Dimension Reduction Technique Based on Supervised Autoencoder for Intrusion Detection of Industrial Control Systems

Table 5

Comparisons of accuracy. The first two rows are dimension reduction methods and classifiers, respectively. The data set number is shown in the first column. The last two rows provide the mean value and standard deviation of the results. “None” indicates using the original features.

No.	None				SupervisedAE				SupervisedAE + PCA
	KNN	DT	AdaBoost	Bagging	KNN	DT	AdaBoost	Bagging	KNN	DT	AdaBoost	Bagging
1	0.8635	0.8276	0.8200	0.8963	0.9265	0.8699	0.8679	0.9255	0.9303	0.8689	0.8661	0.9211
2	0.8580	0.8398	0.8382	0.9041	0.9256	0.8631	0.8712	0.9148	0.9313	0.8674	0.8611	0.9102
3	0.8792	0.8569	0.8556	0.9034	0.9324	0.8702	0.8696	0.9250	0.9278	0.8735	0.8753	0.9151
4	0.8716	0.8358	0.8406	0.9022	0.9385	0.8766	0.8766	0.9275	0.9350	0.8658	0.8660	0.9248
5	0.8727	0.8318	0.8347	0.9014	0.9320	0.8690	0.8615	0.9198	0.9316	0.8617	0.8553	0.9173
6	0.8712	0.8474	0.8496	0.9108	0.9426	0.8863	0.8800	0.9360	0.9418	0.8824	0.8842	0.9330
7	0.8699	0.8382	0.8281	0.8923	0.9332	0.8642	0.8753	0.9278	0.9318	0.8803	0.8684	0.9158
8	0.8666	0.8551	0.8517	0.9101	0.9332	0.8707	0.8745	0.9223	0.9340	0.8766	0.8715	0.9208
9	0.8577	0.8174	0.8208	0.8798	0.9390	0.8629	0.8672	0.9257	0.9348	0.8633	0.8663	0.9213
10	0.8729	0.8373	0.8391	0.8944	0.9427	0.8872	0.8813	0.9325	0.9382	0.8775	0.8836	0.9307
11	0.8756	0.8235	0.8244	0.8983	0.9444	0.8865	0.8915	0.9370	0.9461	0.8835	0.8867	0.9313
12	0.8773	0.8267	0.8245	0.8951	0.9412	0.8786	0.8754	0.9244	0.9418	0.8800	0.8786	0.9242
13	0.8668	0.8425	0.8397	0.9034	0.9418	0.8812	0.8816	0.9275	0.9448	0.8744	0.8831	0.9290
14	0.8659	0.8282	0.8272	0.8972	0.9363	0.8762	0.8704	0.9320	0.9396	0.8751	0.8739	0.9247
15	0.8681	0.8215	0.8254	0.8906	0.9378	0.8753	0.8791	0.9263	0.9375	0.8789	0.8751	0.9229
Mean	0.8691	0.8353	0.8346	0 . 8986	0 . 9365	0.8745	0.8749	0.9269	0 . 9364	0.8740	0.8730	0.9228
Std	0.0061	0.0114	0.0111	0.0076	0.0056	0.0080	0.0071	0.0056	0.0053	0.0068	0.0089	0.0063