Aiming at secure analog speech communication, a homology sound-based algorithm for speech signal interference is proposed in this paper. We first split speech signal into phonetic fragments by a short-term energy method and establish an interference noise cache library with the phonetic fragments. Then we implement the homology sound interference by mixing the randomly selected interferential fragments and the original speech in real time. The computer simulation results indicated that the interference produced by this algorithm has advantages of real time, randomness, and high correlation with the original signal, comparing with the traditional noise interference methods such as white noise interference. After further studies, the proposed algorithm may be readily used in secure speech communication.