Abstract: Covert underwater acoustic communication is typically achieved by embedding information within environmental noise, however, the waveforms generated by existing techniques tend to be relatively regular, making them susceptible to detection through energy variation analysis and other methods. This paper proposes a covert communication system in which symbols are mapped onto the parameter space of a shallow-sea noise model. The tunability of distribution parameters and their significant influence on the probability density function are utilized, with the transmitted information modulated onto the characteristic parameter and skewness parameter of the model. At the receiver, information demodulation is accomplished by estimating the parameter values from the received signal. To address the challenge that the superposition of the transmitted signal and underwater background noise may alter the modulated parameter values and hinder accurate parameter estimation, the variation range of the parameter values of the mixed signal is analyzed, and the decision threshold at the receiver is optimized accordingly. The experimental results showed that the bit error rate can reach 10^-4 under the appropriate generalized signal-to-noise ratio and symbol transmission rate.