If they are difference stationary (unit root), my suggestion is not to take difference and then do regression. But you can do regression on the raw data and do cointegration analysis because in cointegration, the parameter convergence rate is faster ( T versus square root T). Thus it requires less data points. However, 10 data points is still too small.