基于GAF-DenseNet的航空发动机虚假数据注入攻击检测

黄鹏程; 陈丽丹; 祁恬; 张哲; 马永良; 高明

doi:10.13224/j.cnki.jasp.20220627

基于GAF-DenseNet的航空发动机虚假数据注入攻击检测

doi: 10.13224/j.cnki.jasp.20220627

黄鹏程¹,
陈丽丹^2,*, ,,
祁恬³,
张哲⁴,
马永良⁵,
高明⁵

1.
昆明理工大学信息工程与自动化学院，昆明 650504
2.
广州航海学院轮机工程学院，广州 510725
3.
华南理工大学自动化学院，广州 510640
4.
人工智能与数字经济广东省实验室，广州 510335
5.
广州城市理工学院电气工程学院，广州 510800

基金项目: 智能电网四川省重点实验室开放基金（2021-IEPGKLSP-KFZA01）

详细信息

作者简介:
黄鹏程（1999-），男，硕士生，主要研究方向为深空探测中的小行星着陆图像分析

通讯作者:
陈丽丹(1981-)，女，副教授，博士，主要研究方向为电动汽车与交通、能源网交互、智能技术应用。E-mail：chenld@gcu.edu.cn

中图分类号: V263.6；TP18
计量
- 文章访问数: 662
- HTML浏览量: 395
- PDF量: 72
- 被引次数: 0
出版历程
- 收稿日期: 2022-08-28
- 网络出版日期: 2023-06-01

False data injection attacks detection for aero engine system based on GAF-DenseNet

1.
Faculty of Information Engineering and Automation， Kunming University of Science and Technology，Kunming 650504，China
2.
Faculty of Marine Engineering，Guangzhou Maritime University，Guangzhou 510725，China
3.
School of Automation，South China University of Technology，Guangzhou 510640，China
4.
Guangdong Artificial Intelligence and Digital Economy Laboratory，Guangzhou 510335，China
5.
Department of Electrical Engineering， Guangzhou City University of Technology，Guangzhou 510800，China

摘要

摘要:
提出一种基于格拉姆角场（Gramian angular field, GAF）和密集连接卷积网络（densely connected convolutional networks，DenseNet）的航空发动机系统遭受虚假数据注入攻击的机器学习检测方法。首先，基于美国国家航空和宇宙航行局的商用模块化航空推进系统仿真数据集（commercial modular aero-propulsion system simulation，C-MAPSS），构建了连续和间隔虚假数据注入两种攻击模型；其次，通过GAF算法，在保留原始航空发动机传感器获得的时序信号的时间依赖性的前提下，对时间序列数据进行唯一编码，并设计了DenseNet-121网络对图像阵列中内含的传感器信息进行深层挖掘，进而检测航空发动机是否遭受虚假数据注入攻击及攻击类型识别；最后，融合GAF-DenseNet方法在T24、T50和P30传感器上的平均分类准确率为98.46%，与长短期记忆、门控循环单元和卷积神经网络对比分别提高了1.91%、3.82%和0.38%。
- 航空发动机 /
- 商用模块化航空推进系统仿真数据（C-MAPSS） /
- 虚假数据注入攻击 /
- 格拉姆角场（GAF） /
- 密集连接卷积网络（DenseNet）
Abstract:
A machine learning detection method for aero-engine system false data injection attacks based on Gramian angular field (GAF) and densely connected convolutional networks (DenseNet) was proposed. Firstly, two attack models of continuous and interval spurious data injection were constructed based on the simulation dataset of NASA’s commercial modular aero-propulsion system simulation (C-MAPSS). Secondly, the GAF method was proposed to transform the timing signal obtained by the aero-engine sensors into the image signal, and a DenseNet-121 network was designed to detect whether the aero engine was subject to false data injection attack and the type of attack was identified. Finally, the average classification accuracy of GAF-DenseNet method on T24, T50, and P30 sensors was 98.46%, which was 1.91%, 3.82%, and 0.38% better compared with long and short-term memory, gated recurrent units, and convolutional neural networks, respectively.

HTML

图 1 航空发动机预测性维护系统

Figure 1. Aero engine health monitoring system architecture

下载: 全尺寸图片幻灯片

图 2 航空发动机FDIA示意图

Figure 2. Schematic diagram of FDIA of aero engine

下载: 全尺寸图片幻灯片

图 3 连续FDIA

Figure 3. Continuous FDIA

下载: 全尺寸图片幻灯片

图 4 间隔FDIA

Figure 4. Intermittent FDIA

下载: 全尺寸图片幻灯片

图 5 格拉姆数据变换

Figure 5. Data transformation based on GAF

下载: 全尺寸图片幻灯片

图 6 DenseNet的网络结构

Figure 6. Structure of DenseNet

下载: 全尺寸图片幻灯片

图 7 GAF-DenseNet 模型整体结构

Figure 7. Overall structure of GAF-DenseNet model

下载: 全尺寸图片幻灯片

图 8 经标准化处理的原始训练集数据

Figure 8. Standardized processing of original training set data

下载: 全尺寸图片幻灯片

图 9 传感器T24原始数据、受攻击数据图像化

Figure 9. Sensor T24 raw data, attacked data visualization

下载: 全尺寸图片幻灯片

图 10 T24的LSTM、GRU和CNN结果

Figure 10. Results for T24 based on LSTM, GRU and CNN

下载: 全尺寸图片幻灯片

图 11 T50的LSTM、GRU和CNN结果

Figure 11. Results for T50 based on LSTM, GRU and CNN

下载: 全尺寸图片幻灯片

图 12 P30的LSTM、GRU和CNN结果

Figure 12. Results for P30 based on LSTM, GRU and CNN

下载: 全尺寸图片幻灯片

图 13 T24传感器测试集分类预测图

Figure 13. T24 sensor test set classification prediction chart

下载: 全尺寸图片幻灯片

表 1 C-MAPSS数据集示例

Table 1. Samples of dataset of C-MAPSS

发动机ID	运行周期	第1列	第2列	···	第26列
1	1	−0.0007	−0.0004	···	23.4190
1	2	0.0019	−0.0003	···	23.4236
1	3	−0.0043	0.0003	···	23.3442
$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $
1	192	0.0009	0	···	22.9649
2	1	−0.0018	0.0006	···	23.4585
2	2	0.0043	−0.0003	···	23.4085
$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $
2	287	−0.0005	0.0006	···	23.0848
3	1	0.0008	0.0005	···	23.3205
$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $
100	200	−0.0032	−0.0005	···	23.0522

下载: 导出CSV

表 2 DenseNet结构层次

Table 2. Proposed network structure of DenseNet

层类型（layers）	输出结构（output size）	DenseNet-121
卷积（convolution）	112×112	7×7 conv，stride2
池化（pooling）	56×56	3×3 max pool，stride2
密集块1 （dense block_1）	56×56	$\left[ {\begin{array}{*{20}{l} } {1 \times 1{\text{ conv } } } \\ {3 \times 3{\text{ conv } } } \end{array} } \right] \times 6$
过渡层1 （transition layer_1）	56×56	1×1 conv
过渡层1 （transition layer_1）	28×28	2×2 average pool，stride2
密集块2 （dense block_2）	28×28	$\left[ {\begin{array}{*{20}{l} } {1 \times 1{\text{ conv } } } \\ {3 \times 3{\text{ conv } } } \end{array} } \right] \times 12$
过渡层2 （transition layer_2）	28×28	1×1 conv
过渡层2 （transition layer_2）	14×14	2×2 average pool，stride2
密集块3 （dense block_3）	14×14	$\left[ {\begin{array}{*{20}{l} } {1 \times 1{\text{ conv } } } \\ {3 \times 3{\text{ conv } } } \end{array} } \right] \times 24$
过渡层3 （transition layer_3）	14×14	1×1 conv
过渡层3 （transition layer_3）	7×7	2×2 average pool，stride2
密集块4 （dense block_4）	7×7	$\left[ {\begin{array}{*{20}{l} } {1 \times 1{\text{ conv } } } \\ {3 \times 3{\text{ conv } } } \end{array} } \right] \times 16$

下载: 导出CSV

表 3 LSTM、GRU和CNN网络结构

Table 3. Network structure of LSTM, GRU and CNN

网络	隐藏层数	每层的节点数	随机失活比例	批量大小	激活函数
LSTM	4	100	0.2	48	tanh
GRU	4	100	0.2	48	tanh
CNN	4	64	0.2	48	ReLu

下载: 导出CSV

表 4 GAF-DenseNet分类结果

Table 4. Results of GAF-DenseNet

传感器	GAF方法	A:未受FDIA	B:连续FDIA	C:间隔FDIA	正确率/%	BCEloss
T24	GASF	全对	全对	错分2个到B	97.70	0.148
T24	GADF	全对	错分1个到C	错分1个到B	97.70	0.042
T50	GASF	全对	全对	错分1个到B	98.85	0.008
T50	GADF	全对	全对	错分1个到B	98.85	0.006
P30	GASF	全对	全对	错分1个到B	98.85	0.024
P30	GADF	全对	错分1个到C	全对	98.85	0.128

下载: 导出CSV

参考文献(23)

[1]	刘畅,陈雯柏. 一种基于MSDCNN-LSTM的设备RUL预测方法[J]. 西北工业大学学报,2021,39(2): 407-413. doi: 10.1051/jnwpu/20213920407 LIU Chang,CHEN Wenbai. A RUL prediction method of equipments based on MSDCNN-LSTM[J]. Journal of Northwestern Polytechnical University,2021,39(2): 407-413. (in Chinese) doi: 10.1051/jnwpu/20213920407
[2]	刘伟民,胡忠志. 一种基于神经网络的航空发动机剩余寿命预估方法[J]. 航空发动机,2021,47(3): 8-15. LIU Weimin,HU Zhongzhi. An aeroengine remaining useful life prediction method based on neural network[J]. Aeroengine,2021,47(3): 8-15. (in Chinese)
[3]	车畅畅,王华伟,倪晓梅,等. 基于1D-CNN和Bi-LSTM的航空发动机剩余寿命预测[J]. 机械工程学报,2021,57(14): 304-312. doi: 10.3901/JME.2021.14.304 CHE Changchang,WANG Huawei,NI Xiaomei,et al. Residual life prediction of aeroengine based on 1D-CNN and Bi-LSTM[J]. Journal of Mechanical Engineering,2021,57(14): 304-312. (in Chinese) doi: 10.3901/JME.2021.14.304
[4]	ZHAO Z,HUANG Y,ZHEN Z,et al. Data-driven false data-injection attack design and detection in cyber-physical systems[J]. IEEE Transactions on Cybernetics,2021,51(12): 6179-6187. doi: 10.1109/TCYB.2020.2969320
[5]	苗青林,张晓丰,高杨军,等. 双通道深度卷积神经网络的航空发动机剩余使用寿命预测方法[J]. 空军工程大学学报(自然科学版),2022,23(2): 12-18. MIAO Qinglin,ZHANG Xiaofeng,GAO Yangjun,et al. A method of predicting aero-engines remaining useful life based on two-channel deep convolutional neural network[J]. Journal of Air Force Engineering University (Natural Science Edition),2022,23(2): 12-18. (in Chinese)
[6]	项德威,郑前钢,张海波,等. 基于NN-PSM的航空发动机机载自适应稳态模型[J]. 航空动力学报,2022,37(2): 409-423. XIANG Dewei,ZHENG Qiangang,ZHANG Haibo,et al. Aero engine on board adaptive steady state model base on NN-PSM[J]. Journal of Aerospace Power,2022,37(2): 409-423. (in Chinese)
[7]	史国华. 数据驱动的航空发动机RUL预测模型研究[D]. 辽宁大连: 大连理工大学, 2021. SHI Guohua. Research on data-driven aero-engine RUL prediction model[D]. Dalian Liaoning: Dalian University of Technology, 2021. (in Chinese)
[8]	陈保家,郭凯敏,陈法法,等. 基于残差NLSTM网络和注意力机制的航空发动机剩余使用寿命预测[J]. 航空动力学报,2023,38(5): 1176-1184. CHEN Baojia,GUO Kaimin,CHEN Fafa,et al. Prediction of remaining useful life of aero-engine based on residual NLSTM neural network and attention mechanism[J]. Journal of Aerospace Power,2023,38(5): 1176-1184. (in Chinese)
[9]	蔺瑞管,王华伟,车畅畅,等. 基于LSTM分类器的航空发动机预测性维护模型[J]. 系统工程与电子技术,2022,44(3): 1052-1059. LIN Ruiguan,WANG Huawei,CHE Changchang,et al. Predictive maintenance model of areo engine based on LSTM classifier[J]. Systems Engineering and Electronics,2022,44(3): 1052-1059. (in Chinese)
[10]	MUSLEH A S,CHEN G,DONG Z Y. A survey on the detection algorithms for false data injection attacks in smart grids[J]. IEEE Transactions on Smart Grid,2020,11(3): 2218-2234. doi: 10.1109/TSG.2019.2949998
[11]	LIU J,LABEAU F. Detection of false data injection attacks in industrial wireless sensor networks exploiting network numerical sparsity[J]. IEEE Transactions on Signal and Information Processing over Networks,2021,7: 676-688. doi: 10.1109/TSIPN.2021.3122289
[12]	BRILAND M,BOUQUET F. An approach for testing false data injection attack on data dependent industrial devices[J]. Journal of Universal Computer Science,2020,27(7): 774-792.
[13]	MODE G R, CALYAM P, HOQUE K A. Impact of false data injection attacks on deep learning enabled predictive analytics[R]. Budapest, Hungary: 2020 IEEE/IFIP Network Operations and Management Symposium, 2020.
[14]	王文益,古亭亭. 基于1DCNN-BiLSTM网络的ADS-B欺骗式干扰检测[J]. 信号处理,2021,37(6): 984-990. WANG Wenyi,GU Tingting. ADS-B deceptive jamming detection based on 1DCNN-BiLSTM network[J]. Journal of Signal Processing,2021,37(6): 984-990. (in Chinese)
[15]	李妮锶. 基于数据驱动的ADS-B异常检测算法研究[D]. 四川广汉: 中国民用航空飞行学院, 2022. LI Nisi. The research on ADS-B anomaly detection algorithm based on data drive[D]. Guanghan Sichuan: Civil Aviation Flight University of China, 2022. (in Chinese)
[16]	CRETIN A, VERNOTTE A, CHEVROT A, et al. Test data generation for false data injection attack testing in air traffic surveillance[R]. Porto, Portugal: 2020 IEEE International Conference on Software Testing, Verification and Validation Workshops, 2020.
[17]	张公. 基于TrellisNet的航空发动机剩余使用寿命研究[D]. 包头: 内蒙古科技大学, 2021. ZHANG Gong. Research on remaining service life of aeroengine based on TrellisNet[D]. Baotou: Inner Mongolia University of Science and Technology, 2021. (in Chinese)
[18]	YANG H,DING K,QIU R C,et al. Remaining useful life prediction based on normalizing flow embedded sequence-to-sequence learning[J]. IEEE Transactions on Reliability,2021,70(4): 1342-1354. doi: 10.1109/TR.2020.3010970
[19]	FREDERICK D K, DECASTRO J A, LITT J S. User’s guide for the commercial modular aero-propulsion system simulation (C-MAPSS)[R]. NASA/TM-2007-215026, 2007.
[20]	贾驰千,冯冬芹. 基于多目标决策的工控系统设备安全评估方法研究[J]. 自动化学报,2016,42(5): 706-714. JIA Chiqian,FENG Dongqin. Industrial control system devices security assessment with multi-objective decision[J]. Acta Automatica Sinica,2016,42(5): 706-714. (in Chinese)
[21]	PAN S J,YANG Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering,2010,22(10): 1345-1359. doi: 10.1109/TKDE.2009.191
[22]	郑炜,林瑞全,王俊,等. 基于GAF与卷积神经网络的电能质量扰动分类[J]. 电力系统保护与控制,2021,49(11): 97-104. ZHENG Wei,LIN Ruiquan,WANG Jun,et al. Power quality disturbance classification based on GAF and a convolutional neural network[J]. Power System Protection and Control,2021,49(11): 97-104. (in Chinese)
[23]	HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[R]. Honolulu, US: 2017 IEEE Conference on Computer Vision and Pattern Recognition, 2017.