基于Transformer-VAE的ADS-B异常检测方法

doi:10.12305/j.issn.1001-506X.2023.11.37

系统工程与电子技术 ›› 2023, Vol. 45 ›› Issue (11): 3680-3689.doi: 10.12305/j.issn.1001-506X.2023.11.37

基于Transformer-VAE的ADS-B异常检测方法

丁建立¹, 张琪琪¹, 王静²^,³^,*, 霍纬纲¹

1. 中国民航大学计算机科学与技术学院, 天津 300300
2. 中国民航大学信息安全测评中心, 天津 300300
3. 中国民航大学安全科学与工程学院, 天津 300300

收稿日期:2022-10-14 出版日期:2023-10-25 发布日期:2023-10-31
通讯作者: 王静
作者简介:丁建立(1963—), 男, 教授, 博士, 主要研究方向为民航信息系统主动容灾技术、大数据民航应用
张琪琪(1998—), 女, 硕士研究生, 主要研究方向为民航网络安全
王静(1980—), 女, 副教授, 博士, 主要研究方向为民航网络安全、大数据民航应用
霍纬纲(1978—), 男, 教授, 博士, 主要研究方向为时序大数据分析与挖掘
基金资助:
国家自然科学基金民航联合基金(U2033205);中国民航大学信息安全测评中心开放基金(ISECCA-202006)

ADS-B anomaly detection method based on Transformer-VAE

Jianli DING¹, Qiqi ZHANG¹, Jing WANG²^,³^,*, Weigang HUO¹

1. College of Computer Science and Technology, Civil Aviation University of China, Tianjin 300300, China
2. Information Security Evaluation Center of Civil Aviation, Civil Aviation University of China, Tianjin 300300, China
3. College of Safety Science and Engineering, Civil Aviation University of China, Tianjin 300300, China

Received:2022-10-14 Online:2023-10-25 Published:2023-10-31
Contact: Jing WANG

摘要/Abstract

摘要：

针对现有的异常检测算法无法同时捕捉广播式自动相关监视(automatic dependent surveillance-broadcast, ADS-B)报文序列中的长程依赖关系、随机性以及整体性特征的问题, 提出了基于Transformer变分自动编码器(variational autoencoder, VAE)的ADS-B异常检测算法, 以Transformer编码器和解码器作为VAE的推断网络和生成网络来学习ADS-B报文序列的时序分布。由Transformer编码器的自注意力机制建模序列中的长程依赖关系, 生成反映报文序列整体性特征的随机变量, 并采用一个特殊的融合模块融合该随机变量与Transformer解码器的输出以实现报文的重构, 最终通过重构误差来检测异常。实验表明, 所提出的方法在不同攻击场景下, 均优于相关基线算法。

关键词: 广播式自动相关监视, 异常检测, 变分自动编码器, 深度学习

Abstract:

Aiming at the problem that the existing anomaly detection algorithm cannot simultaneously capture the long-range dependency, randomness and overall characteristics in the automatic dependent surveillance-broadcast (ADS-B) message sequence, an anomaly detection algorithm for ADS-B based on variational autoencoder (VAE) and Transformer is proposed. The Transformer encoder and decoder are used as the inference network and generation network of VAE to learn the timing distribution of ADS-B packet sequence. The long-range dependence relationship in ADS-B packet sequence is modeled by the self-attention mechanism of Transformer encoder, and random variables reflecting the global information of ADS-B packet sequence are generated. A special fusion module is used to fuse the global random variables with the output of Transformer decoder to realize the reconstruction of ADS-B packets. Finally, anomaly is detected through reconstruction error. Experiments show that the method outperforms the relevant baseline algorithms under different attack scenarios.

Key words: automatic dependent surveillance-broadcast (ADS-B), anomaly detection, variational autoencoder, deep learning

中图分类号:

V221+.3

丁建立, 张琪琪, 王静, 霍纬纲. 基于Transformer-VAE的ADS-B异常检测方法[J]. 系统工程与电子技术, 2023, 45(11): 3680-3689.

Jianli DING, Qiqi ZHANG, Jing WANG, Weigang HUO. ADS-B anomaly detection method based on Transformer-VAE[J]. Systems Engineering and Electronics, 2023, 45(11): 3680-3689.

图/表 11

图1

图2

图3

图4

图5

图6

图7

表1

表2

表3

图8

参考文献 24

1	WANDELT S , SUN X , FRICKE H . ADS-BI: compressed indexing of ADS-B data[J]. IEEE Trans.on Intelligent Transportation Systems, 2018, 19 (12): 3795- 3806. doi: 10.1109/TITS.2017.2784371
2	KHANDKER S , TURTIAINEN H , COSTIN A , et al. On the (in) security of 1090ES and UAT978 mobile cockpit information systems-an attacker perspective on the availability of ADS-B safety-and mission-critical systems[J]. IEEE Access, 2022, 10, 37718- 37730. doi: 10.1109/ACCESS.2022.3164704
3	COSTIN A, FRANCILLON A. Ghost in the air (traffic): on insecurity of ADS-B protocol and practical attacks on ADS-B devices[C]//Proc. of the Black Hat USA, 2012.
4	FENG Z L, PAN W J, WANG Y. A data authen-tication solution of ADS-B system based on X. 509 certificate[C]//Proc. of the 27th International Congress of the Aeronautical Sciences, 2010.
5	STROHMEIER M, LENDERS V, MARTINOVIC I. Lightweight location verification in air traffic surveillance networks[C]//Proc. of the ACM Workshop on Cyber-Physical System Security, 2015: 49-60.
6	陈蕾, 吴仁彪, 卢丹. 利用多普勒效应的ADS-B欺骗式干扰检测方法[J]. 信号处理, 2018, 34 (6): 722- 728.
	CHEN L , WU R B , LU D . ADS-B spoofing detection method using doppler effect[J]. Journal of Signal Processing, 2018, 34 (6): 722- 728.
7	HABLER E , SHABTAI A . Using LSTM encoder-decoder algorithm for detecting anomalous ADS-B messages[J]. Computers & Security, 2018, 78, 155- 173. doi: 10.3969/j.issn.1001-3695.2018.01.032
8	丁建立, 邹云开, 王静, 等. 基于深度学习的ADS-B异常数据检测模型[J]. 航空学报, 2019, 40 (12): 162- 172.
	DING J L , ZOU Y K , WANG J , et al. ADS-B anomaly data detection model based on deep learning[J]. Acta Aeronautica et Astronautica Sinica, 2019, 40 (12): 162- 172.
9	罗鹏, 王布宏, 李腾耀. 基于BiGRU-SVDD的ADS-B异常数据检测模型[J]. 航空学报, 2020, 41 (10): 323878.
	LUO P , WANG B H , LI T Y . ADS-B anomaly datadetection model based on BiGRU-SVDD[J]. Acta Aeronautica et Astronautica Sinica, 2020, 41 (10): 323878.
10	SU Y, ZHAO Y J, NIU C H, et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network[C]//Proc. of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019: 2828-2837.
11	FRACCARO M , SØNDERBY S K , PAQUET U , et al. Sequential neural models with stochastic layers[J]. Advances in Neural Information Processing Systems, 2016, 29, 2207- 2215.
12	LUO P , WANG B H , LI T Y , et al. ADS-B anomaly data detection model based on VAE-SVDD[J]. Computers & Security, 2021, 104, 102213.
13	VASWANI A , SHAZEER N , PARMAR N , et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30, 6000- 6010.
14	MENG H Y, ZHANG Y X, LI Y X, et al. Spacecraft anomaly detection via transformer reconstruction error[C]//Proc. of the International Conference on Aerospace System Science and Engineering, 2019: 351-362.
15	DUAN T. Unsupervised multivariate time series anomaly detection via transformer-based models and time series encoding[D]. Toronto: University of Toronto, 2021.
16	BOWMAN S R, VILNIS L, VINYALS O, et al. Generating sentences from a continuous space[C]//Proc. of the 20th SIGNLL Conference on Computational Natural Language Learning, 2015: 10-21.
17	BA J L, KIROS J R, HINTON G E. Layer normalization[EB/OL]. [2022-09-25]. https://arxiv.org/abs/1607.06450.
18	KINGMA D P, WELLING M. Auto-encoding variational bayes[EB/OL]. [2022-09-25]. https://arxiv.org/abs/1312.6114.
19	SETIAWAN H, SPERBER M, NALLASAMY U, et al. Variational neural machine translation with normalizing flows[C]//Proc. of the 58th Annual Meeting of the Association for Computational Linguistics, 2020: 7771-7777.
20	BENGIO Y , DUCHARME R , VINCENT P . A neural probabilistic language model[J]. Advances in Neural Information Processing Systems, 2000, 13, 893- 899.
21	KINGMA D P, BA J. Adam: A method for stochastic optimization[EB/OL]. [2022-10-08]. https://arxiv.org/abs/1412.6980.
22	LI K L, HUANG H K, TIAN S F, et al. Improving one-class SVM for anomaly detection[C]//Proc. of the International Conference on Machine Learning and Cybernetics, 2003: 3077-3081.
23	BREUNIG M M, KRIEGEL H P, Ng R T, et al. LOF: identifying density-based local outliers[C]//Proc. of the ACM SIGMOD International Conference on Management of Data, 2000: 93-104.
24	ZHANG K Z, KANG X D, LI S T. Isolation forest for anomaly detection in hyperspectral images[C]//Proc. of the IGARSS IEEE International Geoscience and Remote Sensing Symposium, 2019: 437-440.

模型	评估指标	随机噪声	幽灵飞机	航向偏移	速度偏移	DOS攻击
OC-SVM	精确率	79.23	55.19	76.20	80.46	00.00
	召回率	67.63	9.80	26.53	61.21	00.00
	F1分数	73.01	16.64	40.11	69.52	-
IF	精确率	66.40	48.97	26.72	27.56	00.00
	召回率	82.60	21.68	51.60	57.49	00.00
	F1分数	73.61	30.05	35.20	37.25	-
LOF	精确率	30.16	34.37	31.16	26.67	00.00
	召回率	87.25	92.55	91.03	91.32	00.00
	F1分数	44.74	50.13	46.43	41.28	-
LSTM-AE	精确率	92.01	77.75	79.67	83.81	82.58
	召回率	84.40	89.61	86.48	76.84	90.14
	F1分数	88.04	83.25	82.94	80.17	84.80
LSTM-VAE	精确率	92.99	79.32	80.76	86.50	87.92
	召回率	87.58	88.79	89.03	78.21	91.31
	F1分数	90.20	83.78	84.69	82.14	89.58
Transformer-VAE	精确率	98.78	91.02	94.91	91.77	91.45
	召回率	91.42	93.83	98.32	90.61	93.76
	F1分数	94.96	92.40	96.58	91.19	92.60

模型	精确率	召回率	F1分数
(1)	90.64	87.30	88.94
(2)	91.12	90.89	91.00
(3)	93.17	92.42	92.79

模型	评估指标	k
模型	评估指标	100	200	300	400	500
LSTM-VAE	精确率	88.64	88.89	93.05	90.23	90.90
	召回率	83.70	95.75	86.30	87.70	84.73
	F1分数	86.10	92.19	89.55	88.94	87.71
Transformer-VAE	精确率	96.40	92.93	95.94	92.97	93.45
	召回率	89.93	95.88	94.93	94.91	92.27
	F1分数	93.05	94.38	95.43	93.93	92.85

[1]	汪萌, 诸兵. 不确定性建模在2D和3D目标检测中的应用[J]. 系统工程与电子技术, 2023, 45(8): 2370-2376.
[2]	邵凯, 杜自群, 王光宇. 基于模型剪枝动态调整压缩率的CSI反馈方法[J]. 系统工程与电子技术, 2023, 45(8): 2615-2622.
[3]	崔天舒, 王栋, 黄振. 面向空间认知通信的轻量化网络自动调制分类方法[J]. 系统工程与电子技术, 2023, 45(7): 2220-2226.
[4]	姜雨, 袁琪, 胡志韬, 吴薇薇, 顾欣. 基于气象因素的机场进离港延误预测[J]. 系统工程与电子技术, 2023, 45(6): 1722-1731.
[5]	王文益, 钟仁伟. 基于交替方向乘子法的单天线ADS-B信号分离[J]. 系统工程与电子技术, 2023, 45(5): 1286-1296.
[6]	陈洋, 廖灿辉, 张锟, 刘建, 王鹏举. 基于自监督对比学习的信号调制识别算法[J]. 系统工程与电子技术, 2023, 45(4): 1200-1206.
[7]	张晔, 侯毅, 欧阳克威, 周石琳. 单变量序列数据分类方法综述[J]. 系统工程与电子技术, 2023, 45(2): 313-335.
[8]	邵正途, 许登荣, 徐文利, 王晗中. 基于LSTM和残差网络的雷达有源干扰识别[J]. 系统工程与电子技术, 2023, 45(2): 416-423.
[9]	陈任飞, 彭勇, 李忠文. 基于持续无监督域适应策略的水面漂浮物目标检测方法[J]. 系统工程与电子技术, 2023, 45(11): 3391-3401.
[10]	陶臣嵩, 陈思伟, 肖顺平. 基于深度学习模型的SAR图像间歇采样转发干扰检测[J]. 系统工程与电子技术, 2023, 45(11): 3465-3473.
[11]	何诗扬, 汪玲, 朱岱寅, 钱君. 基于时空记忆解耦RNN的雷暴预测方法[J]. 系统工程与电子技术, 2023, 45(11): 3474-3480.
[12]	李浩然, 熊伟, 崔亚奇. 基于深度特征融合的SAR图像与AIS信息关联方法[J]. 系统工程与电子技术, 2023, 45(11): 3491-3497.
[13]	刘子超, 王江, 何绍溟. 基于深度学习的时间角度控制制导律[J]. 系统工程与电子技术, 2023, 45(11): 3579-3587.
[14]	季策, 宋博翰, 耿蓉, 梁敏骏. 快时变信道下基于深度学习的OFDM系统信道估计[J]. 系统工程与电子技术, 2023, 45(11): 3649-3655.
[15]	宋婷, 兀泽朝, 高艾, 袁建平. 基于CycleGAN的月表图像数据增强方法[J]. 系统工程与电子技术, 2023, 45(10): 3041-3048.

基于Transformer-VAE的ADS-B异常检测方法

ADS-B anomaly detection method based on Transformer-VAE

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 24

相关文章 15

编辑推荐

Metrics

本文评价