This paper presents a digital speech encryption scheme based on symmetrical homomorphic encryption technique (MORE-method), in which a symmetrical key cryptosystems with probabilistic statistics and fully homomorphic properties are used to encrypt speech signals.In the proposed scheme, each sample of a speech signal is firstly multiplied one weight, and then encrypted, the normalization is exploited to make the data expend lossy compression.Finally, a recombination method of the cipher-text is proposed to obtain the corresponding speech cipher-text with good performances.Experimental results show the proposed scheme is correct and satisfy homomorphism, which has strong diffusibility and a large key-space, is robustness to entropy attacks and statistical analysis attacks, and decreased the residual intelligibility as small as possible.Moreover, the encrypted speech can be decrypted completely and the original speech signal can be recovered accurately.Compared with two dimensinal chaotic maps cryptosystem and Paillier cryptosystem, the MORE-method is more security than two dimensinal chaotic maps cryptosystem, and has lower complexity for speech encryption than Paillier cryptosystem, so the proposed scheme especially meets the sensitive speech security in the cloud.