Intrusion Detection Method of ICS Based on Improved CGAN Algorithm

doi:10.3969/j.issn.1671-1122.2023.01.005

Abstract

Abstract:

In this paper, an improved conditional generative adversarial network algorithm was proposed, and the Wasserstein distance was added to measure the distance between synthetic and real samples, for solving the instability problem that cause the generator gradient to disappear when two types of samples were ignored to overlap in CGAN. The effectiveness of the algorithm was verified on the UCI dataset with different imbalance rates. Then the WCGAN-SVM intrusion detection model of industrial control system was constructed and verified on the industrial control dataset SWaT. The experimental results show that the method increases the accuracy of detecting attack samples by 3.51% and decreases the false alarm rate and the false negative rate by 2.29% and 2.19% compared with SVM.

Key words: ICS, intrusion detection, WCGAN, unbalance data, SVM

CLC Number:

TP309

WANG Huazhong, TIAN Zilei. Intrusion Detection Method of ICS Based on Improved CGAN Algorithm[J]. Netinfo Security, 2023, 23(1): 36-43.

Figures/Tables 10

References 18

[1]	THAKKAR A, LOHIYA R. A Survey on Intrusion Detection System: Feature Selection, Model, Performance Measures, Application Perspective, Challenges, and Future Research Directions[J]. Artificial Intelligence Review, 2021, 55(3): 453-563. doi: 10.1007/s10462-021-10037-9 URL
[2]	CASE D U. Analysis of the Cyber Attack on the Ukrainian Power Grid[J]. Electricity Information Sharing and Analysis Center(E-ISAC), 2016, 388: 1-29.
[3]	HOBBS A. The Colonial Pipeline Hack: Exposing Vulnerabilities in Us Cybersecurity[M]. London: SAGE Publications, 2021.
[4]	ZHANG Wenan, HONG Zhen, ZHU Junwei, et al. A Survey of Network Intrusion Detection Methods for Industrial Control Systems[J]. Control and Decision, 2019, 34(11): 2277-2288.
	张文安, 洪榛, 朱俊威, 等. 工业控制系统网络入侵检测方法综述[J]. 控制与决策, 2019, 34(11): 2277-2288.
[5]	INOUE J, YAMAGATA Y, CHEN Yuqi, et al. Anomaly Detection for a Water Treatment System Using Unsupervised Machine Learning[C]// IEEE. 17th IEEE International Conference on Data Mining(ICDMW). New York: IEEE, 2017: 1058-1065.
[6]	MOIZUDDIN M D, JOSE M V. A Bio-Inspired Hybrid Deep Learning Model for Network Intrusion Detection[EB/OL]. (2022-02-28)[2022-03-28]. https://doi.org/10.1016/j.knosys.2021.107894.
[7]	THASEEN I S, BANU J S, LAVANYA K, et al. An Integrated Intrusion Detection System Using Correlation-Based Attribute Selection and Artificial Neural Network[EB/OL]. (2020-08-08)[2022-01-13]. https://doi.org/10.1002/ett.4014.
[8]	LIU Shuo, ZHANG Xinglan. Intrusion Detection Based on Dual Attention[J]. Netinfo Security, 2022, 22(1): 80-86.
	刘烁, 张兴兰. 基于双重注意力的入侵检测系统[J]. 信息网络安全, 2022, 22(1):80-86.
[9]	MBOW M, KOIDE H, SAKURAI K. An Intrusion Detection System for Imbalanced Dataset Based on Deep Learning[C]//IEEE. 9th International Symposium on Computing and Networking(CANDAR). New York: IEEE, 2021: 38-47.
[10]	DOUZAS G, BACAO F, LAST F. Improving Imbalanced Learning through a Heuristic Oversampling Method Based on K-Means and SMOTE[J]. Information Sciences, 2018, 465: 1-20. doi: 10.1016/j.ins.2018.06.056 URL
[11]	ZHOU Xiaokang, HU Yiyong, LIANG Wei, et al. Variational LSTM Enhanced Anomaly Detection for Industrial Big Data[J]. IEEE Transactions on Industrial Informatics, 2021, 17(5): 3469-3477. doi: 10.1109/TII.2020.3022432 URL
[12]	MOAYEDIKIA A, ONG K L, BOO Y L, et al. Feature Selection for High Dimensional Imbalanced Class Data Using Harmony Search[J]. Engineering Applications of Artificial Intelligence, 2017, 57: 38-49. doi: 10.1016/j.engappai.2016.10.008 URL
[13]	GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative Adversarial Nets[C]// ACM. 27th International Conference on Neural Information Processing Systems. New York: ACM, 2014: 2672-2680.
[14]	GAUTHIER J. Conditional Generative Adversarial Nets for Convolutional Face Generation[EB/OL]. [2022-02-20]. http://cs231n.stanford.edu/reports/2015/pdfs/jgauthie_final_report.pdf.
[15]	TANG Xianlun, DU Yiming, LIU Yuwei, et al. Image Recognition with Conditional Deep Convolutional Generative Adversarial Networks[J]. Acta Automatica Sinica, 2018, 44(5): 855-864.
	唐贤伦, 杜一铭, 刘雨微, 等. 基于条件深度卷积生成对抗网络的图像识别方法[J]. 自动化学报, 2018, 44(5):855-864.
[16]	MIRZA M, OSINDERO S. Conditional Generative Adversarial Nets[EB/OL]. (2014-11-06)[2022-01-29]. https://arxiv.org/abs/1411.1784.
[17]	LI Yanxia, CHAI Yi, HU Youqiang, et al. Review of Imbalanced Data Classification Method[J]. Control and Decision, 2019, 34(4): 673-688.
	李艳霞, 柴毅, 胡友强, 等. 不平衡数据分类方法综述[J]. 控制与决策, 2019, 34(4):673-688.
[18]	MATHUR A P, TIPPENHAUER N O. SWaT: A Water Treatment Testbed for Research and Training on ICS Security[C]// IEEE. International Workshop on Cyber-Physical Systems for Smart Water Networks(CySWater). New York: IEEE, 2016: 31-36.

数据集	属性/个	少数类类别	正类数量/个	负类数量/个	IR
glass	9	1	70	144	2.06
carEvaluation	6	acc	384	1344	3.5
segment	16	window	330	3210	6
ecoli	7	imU	35	301	8.6
hungarian	14	2	26	268	10.31

预测实际	正类	负类
正类	TP	FN
负类	FP	TN

数据集	方法	None	SMOTE	ADASYN	WCGAN
glass	LR	0.644	0.662	0.644	0.679
	KNN	0.691	0.697	0.663	0.682
	DT	0.701	0.735	0.721	0.752
	SVM	0.698	0.781	0.762	0.788
carEvaluation	LR	0.685	0.703	0.715	0.763
	KNN	0.711	0.755	0.750	0.759
	DT	0.682	0.711	0.727	0.741
	SVM	0.764	0.779	0.801	0.783
segment	LR	0.524	0.607	0.588	0.597
	KNN	0.438	0.502	0.590	0.621
	DT	0.407	0.459	0.437	0.498
	SVM	0.600	0.645	0.619	0.717
ecoli	LR	0.715	0.764	0.731	0.802
	KNN	0.721	0.741	0.703	0.784
	DT	0.733	0.745	0.777	0.801
	SVM	0.784	0.797	0.791	0.812
hungarian	LR	0.432	0.457	0.433	0.487
	KNN	0.397	0.404	0.471	0.506
	DT	0.607	0.631	0.674	0.688
	SVM	0.579	0.654	0.669	0.696

数据集	方法	None	SMOTE	ADASYN	WCGAN
glass	LR	0.732	0.758	0.731	0.768
	KNN	0.710	0.772	0.746	0.752
	DT	0.771	0.784	0.775	0.813
	SVM	0.818	0.870	0.851	0.853
carEvaluation	LR	0.779	0.818	0.819	0.854
	KNN	0.873	0.886	0.881	0.889
	DT	0.792	0.821	0.835	0.847
	SVM	0.868	0.891	0.902	0.898
segment	LR	0.681	0.688	0.671	0.681
	KNN	0.647	0.659	0.681	0.687
	DT	0.619	0.638	0.621	0.655
	SVM	0.678	0.691	0.667	0.699
ecoli	LR	0.856	0.891	0.877	0.912
	KNN	0.831	0.862	0.819	0.897
	DT	0.848	0.871	0.891	0.905
	SVM	0.911	0.924	0.919	0.955
hungarian	LR	0.623	0.641	0.633	0.649
	KNN	0.591	0.612	0.661	0.672
	DT	0.664	0.691	0.716	0.698
	SVM	0.644	0.691	0.679	0.712

组成部分	结构	激活函数	节点数/个
生成器G	输入层	ReLU	32
	隐藏层	LeakyReLU	256
	输出层	Linear	51
鉴别器D	输入层	ReLU	51
	隐藏层	LeakyReLU	256
	输出层	Linear	1