The rapid development of the Internet, IOT, and big data brings great chance to share data owned by different entities, but it also brings severe challenge to privacy-preserving of private data. Secure multiparty computation is a key privacy-preserving technology, an important field of cryptography, and a focus of international cryptographic community. Privately comparing two numbers is a basic problem of secure multiparty computation. The protocols for this problem are building blocks to construct other privacy-preserving protocols. If the two numbers to be compared is small, there is no reliable solution to this problem that can resist active attacks. In many scenarios, the participants may be malicious and they may actively attack a protocol. If this is the case, there is no solution that can be used to privately compare the numbers. Therefore, it is of important theoretical and practical significance to design a protocol that can resist active attacks. This study first proposes a new technique called encrypt-and-choose and a new technology to resist active attacks:encoding+secure shuffle. Based on these techniques, a secure comparison protocol is first designed that is secure in the semi-honest model. Its security is proved by using the simulation paradigm. All possible active attacks are analyzed that the protocol may suffer from, and ElGamal multiplicative homomorphism and zero-knowledge proof of discrete logarithm and secure shuffle are used to resist possible active attacks. The protocol is then converted to one that can resist active attacks, and it is proved that it is secure against active attacks by using the ideal-real paradigm. Finally, the efficiency of the protocol is analyzed and tested. The experimental results demonstrate that the protocol is practical.