One obvious way to progress is to study the signal at lower energies (at higher energies, we don't see anything). This is, however, easier said than done, since the backgrounds rise steeply at lower energies. One way to handle that is to use larger and larger veto regions as the energy drops; this leaves a smaller and smaller fiducial (active) detector volume, but the signal should also rise at lower energies. Jakob Van Santen pursued this route, in an IceCube paper that was published in Phys. Rev. D. in January, and available on the Cornell preprint server, as arXiv:1410.1749.
The plots below show the results of that study, for the Northern and Southern skies respectively. The backgrounds get larger at lower energies. In the Southern sky, there are two types of backgrounds: penetrating muons and an irreducible background of atmospheric neutrinos, while, in the North, only the neutrinos are present. However, in both cases, it is possible to measure the astrophysical component down to deposited energies of a few TeV. Here, "deposited energy" means the energy visible in the detector.The measured flux is consistent in both hemspheres, and is well fit by a power-law spectrum: phi ~ (E_nu)^-p, where p, the power law index, is 2.46 +/-0.12. The spectrum is significantly softer (i.e. has more low-energy events and fewer high-energy events) than the standard benchmark spectrum, which is dN_nu/dE_nu ~ (E_nu)^-2. This is not a surprise; most of us thought that the -2 index was based on a simplified model which would not survive an encounter with data.
The second analysis, by Gary Binder, also looked at the energy spectrum (with similar results), and also looked at the neutrino flavor ratio: how many of the neutrinos are electron neutrinos (nu_e), vs. muon neutrinos (nu_mu) vs. tau neutrinos (nu_tau) . It is available on the Cornell preprint server, as arXiv:1502.03376. It found a similar spectral index (as have other IceCube studies).
The flavor ratio is somewhat tricky, in that there are three different types of neutrinos which interact via two topologies: long muon tracks from nu_mu, and roughly spherical showers, from nu_e and all flavor charged-current interactions. In this energy range, 83% of nu_tau produce showers, while 17% of them include muons. So, there is some ambiguity. Gary presented his results as a triangle:
The current analysis find a best fit indicated by the cross near the lower left. However, the confidence levels (shown via colors) show that all four models can adequately fit the data. However, the analysis does rule out non-standard models (not shown here, but discussed in the paper) such as some involving sterile (non-interacting) neutrinos.