The aim of the paper is to develop a technique for rebalancing pension fund portfolios in function of their pointwise level of risk. The performance of pension funds is often measured by their global asset returns because of the latter's in influence on periodic contributions and/or future benefits. However, in periods of market crisis attention is focused on the risk level given their social security (and not speculative) function. We describe the process of the global asset return by a multifractional Brownian motion using the function H(t) to detect high or low volatility phases. A procedure is carried out to balance the asset composition when the established local degree of risk is exceeded. The application is carried out on portfolios obtained in accordance with Italian regulations regarding investment limits.