Secure multi-party computation is a focus of international cryptography in recent years. Protocols for Yao's millionaires' problem have become an important building block of many secure multi-party computation protocols. Their efficiency are crucial to the efficiency of many secure multi-party computation protocols. Unfortunately, known protocols for Yao's millionaires' problem have high computational complexity or communication complexity. In this study, based on the 1-out-of-m oblivious transfer and set-inclusion problem, we propose a new protocol to solve this problem. This new protocol is very efficient in terms of both computational and communication complexities. Its privacy-preserving property is also proved by simulation paradigm which is generally accepted in the study of secure multi-party computation. We also compare the information leakage of our new protocol and the known protocols.