Classical nucleation theory (CNT) is commonly used to describe the formation of ice in atmospheric models. A key parameter in CNT is the activation energy, ΔGact, which controls the molecular flux across the ice-liquid interface. ΔGact derives from the assumption that the interfacial flux follows an Arrhenius-type dependency on temperature, T. However it is not clear whether ΔGact can be unambiguously defined and independently measured. Many studies parameterize ΔGact assuming that diffusion across the solid-liquid interface is similar to that in the bulk, neglecting the effect of the ice germ on its environment. This work introduces a new approach to describe ice nucleation that does not require an explicitly definition of ΔGact. According to the new model the interfacial flux is controlled by bulk diffusion and the work dissipated from molecular rearrangement within the interface. Usage of the new model within a general form of CNT leads to an improved prediction of the ice nucleation rate, particularly at T below 230 K where it is strongly underestimated by current models.