The weakly interacting massive particle (WIMP) paradigm gives an elegant mechanism for the generation of dark matter densities with the appropriate properties to be consistent with current observations. While in its minimal implementation, the only opportunity to test the reality of the scenario is by directly detecting the WIMP itself, cosmic ray anomalies observed several years ago imply the possibility of an extended dark matter sector in which secondary annihilations or decays can occur before interacting with the standard model. This opens up the possibility of detecting other parts of dark sector in a much wider set of experiments than those typically considered in connection to dark matter. In this dissertation I study several alternative models, capable of yielding signals across a broad range of experimental approaches.

I study the prospects for detecting a light boson X with mass m_X ≤ 100 MeV at a low energy electron-proton collider. Focus is on the case where X dominantly decays to e⁺e^- as motivated by recent “dark force” models. In order to evade direct and indirect constraints, X must have small couplings to the standard model (α_X ≤ 10^-8) and a sufficiently large mass (m_X ≥ 10 MeV). By comparing the signal and background cross sections for the $e^-p e⁺e^- final state, I conclude that dark force detection requires an integrated luminosity of around 1 ab^-1. This proposal is currently being implements by the DarkLight collaboration at the Thomas Jefferson National Accelerator Facility.

I also investigate the bounds on axion-like states from flavor-changing neutral current b → s decays, assuming the axion couples to the standard model through mixing with the Higgs sector. Such GeV-scale axions have received renewed attention in connection with observed cosmic ray excesses. I find that existing B → K l⁺l^- data impose stringent bounds on the axion decay constant in the multi-TeV range, relevant for constraining the “axion portal” model of dark matter. Such bounds also constrain light Higgs scenarios in the next-to-minimal supersymmetric standard model. These bounds can be improved by dedicated searches in B-factory data and at LHCb.

While looking at direct dark matter detection experiments themselves, it is often assumed that the first evidence for dark matter will come from experiments probing spin-independent interactions, with much higher sensitivities due to coherence effects. I explore models that would be invisible in such experiments, but detectable via spin-dependent interactions. The existence of much larger (or even only) spin-dependent tree-level interactions is not sufficient, due to potential spin-independent subdominant or loop-induced interactions, and I find that in this way most models with detectable spin-dependent interactions would also generate detectable spin-independent interactions. Models in which a light pseudoscalar acts as the mediator seem to uniquely evade this conclusion. In presenting a particular viable dark matter model generating such an interaction, a tens of MeV--GeV-scale axion is found to be an attractive candidate independently of considerations presented earlier.