On Nekhoroshev's estimate at an elliptic equilibrium