Table 4: Experimental results with different settings, where the two values of each cell are of ‘benign’ and ‘malicious’, respectively.

Dimensions & Training optionsPrecisionRecallF1

E = 1599.68 / 100.092.52 / 95.7595.34 / 97.82
E = 3599.73 / 100.094.03 / 97.4796.58 / 98.71
Drop-out with 0.599.70 / 100.093.53 / 97.3796.23 / 98.66
L2 gradient-clipping with 0.399.74 / 100.095.25 / 97.1797.17 / 98.55
L2 gradient-clipping with 0.799.70 / 100.092.11 / 97.7895.14 / 98.86
w/o batch-normalization99.70 / 100.095.96 / 97.2797.68 / 98.61
P = 5099.70 / 100.095.05 / 97.0797.13 / 98.50
P = 15099.76 / 100.093.83 / 97.8896.29 / 98.92