We show that cold dark matter particles interacting through a Yukawa potential could naturally explain the recently observed cores in dwarf galaxies without affecting the dynamics of objects with a much larger velocity dispersion, such as clusters of galaxies. The velocity dependence of the associated cross-section as well as the possible exothermic nature of the interaction alleviates earlier concerns about strongly interacting dark matter. Dark matter evaporation in low-mass objects might explain the observed deficit of satellite galaxies in the Milky Way halo and have important implications for the first galaxies and reionization. PACS numbers: 95.35+d, Introduction. The collisionless cold dark matter (CDM) model has been highly successful in accounting for the gravitational growth of density perturbations from their small observed amplitude at early cosmic times (as imprinted on the cosmic microwave background anisotropies [1]) to the present-day structure of the Universe on large scales. However, it is far from clear that the predictions of this model are valid on small scales.New data on low mass galaxies indicate that their dark matter distribution has a core [2], in contrast to the cusped profile expected from collisionless CDM simulations [3]. The mean value of the inner logarithmic slope of the mass density profile in seven dwarf galaxies within the THINGS survey is observed to be −0.29±0.07[4], much shallower than the expected slope of ∼ −1 from pure CDM simulations. Moreover, the dynamics of dwarf spheroidal galaxies, such as Fornax [5], Ursa-Minor [6], and Sculptor [7], whose luminosities and dynamical masses are smaller by 2-3 orders of magnitude than the THINGS galaxies, indicates a characteristic core density of ∼ 0.1 ± 0.05M pc −3 = (7 ± 4) × 10 −24 g cm −3 . Since these dwarf spheroidals are dominated by dark matter throughout, it is challenging to explain their inferred cores by the gravitational interaction of the dark matter with the baryons [8]. Although it is conceivable that powerful gas outflows from an early baryon-dominated nucleus would reduce the central dark matter density in luminous galaxies [9,10], the formation of a massive baryonic nucleus would initially compress the CDM [11] and exacerbate the discrepancy that needs to be resolved [8], and also potentially violate the observed low luminosities from dwarf galaxies at higher redshifts [12,13]. High-redshift observations of dwarf galaxies must find evidence for the required strong feedback phase, or else an alternative process is at play. Some recent simulations that include feedback do not observe the appearance of cores within the lowest luminosity galaxies [14].To alleviate early signs of the above discrepancy, Spergel & Steinhardt [15] adopted the Strongly-