The human skin is modeled as a stack of homogeneous layers in the terahertz and submillimeter waves regions with some anisotropy due to the helical sweat glands and other elongated entities. A dielectric model for the skin is presented, valid for a wider frequency range (up to the terahertz region) taking into account the dispersive nature of the effective conductivity. Polarized reflectivity and generalized ellipsometric parameters are calculated versus angle and wavelength. Recent studies have claimed that the helical sweat ducts act as an array of low-Q helical antennae and are dominant in shaping the spectral response in the sub-terahertz region. We found that water absorption, dispersion and multiple interference effects play the major role in shaping the spectrum without the need for the assumption of the sweat ducts acting as low-Q helical antennae. High sensitivities to the water content are found particularly in the ellipsometric parameters at large incidence angles. Hence a new methodology is proposed to detect skin cancer using variable angle ellipsometry or polarized reflectometry. The parameter found with the highest sensitivity to water content is cos Δpp with Δpp being the phase of the on-diagonal reflection matrix ratio between p-to-p polarization.