With the development of smart city, user data are all transformed in encrypted form to protect user privacy. Secure comparison in encrypted form is the fundamental operation of many secure encrypted data analysis tasks, such as secure k-NN query and classification, Bayesian classification. Thus, it is important to achieve an efficient secure encrypted data comparison scheme. Recently, some methods have been put forward to support secure comparison over encrypted data. Nevertheless, the existing solutions are still inefficient in practical. In this paper, we propose a novel encrypted data comparison protocol based on a hybrid approach of Paillier cryptosystem and garbled circuits. Our scheme reveals nothing about encrypted data and comparison result, and is provably secure under semi-honest model. Additionally, our proposed protocol can achieve higher efficiency, compared with the state-of-the-art scheme. Finally, we indicate the security and efficiency of our scheme by theoretical analysis and experiment evaluations.