Complex organizations need to establish access control policies in order to manage access to restricted resources. Role Based Access Control paradigm has been introduced in ’90 years aiming at simplifying the management of centralized access control. The definition of a good set of roles in order to match the organizational requirements of a company is a problem partially solved by role mining techniques, which return automatically a set of roles compatible with the permissions assigned to users. Unfortunately, the problem of finding an optimal role set has been proved to be NP-hard; so heuristics have been introduced in order to approximate the optimal solution. In this work we propose a novel heuristic and compare its results showing its efficiency and effectiveness.