In this paper, we consider a gravitational action containing a combination of the Ricci scalar, R, and the topological Gauss-Bonnet term, G. Specifically, we study the cosmological features of a particular class of modified gravity theories selected by symmetry considerations, namely the f (R, G) = R n G 1−n model. In the context of a spatially flat, homogeneous and isotropic background, we show that the currently observed acceleration of the Universe can be addressed through geometry, hence avoiding de facto the shortcomings of the cosmological constant. We thus present a strategy to numerically solve the Friedmann equations in presence of pressureless matter and obtain the redshift behavior of the Hubble expansion rate. Then, to check the viability of the model, we place constraints on the free parameters of the theory by means of a Bayesian Monte Carlo method applied to late-time cosmic observations. Our results show that the f (R, G) model is capable of mimicking the low-redshift behavior of the standard ΛCDM model, though substantial differences emerge when going toward high redshifts, leading to the absence of a standard matter-dominated epoch. Finally, we investigate the energy conditions and show that, under suitable choices for the values of the cosmographic parameters, they are all violated when considering the mean value of n obtained from our analysis, as occurs in the case of a dark fluid.