In page 6 the correct data leakage probability is "1-(1-prob_a)^i", the results of the paper are not affected by this error.