The presence of dark matter is inferred at scales ranging from rotations of galaxies to imprints in the CMB – the Big Bang after-glow. The nature of dark matter is, however, still unknown as no detection other than the gravitational one has been made. This thesis presents two analyses searching for a neutrino signal from dark matter annihilations in the Milky Way.

The first analysis searched for an excess of ν_μ charged current events with directions from the central region of the dark matter halo and, was focused on low energy events, thus probing low dark matter particle masses. Approximately 319 days of data collected with the 79-string configuration of the IceCube detector was used in the analysis. Despite a large deficit in the number of observed events the data were found to be consistent with background and upper limits were set on <σⱴ>. At the time of the analysis these limits were the strongest set by a neutrino experiment below 100 GeV.

The second analysis was performed on a data sample originally used in an unfolding analysis of the atmospheric and astrophysical neutrino spectra. The data consisted of contained cascade events above 1 TeV collected with the 79-string configuration and the completed detector in the 86-string configuration during two years of data-taking. The limits set by this analysis were more constraining by up to a factor of 10 compared to previous IceCube analyses, and the most competitive limits are set assuming a Burkert halo profile. These two analyses prompted the development of a signal subtraction likelihood method to address the problem of signal contamination in background estimates based on scrambled data.

Additionally a study concerning future extensions of IceCube in the Gen2 project is presented. The cascade reconstruction performance was examined and compared for different proposed detector extensions.