Anomaly Detection for Industrial Control System Based on Autoencoder Neural Network

Table 4

Comparisons with other models. The results of TABOR are taken from paper [23]. And the results of SVM and DNN are taken from [24].

No.	Target	Detection	Recall
			DNN	SVM	TABOR	Com-AE
1	MV-101	FIT-101, MV-101, PIT-502	0	0	0.049	0.966
2	P-102	P-102, P-302, FIT301	0	0	0.930	0.919
3	LIT-101	P-302, FIT-301, DPIT-301	0	0	0	0.120
4	MV-504	—	0	0.035	0.328	0
5	AIT-202	AIT-202, P-203, P-302	0.717	0.720	0.995	1.0
6	LIT-301	P-205, P-203, P101	0	0.888	0	0.354
7	DPIT-301	P-302, DPIT-301, FIT-301	0.927	0.919	0.992	0.992
8	FIT-401	FIT-401, UV-401, P-501	1	0.433	0.994	0.150
9	FIT-401	UV-401, FIT-401, P-501	0.978	1	0.998	1.0
10	MV-304	—	0	0	0	0
11	MV-303	—	0	0	0	0
12	LIT-301	—	0	0	0	0
13	MV-303	P-101, MV301, P-302	0	0	0.597	0.275
14	AIT-504	—	0.123	0.13	0.004	0
15	AIT-504	AIT-504, AIT-503, AIT-501	0.845	0.848	0.997	0.929
16	MV-101,LIT-101	—	0	0.0167	0.083	0
17	UV-401,AIT-502,P-501	UV-401, P-501, FIT-504	0.998	1	0.998	0.961
18	P-602,DPIT-301,MV-302	DPIT-301, P-302, FIT-301	0.867	0.875	0	0.984
19	P-203,P-205	P-203, P-205, P-204	0	0	0	0.523
20	LIT-401,P402	—	0	0.009	0	0
21	P-101,LIT-301	P-102, AIT-503, FIT-201	0	0	0.999	0.976
22	P-302,LIT-401	—	0	0	0.196	0
23	P-302	P-302, MV-304, FIT-301	0.936	0.936	1.000	0.977
24	P-201,P-203,P-205	—	0	0	0	0
25	P-101,MV-101,LIT-101	FIT-201, P-102, AIT-503	0	0.003	0.999	0.964
26	LIT-401	—	0	0	0	0
27	LIT-301	P-205, P-203, P-101	0	0.905	0	0.045
28	LIT-101	P-302, FIT-301, DPIT-301	0	0	0.890	0.225
29	P-101	P-102, P-101, AIT-503	0	0	0.990	0.178
30	P-101,P-102	AIT-202, P-101, FIT-201	0	0	0.258	0.507
31	LIT-101	AIT503, MV304, LIT-101	0	0.119	0.889	0.796
32	P-501,FIT-502	FIT-504, FIT-503, PIT-502	1	1	0.998	0.795
33	AIT402,AIT502	FIT-101, AIT-502, AIT-402	0.923	0.927	0.996	1.0
34	FIT-401,AIT-502	AIT-503, FIT-401, AIT-502	0.940	0	0.369	0.788
35	FIT-401	FIT-401, P-501, UV-401	0.933	0.927	0.997	0.805
36	LIT-301	LIT-301, AIT-402, AIT-502	0	0.357	0	0.321