Seismic attenuation has been an interesting topic of research, for it reflects the inherent media characteristics in which seismic waves propagate. There are many factors that cause seismic wave attenuation, such as geometry attenuation caused by energy dissipating during propagation, friction attenuation by relative sliding among rock grains, and scattering attenuation by rock heterogeneity. In this paper we study P-wave scattering attenuation in a random elastic medium by numerical simulations from a statistical point of view. A random elastic medium model is built based on general stochastic process theory. Then a staggered-grid pseudo-spectral method is used to simulate wave propagation. Scattering attenuation is estimated by the spectral ratio method based on virtual detector records. Random elastic media numerical scatter results with various heterogeneity levels show that the higher heterogeneous levels cause greater scattering attenuation. When the scatter sizes are smaller than a wave length, the larger scatters give a greater attenuation. Finally, we propose a method to evaluate fluid-flow attenuation in porous media. The fluid-flow attenuation is derived from total attenuation and scattering attenuation in random porous media and the attenuation is estimated quantitatively. Results show that in the real seismic frequency range when the heterogeneous scale is about 101 meters (less than one wave length), scattering attenuation is larger than fluid-flow attenuation in random porous media and scattering attenuation is the main factor of seismic attenuation in real heterogeneous porous media.