The modelling of stochastic media is of considerable interest for criticality safety and reactor physics applications. Novel methods for modelling highly disordered, random heterogeneous mixtures based on randomized Weierstrass functions and Perlin noise have been implemented in the MONK® Monte Carlo code using Woodcock tracking. Perlin noise in particular is shown to be well suited to the application, efficiently producing random geometries with much greater physical realism than the artificial nature of alternative methods. The effect of geometry randomization on neutron multiplication and reactor kinetics parameters in a simple benchmark configuration has been investigated, and results are compared with previous research based on tessellation algorithms. While the results are found to be broadly consistent with previous work, the new methods presented here offer significant benefits in terms of increased physical realism, conservation of material volume fractions, computational efficiency and ease of use.