基于安全联邦蒸馏GAN的工业CPS协作入侵检测系统

doi:10.11959/j.issn.1000-436x.2023216

通信学报 ›› 2023, Vol. 44 ›› Issue (12): 230-244.doi: 10.11959/j.issn.1000-436x.2023216

• 学术通信 • 上一篇

基于安全联邦蒸馏GAN的工业CPS协作入侵检测系统

梁俊威¹, 杨耿¹, 马懋德², Muhammad Sadiq¹

¹ 深圳信息职业技术学院软件学院，广东深圳 518172
² 南洋理工大学电子与电气工程学院，新加坡 639798

修回日期:2023-12-11 出版日期:2023-12-01 发布日期:2023-12-01
作者简介:梁俊威（1992- ），男，广东深圳人，博士，深圳信息职业技术学院讲师，主要研究方向为信息安全、人工智能、无线通信网络等
杨耿（1986- ），男，广西贵港人，博士，深圳信息职业技术学院高级工程师，主要研究方向为无线通信网络、模式识别等
马懋德（1964- ），男，加拿大人，博士，南洋理工大学教授、博士生导师，主要研究方向为无线通信网络、信息安全、人工智能等
Muhammad Sadiq（1982- ），男，巴基斯坦人，博士，深圳信息职业技术学院助理教授，主要研究方向为人工智能、模式识别、信息安全等
基金资助:
广东省青年创新人才基金资助项目(2022KQNCX233);公共大数据国家重点实验室基金资助项目(PBD2022-14);深圳市自然科学基金资助项目(20220820003203001)

Secure federated distillation GAN for CIDS in industrial CPS

Junwei LIANG¹, Geng YANG¹, Maode MA², Sadiq Muhammad¹

¹ College of Software Engineering, Shenzhen Institute of Information Technology, Shenzhen 518172, China
² School of Electronic and Electrical Engineering, Nanyang Technological University, Singapore 639798, Singapore

Revised:2023-12-11 Online:2023-12-01 Published:2023-12-01
Supported by:
The Guangdong Provincial Research Platform and Project(2022KQNCX233);The Foundation of State Key Laboratory of Public Big Data(PBD2022-14);The Shenzhen Natural Science Foundation(20220820003203001)

摘要/Abstract

摘要：

针对敏感信息保密必要性导致的数据孤岛问题，提出了一种适用于工业信息物理系统（CPS）的安全协作入侵检测系统（PFD-GAN）。具体来说，首先通过融入Wasserstein距离和标签条件，改进外部分类器生成对抗网络（EC-GAN），构建了一种新型半监督入侵检测模型，以产生能够实用的生成数据来增强分类性能。同时，在改进EC-GAN的训练中，融合本地差分隐私技术，防止敏感信息的泄露、保障协作过程的隐私安全。此外，设计了基于去中心化联邦蒸馏的协作方式，允许多个工业CPS共同构建一个综合的入侵检测系统，以识别整个网络系统下的威胁，而无须共享统一的模板模型。对真实工业CPS数据集的实验评估和理论分析表明，PFD-GAN可以在免受隐私泄露风险的同时，高效地检测针对工业CPS的各种类型攻击。

关键词: 入侵检测系统, 信息物理系统, 生成对抗网络, 本地差分隐私, 去中心化联邦蒸馏

Abstract:

Aiming at the data island problem caused by the imperativeness of confidentiality of sensitive information, a secure and collaborative intrusion detection system (CIDS) for industrial cyber physical systems (CPS) was proposed, called PFD-GAN.Specifically, a novel semi-supervised intrusion detection model was firstly developed by improving external classifier-generative adversarial network (EC-GAN) with Wasserstein distance and label condition, to strengthen the classification performance through the use of synthetic data.Furthermore, local differential privacy (LDP) technology was incorporated into the training process of developed EC-GAN to prevent sensitive information leakage and ensure privacy and security in collaboration.Moreover, a decentralized federated distillation (DFD)-based collaboration was designed, allowing multiple industrial CPS to collectively build a comprehensive intrusion detection system (IDS) to recognize the threats under the entire cyber systems without sharing a uniform template model.Experimental evaluation and theory analysis demonstrate that the proposed PFD-GAN is secure from the threats of privacy leaking and highly effective in detecting various types of attacks on industrial CPS.

Key words: intrusion detection system, cyber physical system, generative adversarial network, local differential privacy, decentralized federated distillation

中图分类号:

TP393.08

梁俊威, 杨耿, 马懋德, Muhammad Sadiq. 基于安全联邦蒸馏GAN的工业CPS协作入侵检测系统[J]. 通信学报, 2023, 44(12): 230-244.

Junwei LIANG, Geng YANG, Maode MA, Sadiq Muhammad. Secure federated distillation GAN for CIDS in industrial CPS[J]. Journal on Communications, 2023, 44(12): 230-244.

图/表 13

图1

图2

图3

表1

仿真实验参数"

参数名	参数值
潜在空间纬度dim(z)/维	24
批量采样数B/个	$B = \frac{\| S \|}{T}$
学习率 $α_{φ / ω / θ}$	5×10^-4
裁剪阈值 $C_{ω}, C_{θ}$	0.01
本地迭代次数 $T_{[φ, ω, θ]}$ /次	[50,5,100]
全局迭代次数T /次	$T = T_{φ} = \frac{T_{θ}}{2}$
无监督损失权重λ	0.1
伪标签阈值τ	0.7
LDP噪声尺度(ε, )δ	(6,1×10^-5)

表1

表2

图4

图5

图6

图7

表3

表4

图8

图9

参考文献 24

[1]	PIVOTO D G , ALMEIDA L F , RIGHI R ,et al. Cyber-physical systems architectures for industrial Internet of things applications in industry 4.0:a literature review[J]. Journal of Manufacturing Systems, 2021,58: 176-192.
[2]	SALAU B A , RAWAL A , RAWAT D B . Recent advances in artificial intelligence for wireless Internet of things and cyber–physical systems:a comprehensive survey[J]. IEEE Internet of Things Journal, 2022,9(15): 12916-12930.
[3]	吕思才, 张格, 张耀方 ,等. 一种面向工控系统的PU学习入侵检测方法[J]. 信息安全学报, 2021,6(4): 72-89.
	LV S C , ZHANG G , ZHANG Y F ,et al. A PU learning intrusion detection method for industrial control system[J]. Journal of Cyber Security, 2021,6(4): 72-89.
[4]	NESPOLI P , DíAZ-LóPEZ D , MáRMOL F G . Cyberprotection in IoT environments:a dynamic rule-based solution to defend smart devices[J]. Journal of Information Security and Applications, 2021,60:102878.
[5]	RAJASOUNDARAN S , PRABU A V , SUBRAHMANYAM J B V ,et al. WITHDRAWN:secure watchdog selection using intelligent key management in wireless sensor networks[J]. Materials Today:Proceedings, 2021:doi.org/10.1016/j.matpr.2020.12.1027.
[6]	NAQASH T , TANVEER M H , SHAH S H ,et al. Statistical analysis-based intrusion detection for software defined network[C]// Smart Trends in Computing and Communications. Singapore:Springer, 2022: 279-289.
[7]	IERACITANO C , ADEEL A , MORABITO F C ,et al. A novel statistical analysis and autoencoder driven intelligent intrusion detection approach[J]. Neurocomputing, 2020,387: 51-62.
[8]	ALDRIBI A , TRAORé I , MOA B ,et al. Hypervisor-based cloud intrusion detection through online multivariate statistical change tracking[J]. Computers ＆ Security, 2020,88:101646.
[9]	YAO R Z , WANG N , LIU Z H ,et al. Intrusion detection system in the advanced metering infrastructure:a cross-layer feature-fusion CNN-LSTM-based approach[J]. Sensors, 2021,21(2): 626.
[10]	CHADZA T , KYRIAKOPOULOS K G , LAMBOTHARAN S . Analysis of hidden Markov model learning algorithms for the detection and prediction of multi-stage network attacks[J]. Future Generation Computer Systems, 2020,108: 636-649.
[11]	LI W , MENG W . Collaborative intrusion detection in the era of IoT:recent advances and challenges[J]. Security and Privacy in the Internet of Things:Architectures,Techniques,and Applications, 2021: 123-149.
[12]	SHU J G , ZHOU L , ZHANG W Z ,et al. Collaborative intrusion detection for VANETs:a deep learning-based distributed SDN approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2021,22(7): 4519-4530.
[13]	RUZAFA-ALCáZAR P , FERNáNDEZ-SAURA P , MáRMOL-CAMPOS E ,et al. Intrusion detection based on privacy-preserving federated learning for the industrial IoT[J]. IEEE Transactions on Industrial Informatics, 2021,19(2): 1145-1154.
[14]	YU P Q , WYNTER L , LIM S H . Fed+:a family of fusion algorithms for federated learning[J]. arXiv Preprint,arXiv:2009.06303, 2020.
[15]	ZHANG Z , ZHANG Y , GUO D ,et al. SecFedNIDS:robust defense for poisoning attack against federated learning-based network intrusion detection system[J]. Future Generation Computer Systems, 2022,134: 154-169.
[16]	ABDEL-BASSET M , MOUSTAFA N , HAWASH H ,et al. Federated intrusion detection in blockchain-based smart transportation systems[J]. IEEE Transactions on Intelligent Transportation Systems, 2022,23(3): 2523-2537.
[17]	SHENDE O , PATERIYA R K , VERMA P ,et al. CEBM:collaborative ensemble blockchain model for intrusion detection in IoT environment[J]. Europe PMC, 2021:doi.org/10.21203/rs.3.rs-702181/v1.
[18]	AGRAWAL S , SARKAR S , AOUEDI O ,et al. Federated Learning for intrusion detection system:concepts,challenges and future directions[J]. Computer Communications, 2022,195: 346-361.
[19]	MORRIS T , GAO W . Industrial control system traffic data sets for intrusion detection research[C]// International Conference on Critical Infrastructure Protection. Berlin:Springer, 2014: 65-78.
[20]	CHATZOGLOU E , KAMBOURAKIS G , KOLIAS C . Empirical evaluation of attacks against IEEE 802.11 enterprise networks:the AWID3 dataset[J]. IEEE Access, 2021,9: 34188-34205.
[21]	XU C G , REN J , ZHANG D Y ,et al. GANobfuscator:mitigating information leakage under GAN via differential privacy[J]. IEEE Transactions on Information Forensics and Security, 2019,14(9): 2358-2371.
[22]	ZHAO R J , WANG Y J , XUE Z ,et al. Semisupervised federated-learning-based intrusion detection method for Internet of things[J]. IEEE Internet of Things Journal, 2023,10(10): 8645-8657.
[23]	IDRISSI M J , ALAMI H , MAHDAOUY A E ,et al. Fed-ANIDS:federated learning for anomaly-based network intrusion detection systems[J]. Expert Systems With Applications, 2023,234:121000.
[24]	PHAN N , WU X T , HU H ,et al. Adaptive laplace mechanism:differential privacy preservation in deep learning[C]// Proceedings of 2017 IEEE International Conference on Data Mining (ICDM). Piscataway:IEEE Press, 2017: 385-394.

IDS模型		Precision	Recall	F1值	FAR	FLOPS/(flop·s^-1)	Latency/ms
不使用LDP技术	CNN	95.62%	94.56%	94.92%	5.41%	89 216	4.29
	EC-GAN	96.55%	95.89%	96.11%	3.93%	89 216	4.20
	Dev EC-GAN	97.13%	96.76%	96.81%	2.20%	89 216	4.26
	Fix PFD-GAN	98.65%	98.62%	98.60%	1.14%	89 216	4.31
	PFD-GAN	99.00%	98.99%	98.97%	0.79%	74 470	3.58
使用LDP技术	CNN	93.76%	92.77%	93.00%	7.85%	89 216	4.36
	EC-GAN	95.63%	94.14%	94.68%	4.57%	89 216	4.34
	Dev EC-GAN	96.41%	95.60%	95.82%	3.03%	89 216	4.38
	Fix PFD-GAN	97.94%	97.75%	97.77%	1.54%	89 216	4.29
	PFD-GAN	98.16%	97.97%	97.99%	1.47%	74 470	3.49

IDS模型	J_normal	Jf_looding	J_{impersonation}	J_injection	FLOPS/(flop·s^-1)	Latency/ms
DL-GAN	96.80%	54.28%	96.39%	95.65%	5.22×10 ⁵	25.12
MD-BiGAN	98.11%	74.35%	82.97%	96.62%	5.14×10 ⁵	24.73
Fed-ANIDS	98.33%	71.55%	80.58%	96.84%	2.51×10 ⁶	120.97
SSFL	98.70%	83.33%	91.21%	96.69%	2.48×10 ⁶	119.40
PFD-GAN	99.30%	87.94%	97.11%	97.02%	4.77×10 ⁵	22.93

IDS解决方案		Precision	Recall	F1值	FAR	Latency/ms
无任何辅助机制的IDS	GHSOM	96.17%	96.11%	95.99%	5.48%	3.40
	Hypothesis Testing	94.28%	94.00%	93.96%	7.15%	0.43
	Dolphine + SVM	95.99%	95.97%	95.87%	5.59%	1.44
	Markov Model	91.33%	93.17%	92.00%	10.40%	1.56
带辅助机制的IDS	GHSOM + MOEA	97.43%	97.03%	97.01%	2.70%	2.03
	t-test + Bayesian	96.46%	96.18%	96.08%	5.42%	0.57
	PFD-GAN w/o LDP	98.65%	98.62%	98.60%	1.14%	3.52
差分隐私保护的联邦IDS	FedAvg	93.20%	92.95%	92.82%	7.96%	5.74
	Fed+	93.49%	93.05%	93.09%	7.68%	5.89
	SecFedNIDS	95.89%	94.82%	95.06%	3.65%	4.28
	PFD-GAN w LDP	97.94%	97.75%	97.77%	1.54%	3.65

基于安全联邦蒸馏GAN的工业CPS协作入侵检测系统

Secure federated distillation GAN for CIDS in industrial CPS

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 24

相关文章 15

Metrics

推荐阅读 0

[1]	陈晓霖, 昝道广, 吴炳潮, 关贝, 王永吉. 面向纵向联邦学习的对抗样本生成算法[J]. 通信学报, 2023, 44(8): 1-13.
[2]	苏新, 田天, Ziyang Gong, 周一青. 基于异常行为的海洋气象传感网的入侵检测方法研究[J]. 通信学报, 2023, 44(7): 86-99.
[3]	张佳乐, 朱诚诚, 孙小兵, 陈兵. 基于GAN的联邦学习成员推理攻击与防御方法[J]. 通信学报, 2023, 44(5): 193-205.
[4]	苏新, 张桂福, 行鸿彦, Zenghui Wang. 基于平衡生成对抗网络的海洋气象传感网入侵检测研究[J]. 通信学报, 2023, 44(4): 124-136.
[5]	于季弘, 林子砚, 叶能, 杨凯, 安建平. 基于三方生成对抗网络的隐蔽通信方法[J]. 通信学报, 2023, 44(11): 225-236.
[6]	王延文, 雷为民, 张伟, 孟欢, 陈新怡, 叶文慧, 景庆阳. 基于生成模型的视频图像重建方法综述[J]. 通信学报, 2022, 43(9): 194-208.
[7]	刘延华, 李嘉琪, 欧振贵, 高晓玲, 刘西蒙, MENG Weizhi, 刘宝旭. 对抗训练驱动的恶意代码检测增强方法[J]. 通信学报, 2022, 43(9): 169-180.
[8]	李昂, 陈建新, 魏昕, 周亮. 面向6G的跨模态信号重建技术[J]. 通信学报, 2022, 43(6): 28-40.
[9]	向夏雨, 王佳慧, 王子睿, 段少明, 潘鹤中, 庄荣飞, 韩培义, 刘川意. 基于生成对抗网络技术的医疗仿真数据生成方法[J]. 通信学报, 2022, 43(3): 211-224.
[10]	段雪源, 付钰, 王坤. 基于VAE-WGAN的多维时间序列异常检测方法[J]. 通信学报, 2022, 43(3): 1-13.
[11]	刘威, 陈成, 江锐, 卢涛. 四通道无监督学习图像去雾网络[J]. 通信学报, 2022, 43(10): 210-222.
[12]	陆彦辉, 柳寒, 李航, 朱光旭. 基于多鉴别器生成对抗网络的时间序列生成模型[J]. 通信学报, 2022, 43(10): 167-176.
[13]	周志立, 王美民, 杨高波, 朱剑宇, 孙星明. 基于轮廓自动生成的构造式图像隐写方法[J]. 通信学报, 2021, 42(9): 144-154.
[14]	邹福泰, 谭越, 王林, 蒋永康. 基于生成对抗网络的僵尸网络检测[J]. 通信学报, 2021, 42(7): 95-106.
[15]	王洪雁, 杨晓, 姜艳超, 汪祖民. 基于多通道GAN的图像去噪算法[J]. 通信学报, 2021, 42(3): 229-237.