We consider a simple renormalizable dark matter model consisting of two real scalars with a mass splitting $\delta$, interacting with the SM particles through the Higgs portal. We find a viable parameter space respecting all the bounds imposed by invisible Higgs decay experiments at the LHC, the direct detection experiments by XENON100 and LUX and the dark matter relic abundance provided by WMAP and Planck. Despite the singlet scalar dark matter model that is fragile against the future direct detection experiments, the scalar split model introduced here survives such forthcoming bounds. We emphasize on the role of the co-annihilation processes and the mixing effects in this feature. For $m_{\text{DM}} \sim 63$ GeV in this model we can explain as well the observed gamma-ray excess in the analyses of the Fermi-LAT data at Galactic latitudes $2^{\circ} \leq |b| \leq 20^{\circ}$ and Galactic longitudes $|l| < 20^{\circ}$.

We propose a renormalizable dark matter model in which a fermionic dark matter (DM) candidate communicates with the standard model particles through two distinct portals: Higgs and vector portals. The dark sector is charged under a $U(1)'$ gauge symmetry while the standard model has a leptophobic interaction with the dark vector boson. The leading contribution of DM-nucleon elastic scattering cross section begins at one-loop level. The model meets all the constraints imposed by direct detection experiments provided by LUX and XENON100, observed relic abundance according to WMAP and Planck, and the invisible Higgs decay width measured at the LHC. It turns out that the dark matter mass in the viable parameter space can take values from a few GeV up to 1 TeV. This is a new feature which is absent in the models with only one portal. In addition, we can find in the constrained regions of the parameter space a DM mass of $\sim 34$ GeV annihilating into $b$ quark pair, which explains the Fermi-LAT gamma-ray excess.