Assessing the vulnerability of critical infrastructure objects is of major concern when dealing with the process of dependability and risk management. Special attention is paid to the objects of higher interest, such as e.g. nuclear power plants. In spite of the protection of these objects, there is still a certain level of a potential threat. The aim of the paper is to describe potential model of a possible way of attacking on the object in order to get into a particular part of it. For this reason as well as for modelling adversary´s behaviour the diffusion processes have been used. The intention is to explain how and when the goal can be achieved by an adversary. The results are assumed to be possibly used for setting the efficiency of an object physical protection system, or the algorithm development of the way a possible adversary selects the goal.
