Searching for dark matter in cosmic ray experiments is a Sisyphean task. An analogy would be searching for new physics at the LHC without knowing the parton distribution function inside the proton. Indeed, recently we have witnessed several astrophysical excesses (positrons in PAMELA, electrons in Fermi) hailed as "dark matter" by particle theorists and as "pulsars-or-whatever" by astrophysicists, with both sides on an equally firm footing. The feeling is that an unambiguous discovery of dark matter in cosmic rays is impossible, although we may well discover dark matter by other means and then use astrophysics to constrain its properties.
There is however one exception from that rule. The lore is that only dark matter can produce a monochromatic photon line; all standard cosmic phenomena we're aware of produce a continuous spectrum of photons that can usually be well approximated by a power law. On the other hand, gamma ray line can be easily produced by annihilation of weak-scale dark matter particles in the galactic center. Today, the average velocity of dark matter particles in our galaxy is about 1/1000 of the speed of light, thus they are practically at rest from the point of view of the relativistic kinematics. If two dark matter particles meet and annihilate into 2 photons (or 1 photon plus 1 other neutral particle) conservation of momentum implies that the energy of the outgoing photons must be equal to the dark matter mass. Therefore an observation of a gamma ray line from the galactic center would be considered a smoking gun signal of dark matter, and as a bonus it would give us an estimate of the mass of the dark matter particle.
That sounds too good to be true, and indeed there is a catch. Dark matter is...well... dark, so it does not directly couple to photons. In a generic model of dark matter, the 2-to-2 annihilation into photons is a loop-mediated process whose rate is orders of magnitudes smaller than the annihilation into other Standard Model particles. The latter does not produce a photon line but rather a continuous spectrum (for example, annihilation into electrons would produce photons via a bremsstrahlung off the final state). For this reason, a random model from the zoo, such as Minimal Dark Matter or the MSSM neutralino, cannot predict a gamma-ray line at an observable level, given current constraints on the gamma ray continuum flux. Nevertheless, there exists a handful of models where the 2-to-2 annihilation into photons is enhanced above the naive expectation, and the resulting line is potentially observable, see for example here or here.
Now, last Friday a new paper appeared on arXiv which claims that a gamma ray line is present in the data collected during the last 4 years by the Fermi satellite! The line can be found by a sophisticated analysis that selects the region of the sky with the optimal signal-to-background ratio depending on the assumed density profile of dark matter. Such an analysis reveals a bump over the gamma ray continuum near 130 GeV. This would correspond to a dark matter particle with the mass of 130±2 GeV annihilating into 2 photons (or a ~145 GeV particle annihilating into γ+Z boson, or a ~155 GeV particle annihilating into γ+Higgs). The local significance of the bump is 4.6 sigma, or 3.3 sigma after the look-elsewhere effect is taken into account. The best fit annihilation cross section is slightly above 10^-27 cm^3/s, which is a typical cross section value for the weak interaction processes, and about 1/10 of the total annihilation cross section expected if dark matter is a thermal relic from the big bang.
One should note the analysis has been performed not by the Fermi collaboration but by an outsider. In fact, a similar analysis by Fermi himself found no significant gamma ray line signal (but using less data and less fancy statistical methods). Unlike in particle physics, data collected by astrophysics experiments are often publicly available and can be independently analyzed by anyone with enough skill and will. Of course, an outsider's analysis carries less weight because some relevant information needed to precisely assess systematic errors is known only to the collaboration members. Nevertheless, in the past at least one discovery in the Fermi data was first claimed in an independent analysis and only later confirmed and blessed by the collaboration. In the case at hand, the members of the collaboration admit off the record that the line is indeed present in their data, although with a somewhat smaller statistical significance than claimed in the paper.
Clearly, at this point we cannot yet claim to have observed a dark matter signal. But the excess seems interesting enough to keep a close eye on further developments and wait for the official word from Fermi. And maybe indulge in some ambulance chasing in the meantime ;-)