基于注意力机制的GRU神经网络安全态势预测方法

doi:10.3969/j.issn.1001-506X.2021.01.32

摘要/Abstract

摘要：

传统的网络安全态势预测方法依赖于历史态势值的准确性,并且各种网络安全因素之间存在相关性和重要程度差异性。针对以上问题,提出一种基于注意力机制的循环门控单元(recurrent gate unit, GRU)编码预测方法,该方法利用GRU神经网络挖掘网络安全态势数据之间的时间相关性;引入注意力机制计算安全指标的分配权重并将其编码为网络安全态势值;利用改进的粒子群优化(particle swarm optimization, PSO)算法进行超参数寻优,以加速GRU神经网络的训练。仿真分析表明,所提方法具有更快的收敛速度和较低的复杂度,并且在不同的预测时长下具有较小的均方误差和平均绝对误差。

关键词: 网络安全态势预测, 注意力机制, 循环门控单元, 粒子群优化算法

Abstract:

Traditional network security situation prediction methods rely on the accuracy of the historical situation value, and there are differences in the correlation and importance between various network security factors. For the above mentioned problems, the recurrent gated unit (GRU) coding prediction method based on the attention mechanism is proposed. Attention mechanism is introduced to calculate the weight of the security index and it is coded as the network security situation value. The improved particle swarm optimization (PSO) algorithm is used to optimize the super parameters to accelerate the training of GRU neural network. Simulation results show that the proposed method has faster convergence speed and lower complexity, smaller mean square error (MSE) and mean absolute error (MAE) under different prediction time.

Key words: network security situation prediction, attention mechanism, recurrent gated unit(GRU), particle swarm optimization(PSO)algorithm

中图分类号:

TP393

何春蓉, 朱江. 基于注意力机制的GRU神经网络安全态势预测方法[J]. 系统工程与电子技术, 2021, 43(1): 258-266.

Chunrong HE, Jiang ZHU. Security situation prediction method of GRU neural network[J]. Systems Engineering and Electronics, 2021, 43(1): 258-266.

图/表 13

图1

图2

图3

图4

图5

图6

表1

表2

图7

图8

表3

图9

表4

参考文献 26

1	WEI X, JIANG X. Comprehensive analysis of network security situational awareness methods and models[C]//Proc.of the IEEE 2nd International Symposium on Instrumentation and Measurement, Sensor Network and Automation, 2013: 176-179.
2	SHANG L, ZHAO W, ZHANG J J, et al. Network security situation prediction based on long short-term memory network[C]//Proc.of the 20th Asia-Pacific Network Operations and Management Symposium, 2019.
3	LI J, DONG C B. Research on network security situation prediction-oriented adaptive learning neuron[C]//Proc.of the 2nd International Conference on Networks Security, Wireless Communications and Trusted Computing, 2010: 483-485.
4	HU J J , MA D Y , LIU C , et al. Network security situation prediction based on MR-SVM[J]. IEEE Access, 2019, 7, 130937- 130945.
5	HU G Y , QIAN P L . Cloud belief rule base model for network security situation prediction[J]. IEEE Communications Letters, 2016, 20 (5): 914- 917.
6	WANG B W , KONG W , GUAN H , et al. Air quality forecasting based on gated recurrent long short term memory model in internet of things[J]. IEEE Access, 2019, 7, 69524- 69534.
7	张任川, 张玉臣, 刘璟, 等. 应用改进卷积神经网络的网络安全态势预测方法[J]. 计算机工程与应用, 2019, 55 (6): 86- 93.
	ZHAN R C , ZHAN Y C , LIU J , et al. Network security situation prediction method using improved convolution neural network[J]. Computer Engineering and Applications, 2019, 55 (6): 86- 93.
8	HAN Z Z , SHANG M Y , LIU Z B , et al. SeqViews 2 SeqLabels: learning 3D global features via aggregating sequential views by RNN with attention[J]. IEEE Trans.on Image Processing, 2019, 28 (2): 658- 672.
9	ZHAO R , WANG D Z , YAN R Q , et al. Machine health monitoring using local feature-based gated recurrent unit networks[J]. IEEE Trans.on Industrial Electronics, 2018, 65 (2): 1539- 1548.
10	WANG B W , KONG W , GUAN H , et al. Air quality forecasting based on gated recurrent long short term memory model in internet of things[J]. IEEE Access, 2019, 7, 69524- 69534.
11	朱张莉, 饶元, 吴渊, 等. 注意力机制在深度学习中的研究进展[J]. 信息学报, 2019, 33 (6): 1- 11.
	ZHU Z L , RAO Y , WU Y , et al. Research progress of attention mechanism in deep learning[J]. Journal of Chinese Information Processing, 2019, 33 (6): 1- 11.
12	WANG Q R , YUAN C , WANG J D , et al. Learning attentional recurrent neural network for visual tracking[J]. IEEE Trans.on Multimedia, 2019, 21 (4): 930- 942.
13	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[EB/OL].[2017-06-12]http://arxiv.org/abs/1706.03762.
14	SRIVASTAVA D, SINGH Y, SAHOO A. Auto tuning of RNN hyper-parameters using cuckoo search algorithm[C]//Proc.of the 12th International Conference on Contemporary Computing, 2019.
15	GUO X, STEIN B V. A new approach towards the combined algorithm selection and hyper-parameter optimization problem[C]//Proc.of the IEEE Symposium Series on Computational Intelligence, 2019: 2042-2049.
16	HAN H G , LU W , HOU Y , et al. An adaptive-PSO-based self-organizing RBF neural network[J]. IEEE Trans.on Neural Networks and Learning Systems, 2018, 29 (1): 104- 117.
17	HAN H G , WU X L , ZHANG L , et al. Self-organizing RBF neural network using an adaptive gradient multi-objective particle swarm optimization[J]. IEEE Trans.on Cybernetics, 2019, 49 (1): 69- 82.
18	LIU W B , WANG Z D , YUAN Y , et al. A novel sigmoid-function-based adaptive weighted particle swarm optimizer[J]. IEEE Trans.on Cybernetics, 2019, 1- 10.
19	CHANG Z H, ZHANG Y, CHEN W B. Effective Adam-optimized LSTM neural network for electricity price forecasting[C]//Proc.of the IEEE 9th International Conference on Software Engineering and Service Science, 2018: 245-248.
20	BOCK S, WEIS M. A proof of local convergence for the Adam optimizer[C]//Proc.of the International Joint Conference on Neural Networks, 2019.
21	GUAN N Y, SHAN L, YANG C Q, et al. Delay compensated asynchronous Adam algorithm for deep neural networks[C]//Proc.of the IEEE International Symposium on Parallel and Distributed Processing with Applications and IEEE International Conference on Ubiquitous Computing and Communications, 2017: 852-859.
22	中国互联网信息中心: 2017至2019年网络安全信息与动态周报[EB/OL].[2019-12-08]. https://www.cert.org.cn/publish/main/index.html.
	CNCERT/CC.Weekly report on network security information from 2017 to 2019[EB/OL].[2019-12-08]. https://www.cert.org.cn/publish/main/index.html.
23	姜万菲.基于多模型权重提取与融合的网络安全态势预测研究[D].兰州:兰州理工大学, 2016.
	JIANG W F. Research on network security situation prediction based on multi-model weight extraction and fusion[D]. Lanzhou: Lanzhou University of Technology, 2016.
24	孙卫喜. 用于网络安全态势预测的粒子群与支持向量机算法研究[J]. 计算机应用与软件, 2019, 36 (6): 308- 316.
	SUN W X . PSO and SVM for network security situation prediction[J]. Computer Applications and Software, 2019, 36 (6): 308- 316.
25	MUNKHDALAI L . An end-to-end adaptive input selection with dynamic weights for forecasting multivariate time series[J]. IEEE Access, 2019, 7, 99099- 99114.
26	BAHDANAU D, KYUNGHYUN C, BENGIO Y. Neural machine translation by jointly learning to align and translate[C]//Proc.of the 3rd International Conference on Learning Representations, 2015.

预测模型	MAE	RMSE
Attention-GRU	0.001 6	0.002 7
AIS-LSTM	0.004 1	0.005 4
Attention-RNN	0.007 4	0.007 8
GRU	0.074 5	0.089 5
BP	0.008 7	0.009 3
PSO_SVM	0.111 7	0.153 3

序号	Attention-GRU	AIS-LSTM	Attention-RNN	GRU	PSO_SVM	BP
1	0.000 67	0.002 64	0.005 58	0.090 65	0.002 93	0.010 12
2	0.002 52	0.002 09	0.005 45	0.063 53	0.018 31	0.007 89
3	0.002 88	0.002 11	0.007 05	0.082 33	0.039 51	0.008 01
4	0.001 14	0.006 35	0.005 08	0.198 15	0.014 23	0.011 51
5	0.000 35	0.003 75	0.004 94	0.108 85	0.015 66	0.010 39
6	0.000 15	0.004 41	0.005 78	0.105 82	0.011 87	0.010 04
7	0.000 49	0.002 07	0.006 72	0.026 61	0.074 79	0.010 41
8	0.000 69	0.002 21	0.006 32	0.038 06	0.057 84	0.009 02
9	0.002 03	0.000 45	0.012 05	0.008 23	0.253 50	0.000 24
10	0.000 12	0.000 73	0.009 02	0.054 72	0.149 55	0.008 47
11	0.003 29	0.008 81	0.009 53	0.019 44	0.247 28	0.002 66
12	0.003 41	0.024 17	0.011 67	0.047 68	0.356 41	0.007 31
13	0.002 74	0.005 85	0.007 54	0.091 61	0.180 83	0.004 81
14	0.001 55	0.004 12	0.006 28	0.154 76	0.071 31	0.009 16
15	0.002 09	0.004 83	0.004 22	0.126 61	0.011 03	0.009 87
16	0.000 43	0.002 56	0.007 66	0.034 64	0.107 37	0.009 07
17	0.001 05	0.000 33	0.007 76	0.020 08	0.133 93	0.008 73
18	0.003 83	0.018 14	0.010 03	0.070 82	0.264 32	0.001 14

模型	复杂度
Attention-GRU	O(Q3(2Nm+m²))
AIS-LSTM	O(Q4(2Nm+m²))
Attention-RNN	O(Q(2Nm+m²))
GRU	O(Q3(Nm+m²))
BP	O(QNm)
PSO_SVM[13]	O(N³)

模型	时间/s
模型	PSO算法	模型	预测	总和
Attention-GRU	4 701	182	0.57	4 883.57
AIS-LSTM	7 361	223	0.73	7 584.73
Attention-RNN	2 872	10	0.32	2 882.32
GRU	2 185	4	0.06	2 189.06
BP	240	0.96	0.014	240.97
PSO_SVM	3.85	0.02	0.001	3.871

[1]	韩啸, 陈世文, 陈蒙, 杨锦程. 基于互易点学习的LPI信号开集识别[J]. 系统工程与电子技术, 2022, 44(9): 2752-2759.
[2]	杨建峰, 肖和业, 李亮, 白俊强, 董维浩. 基于模糊聚类和专家评分机制的无人机多层次模块划分方法[J]. 系统工程与电子技术, 2022, 44(8): 2530-2539.
[3]	徐平亮, 崔亚奇, 熊伟, 熊振宇, 顾祥岐. 生成式中断航迹接续关联方法[J]. 系统工程与电子技术, 2022, 44(5): 1543-1552.
[4]	杜思予, 全英汇, 沙明辉, 方文, 邢孟道. 基于进化PSO算法的稀疏捷变频雷达波形优化[J]. 系统工程与电子技术, 2022, 44(3): 834-840.
[5]	吴涛, 王伦文, 朱敬成. 基于迁移学习和注意力机制的伪装图像分割[J]. 系统工程与电子技术, 2022, 44(2): 376-384.
[6]	金涛, 王晓峰, 田润澜, 张歆东. 基于改进1DCNN+TCN的雷达辐射源快速识别方法[J]. 系统工程与电子技术, 2022, 44(2): 463-469.
[7]	马钰棠, 孙鹏, 张杰勇, 王鹏, 闫云飞, 赵亮. 样本不平衡下的空中群组意图识别方法[J]. 系统工程与电子技术, 2022, 44(12): 3747-3755.
[8]	唐一强, 杨霄鹏, 朱圣铭. 基于注意力机制的混合CNN-BiLSTM低轨卫星信道预测算法[J]. 系统工程与电子技术, 2022, 44(12): 3863-3870.
[9]	曲凌志, 杨俊安, 刘辉, 黄科举. 嵌入注意力机制的通信辐射源个体识别方法[J]. 系统工程与电子技术, 2022, 44(1): 20-27.
[10]	刘紫燕, 马珊珊, 梁静, 朱明成, 袁磊. 注意力机制CNN的毫米波大规模MIMO系统信道估计算法[J]. 系统工程与电子技术, 2022, 44(1): 307-312.
[11]	崔邦彦, 田润澜, 王东风, 崔钢, 石静苑. 基于注意力机制和改进CLDNN的雷达辐射源识别[J]. 系统工程与电子技术, 2021, 43(5): 1224-1231.
[12]	王坤, 侯树贤, 王力. 基于自适应变异PSO-SVM的APU性能参数预测模型[J]. 系统工程与电子技术, 2021, 43(2): 526-536.
[13]	高诗飏, 董会旭, 田润澜, 张歆东. 基于SRNN+Attention+CNN的雷达辐射源信号识别方法[J]. 系统工程与电子技术, 2021, 43(12): 3502-3509.
[14]	赵帅, 刘松涛, 汪慧阳. 基于PSO-CNN的LPI雷达波形识别算法[J]. 系统工程与电子技术, 2021, 43(12): 3552-3563.
[15]	赵若辰, 王敬东, 林思玉, 顾东泽. 基于卷积神经网络的小型建筑物检测算法[J]. 系统工程与电子技术, 2021, 43(11): 3098-3106.