The problem of estimating the time-varying net reactivity from flux measurements is solved for a point reactor kinetics model using a linear filtering technique in an H∞ setting. In order to use this technique, an appropriate dynamical model of the reactivity is constructed that can be embedded into the reactor model as one of its state variables. A filter, which minimizes the norm of the estimation error power spectrum, operates on neutron density measurements corrupted by noise and provides an estimate of the dynamic net reactivity. Computer simulations are performed to reveal the basic characteristics of the H∞ optimal filter. The results of the simulation indicate that the filter can be used to determine the time-varying reactivity from neutron density measurements that have been corrupted by noise.