being timely (R1) and significant/valuable (R3); extensive analysis and experimental results provided (R1

We thank the reviewers for all remarks/suggestions. As for the concerns/questions raised, we believe we successfully addressed every single one, as explained below. Our method offers great options to protect system with reasonable extra computation. Further, using reduced minibatch size, effective redundancy reduces to as low as 1.25, for example, while providing a This is a tough but highly relevant question. We feel a more effective solution is reducing mini-batch size as shown in A.4 in SM.