A good observation of preflare activities is important for us to understand the origin and triggering mechanism of solar flares, and to predict the occurrence of solar flares. This work presents the characteristics of microwave spectral fine structures as preflare activities of four solar flares observed by Ond\v{r}ejov radio spectrograph in the frequency range of 0.8--2.0 GHz. We found that these microwave bursts which occurred 1--4 minutes before the onset of flares have spectral fine structures with relatively weak intensities and very short timescales. They include microwave quasi-periodic pulsations (QPP) with very short period of 0.1-0.3 s and dot bursts with millisecond timescales and narrow frequency bandwidths. Accompanying these microwave bursts, there are filament motions, plasma ejection or loop brightening on the EUV imaging observations and non-thermal hard X-ray emission enhancements observed by RHESSI. These facts may reveal certain independent non-thermal energy releasing processes and particle acceleration before the onset of solar flares. They may be conducive to understand the nature of solar flares and predict their occurrence.

We present a model for dm-fiber bursts that is based on assuming fast sausage magnetoacoustic wave trains that propagate along a dense vertical filament or current sheet. Eight groups of dm-fiber bursts that were observed during solar flares were selected and analyzed by the wavelet analysis method. To model these fiber bursts we built a semi-empirical model. We also did magnetohydrodynamic simulations of a propagation of the magnetoacoustic wave train in a vertical and gravitationally stratified current sheet. In the wavelet spectra of the fiber bursts computed at different radio frequencies we found the wavelet tadpoles, whose head maxima have the same frequency drift as the drift of fiber bursts. It indicates that the drift of these fiber bursts can be explained by the propagating fast sausage magnetoacoustic wave train. Using new semi-empirical and magnetohydrodynamic models with a simple radio emission model we generated the artificial radio spectra of the fiber bursts, which are similar to the observed ones.