A Stacked Deep Learning Approach for IoT Cyberattack Detection

Table 4

The accuracy of our method compared with the tested methods on the power system datasets.

Model	Power system
	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
Ours	97.7	97.9	97.2	96.0	97.1	96.8	98.3	97.2	96.7	98.0	97.9	98.0	98.5	98.4	96.9
[16]	93.4	89.3	89.8	88.8	32.5	90.2	75.2	74.9	89.1	88.5	89.8	89.6	92.7	90.1	88.5
[21]	90.0	88.5	86.8	84.9	88.6	85.6	88.5	89.2	87.2	88.8	86.7	86.3	88.9	88.1	87.4
RF	93.8	93.1	94.5	93.3	94.0	93.4	94.3	94.0	93.2	94.0	94.3	93.9	96.0	94.8	92.9
LR	87.7	71.0	72.7	67.9	75.0	71.3	77.5	74.3	69.9	73.1	76.3	69.4	78.5	75.0	68.7
DT	92.5	90.0	89.5	89.4	92.3	91.3	89.2	92.3	90.1	87.9	91.2	89.7	92.7	90.6	88.5
LDA	77.6	73.2	72.1	69.0	76.2	72.3	78.7	74.0	70.5	73.2	77.6	69.6	79.1	74.9	70.0
QDA	50.2	51.4	48.8	50.7	44.9	47.5	49.3	47.8	52.8	50.6	43.2	54.3	49.7	46.4	57.3
KNN	89.4	87.4	86.7	85.1	89.2	85.8	88.0	89.2	86.6	88.4	86.0	86.0	88.9	88.1	87.5
NB	25.2	34.1	33.5	39.2	31.5	32.1	29.2	60.8	35.6	33.4	75.8	39.0	27.6	28.9	38.7
XGB	88.3	84.2	88.4	81.7	81.7	82.6	84.7	86.5	84.4	86.4	84.5	80.9	86.3	83.6	83.4
MLP	79.1	69.4	71.1	62.1	71.5	32.0	73.9	28.0	65.7	71.6	76.5	65.2	76.2	28.4	62.7
RSS	93.8	92.1	93.8	92.8	94.2	93.1	92.8	93.9	92.0	92.7	92.0	92.4	95.5	94.3	92.0
GB	90.0	85.7	87.5	82.5	83.5	83.9	85.6	87.3	85.5	87.7	86.0	84.6	87.1	85.7	85.3