Abstract: Speech information hiding has long been a hot point and also a problem in domain of information security. In this paper, a new speech information hiding model based on theories of masking effect and spread spectrum codes is introduced, which is highly transparent using hopping fractional fourier transform. Introduction of this model is started with two algorithms derived from frequency masking—a confidential information embedding algorithm in frational fourier transform domain and an extracting algorithm in frational correlation cepstrum domain. Another wide interval multi-system pseudo-random sequence is also put forward in order to control the fractional order factors. According to the theoretical analysis and experimental result, this model is highly transparent that confidential information can not be found in time domain, frequency domain or even cepstrum domain. also, this model performs well in anti-noise. Our study suggests that this model has high value and good prospect in application.