A linear system has two properties: scaling (f(Ax)=Af(x)) and superposition (f(Ax+By)=Af(x)+Bf). Try to test the second property, when x is a sinewave and y is a constant positive level offset. In a linear system you would obtain the same half-period waveform, but shifted above zero. In rectifier the output will be different from that. Anyway, it's not the strict proof. Someone, plz, correct me if i'm wrong