The potential of low-salinity (LS) water injection as an oil recovery technique has been the source of much recent debate within the petroleum industry. Evidence from both laboratory and field-level studies has indicated significant benefits compared to conventional high-salinity (HS) waterflooding, but many conflicting results have also been reported and, to date, the underlying mechanisms remain poorly understood. In this paper, we aim to address this uncertainty by developing a novel, steady-state pore network model in which LS brine displaces oil from a HS-bearing network. The model allows systematic investigation of the crude oil/brine/rock parameter space, with the goal of identifying features that may be critical to the production of incremental oil following LS brine injection. By coupling the displacement model to a salinity-tracking tracer algorithm, and assuming that a reduction of water salinity within the pore network leads to localised wettability alteration, substantial perturbations to standard pore filling sequences are predicted. The results clearly point to two principal effects of dynamic contact angle modification at the pore scale: a "pore sequence" effect, characterised by an alteration to the distribution of displaced pore sizes, and a "sweep efficiency" effect, demonstrated by a change in the overall fraction of pores invaded. Our study indicates that any LS effect will depend on the relative (scenario-dependent) influence of each mechanism, where factors such as the initial wettability state of the system and the pore size distribution of the underlying network are found to play crucial roles. In addition, we highlight the important role played by end-point capillary pressure in determining LS efficacy.