The CRESST-II cryogenic Dark Matter search, aiming at detection of WIMPs via elastic scattering off nuclei in CaWO$_4$ crystals, completed 730 kg days of data taking in 2011. We present the data collected with eight detector modules, each with a two-channel readout; one for a phonon signal and the other for coincidently produced scintillation light. The former provides a precise measure of the energy deposited by an interaction, and the ratio of scintillation light to deposited energy can be used to discriminate different types of interacting particles and thus to distinguish possible signal events from the dominant backgrounds. Sixty-seven events are found in the acceptance region where a WIMP signal in the form of low energy nuclear recoils would be expected. We estimate background contributions to this observation from four sources: 1) "leakage" from the e/\gamma-band 2) "leakage" from the \alpha-particle band 3) neutrons and 4) Pb-206 recoils from Po-210 decay. Using a maximum likelihood analysis, we find, at a high statistical significance, that these sources alone are not sufficient to explain the data. The addition of a signal due to scattering of relatively light WIMPs could account for this discrepancy, and we determine the associated WIMP parameters.

In experiments for direct dark matter searches, neutrinos coherently scattering off nuclei can produce similar events as Weakly Interacting Massive Particles (WIMPs). To reach sensitivities better than about 10^-10 pb for the elastic WIMP nucleon spin-independent cross section in the zero-background limit, energy thresholds for nuclear recoils should be >2.05 keV for CaWO_4, >4.91 keV for Ge, >2.89 keV for Xe, >8.62 keV for Ar and >15.93 keV for Ne as target material. Atmospheric neutrinos limit the achievable sensitivity for the background-free direct dark matter search to >10^-12 pb.