology with homomorphic encryption. Through experiments on encrypting and decrypting plaintexts of different sizes, the results show that the introduction of GPU significantly improves the encryption and decryption speed. Compared with other parallel processing methods, the acceleration effect of our proposed method is more significant.