Arctic sea-ice contains imperfections such as cracks, leads and pressure ridges that scatter flexuralgravity waves. Models for predicting scattering have been described in the literature, concentrating mainly on singular isolated features with simplified shapes or on arrays of such features. In reality ridges are seldom simple and leads are rarely entirely free of ice. Here we describe a model in which the scattering by a sheet of arbitrary thickness can be simulated. Linear wave theory and Green's functions are used to derive the governing equations for a numerical model of a two-dimensional (in the vertical) system. We examine wave scattering by random ice sheets, identifying trends in behavior as the wave period and the length, median thickness and variance of the sheet are changed. It has been suggested that wave scattering could be used to identify sea-ice thickness, a task which is difficult or expensive by other methods, and here we examine a technique by which this could potentially be achieved. However, a large data base is necessary for this to work and this may limit the practicality of the approach.