IceTop, the surface component of the IceCube Neutrino Observatory detects extensive air showers (EAS) initiated by cosmic rays and gamma rays in the energy range of PeV to EeV. IceTop is sensitive to air shower characteristics such as the muon content, local fluctuations and shower-front curvature. These characteristics correlate with the mass of the primary particle for a given energy and arrival direction. Using IceTop observables, we construct three two-dimensional probability distribution functions (PDFs) that reflect these shower-front properties. A log-likelihood ratio is constructed using these PDFs, assuming a pair of hypotheses corresponding to primaries with distinct mass. We show the effectiveness of this parameter for discriminating gamma rays from cosmic rays and heavy from light cosmic ray primaries.