ReXNet：融合不确定性量化与可解释性的空天安全可信框架

刘壮; 陈雨然; 张嘉桐; 蒋雨静; 汪旭晖

doi:10.11999/JEIT251159

ReXNet：融合不确定性量化与可解释性的空天安全可信框架

doi: 10.11999/JEIT251159 cstr: 32379.14.JEIT251159

1.
东北财经大学金融科技学院大连 116025
2.
东北财经大学管理科学与工程学院大连 116025
3.
东北财经大学工商管理学院大连 116025

基金项目: 国家自然科学基金专项项目(72442025)

详细信息

作者简介:
刘壮：男，副教授、博士/硕士研究生导师，研究方向为图像处理与智能识别、空天信息网络、智能与协同控制、机器学习等

陈雨然：女，研究方向为机器学习、人工神经网络与计算等

张嘉桐：女，研究方向为机器学习、人工神经网络与计算等

蒋雨静：女，研究方向为机器学习、人工神经网络与计算等

汪旭晖：男，教授、博士/硕士研究生导师，研究方向为专家系统

通讯作者:
陈雨然　yuranchen06@gmail.com

中图分类号: TN915.08; TP393.081
计量
- 文章访问数: 125
- HTML全文浏览量: 57
- PDF下载量: 20
- 被引次数: 0
出版历程
- 收稿日期: 2025-11-01
- 修回日期: 2025-01-28
- 录用日期: 2026-01-30
- 网络出版日期: 2026-02-12

ReXNet: A Trustworthy Framework for Space-air Security Integrating Uncertainty Quantification and Explainability

1.
Dongbei University of Finance and Economics, Dalian 116025, China
2.
School of Management Science and Engineering, Dongbei University of Finance and Economics, Dalian 116025, China
3.
School of Business Administration, Dongbei University of Finance and Economics, Dalian 116025, China

Funds: The Special Funds of The National Natural Science Foundation of China (72442025)

摘要

摘要: 随着空天地一体化网络日益发展，成为国家战略前沿，其深度融合的卫星遥感、导航定位和通信应用，均对人工智能的可靠性与透明度提出了严苛要求。特别地，空天信息系统面临着物理层、网络层到应用层的复合式安全挑战，在这些高风险敏感性场景中，发展高稳健性与可信度的智能检测技术已成为当务之急。为应对这一挑战，该文提出了一个新颖的可信人工智能框架ReXNet。该框架深度整合了不确定性量化与可解释人工智能技术，并允许灵活替换骨干模型，以适配多样化的空天安全任务。通过在入侵检测、故障诊断及广播式自动相关监测(ADS-B)注入攻击等空天地3层典型安全场景数据集上的实验验证，ReXNet框架在保持高精度异常检测性能的同时，能有效量化预测置信度、识别模型知识边界外的未知样本，并为决策提供逻辑一致且可追溯的归因解释。该框架的模块化与灵活性创新，为解决人工智能在安全关键系统中的应用瓶颈提供了有效的技术路径。通过系统性地提升模型的可靠性与透明度，该研究旨在推动智能检测技术在空天安全领域的应用范式从追求单一的“高精度”向兼顾“高可信”转变，显著增强了其场景适用性与整体可信度。
- 空天地一体化网络 /
- 空天安全 /
- 可信人工智能 /
- 不确定性量化 /
- 可解释性
Abstract: Objective The Space-Air-Ground Integrated Network (SAGIN) has emerged as a strategic infrastructure for national development. However, its security vulnerabilities are increasingly evident. The physical, network, and application layers of SAGIN face different security challenges that require targeted protection strategies. Aerospace scenarios require both high predictive accuracy and transparent decision making. Therefore, more robust, reliable, and interpretable intelligent methods are needed to support network security and system trustworthiness. Methods A detection framework is proposed that integrates Uncertainty Quantification (UQ) and eXplainable Artificial Intelligence (XAI). In the front-end stage, a Bayesian deep learning method based on Monte Carlo Dropout is adopted to enable probabilistic prediction modeling. This approach separates and quantifies epistemic uncertainty and aleatoric uncertainty, which improves model reliability. In the back-end stage, SHAP and LIME are applied to provide feature attribution for each prediction, improving model interpretability and transparency. Moreover, the intermediate layer of the framework allows flexible replacement of deep learning backbones, enabling adaptation to different space and aerospace application scenarios. Results and Discussions Extensive experiments were conducted on representative space-air security datasets, including UAV swarm fault detection, ADS-B injection attacks, and network fraud detection. The experimental results show that the proposed framework achieves high-precision anomaly detection. It also evaluates prediction confidence and identifies unknown samples outside the model knowledge boundary. In addition, the framework generates logically consistent and traceable explanations for model decisions, which improves interpretability and operational reliability. The results indicate that the combined use of UQ and XAI improves the robustness and trustworthiness of intelligent models in aerospace security applications. Conclusions This study improves the reliability and transparency of anomaly detection models in the space-air domain. It reflects a transition in artificial intelligence applications from focusing only on prediction accuracy to emphasizing system trustworthiness. Future work will promote practical deployment of the framework. The focus will include real-time processing capability, lightweight implementation, and operation in resource-constrained environments such as onboard and on-orbit systems. These efforts support more secure, autonomous, and efficient operation of SAGIN and contribute to the sustainable development of future space-air information networks.
- Space-Air-Ground Integrated Network (SAGIN) /
- Aerospace security /
- Trustworthy Artificial Intelligence(AI) /
- Uncertainty Quantification (UQ) /
- EXplainable Artificial Intelligence (XAI)

HTML全文

图 1 ReXNet框架结构层级概述图

下载: 全尺寸图片幻灯片

图 2 不确定性量化分析

下载: 全尺寸图片幻灯片

图 3 基于SHAP值的特征全局重要性及影响分布

下载: 全尺寸图片幻灯片

图 4 SHAP依赖图：bwd_pkt_len_mean (排名第1) 特征的详细效应

下载: 全尺寸图片幻灯片

图 5 基于SHAP与LIME的单样本可解释性分析

下载: 全尺寸图片幻灯片

图 6 最重要特征的PDP-UQ协同分析图

下载: 全尺寸图片幻灯片

图 7 ReXNet框架的扩展性可解释分析

下载: 全尺寸图片幻灯片

表 1 不同模型在 UAV-GCS-IDS 数据集上的性能对比

模型类别	模型架构	Accuracy	F1-Score
基准模型	TabNet	0.909	0.952
	XGBoost	0.929	0.962
	Transformer	0.910	0.953
骨干模型	DNN	0.912	0.954
	CNN	0.907	0.950
	LSTM	0.903	0.948
	ResNet	0.911	0.953
	GatedNet	0.913	0.955
ReXNet 模式1(EU 量化)	B-DNN	0.919	0.957
	B-CNN	0.911	0.953
	B-LSTM	0.909	0.951
	B-ResNet	0.912	0.961
	B-GatedNet	0.912	0.954
ReXNet 模式2(完整 UQ)	Full-DNN	0.925	0.960
	Full-CNN	0.917	0.956
	Full-LSTM	0.913	0.954
	Full-ResNet	0.933	0.965
	Full- GatedNet	0.927	0.961

下载: 导出CSV

表 2 不同模型在SAGIN不同层次上的F1-Score性能对比

模型类别	模型架构	数据集
		物理层	网络层		应用层
		C-MAPSS	T-ITS	ADS-B	GUIDE
基准模型	TabNet	0.836	0.924	0.986	0.772
	XGBoost	0.877	0.937	0.985	0.917
	Transformer	0.689	0.919	0.653	0.744
骨干模型	DNN	0.940	0.938	0.971	0.915
	CNN	0.938	0.945	0.977	0.913
	LSTM	0.941	0.950	0.984	0.918
	ResNet	0.942	0.951	0.979	0.921
	GatedNet	0.940	0.949	0.967	0.919
ReXNet 模式2(EU 量化)	B-DNN	0.959	0.941	0.977	0.922
	B-CNN	0.955	0.948	0.978	0.920
	B-LSTM	0.957	0.954	0.985	0.925
	B-ResNet	0.960	0.954	0.982	0.928
	B-GatedNet	0.958	0.952	0.981	0.926
ReXNet 模式1(完整 UQ)	Full-DNN	0.986	0.946	0.979	0.931
	Full-CNN	0.985	0.952	0.980	0.929
	Full-LSTM	0.987	0.956	0.986	0.934
	Full-ResNet	0.990	0.957	0.989	0.938
	Full-GatedNet	0.988	0.955	0.983	0.936

下载: 导出CSV

参考文献(25)

[1]	刘光毅, 张慧敏, 佟舟, 等. 6G移动信息网络架构: 从通信到一切皆服务的变迁[J]. 中国科学: 信息科学, 2024, 54(5): 1236–1266. doi: 10.1360/SSI-2023-0339. LIU Guangyi, ZHANG Huimin, TONG Zhou, et al. 6G mobile information network architecture: Migrate from communication to XaaS[J]. Scientia Sinica Informationis, 2024, 54(5): 1236–1266. doi: 10.1360/SSI-2023-0339.
[2]	徐金雷, 赵俊湦, 卢华兵, 等. 面向6G的多维扩展通感一体化研究综述[J]. 电子与信息学报, 2024, 46(5): 1672–1683. doi: 10.11999/JEIT231045. XU Jinlei, ZHAO Junsheng, LU Huabing, et al. An overview on multi-dimensional expanded integrated sensing and communication for 6G[J]. Journal of Electronics & Information Technology, 2024, 46(5): 1672–1683. doi: 10.11999/JEIT231045.
[3]	尹浩, 黄宇红, 韩林丛, 等. 6G通信–感知–计算融合网络的思考[J]. 中国科学: 信息科学, 2023, 53(9): 1838–1842. doi: 10.1360/SSI-2023-0135. YIN Hao, HUANG Yuhong, HAN Lincong, et al. Thoughts on 6G integrated communication, sensing and computing networks[J]. Scientia Sinica Informationis, 2023, 53(9): 1838–1842. doi: 10.1360/SSI-2023-0135.
[4]	ZHUO Ming, LIU Leyuan, ZHOU Shijie, et al. Survey on security issues of routing and anomaly detection for space information networks[J]. Scientific Reports, 2021, 11(1): 22261. doi: 10.1038/s41598-021-01638-z.
[5]	LIU Zhuang, QIAN Shiyao, CAO Shuirong, et al. Mitigating age-related bias in large language models: Strategies for responsible artificial intelligence development[J]. INFORMS Journal on Computing, 2025. doi: 10.1287/ijoc.2024.0645.
[6]	XIAO Yue, YE Ziqiang, WU Mingming, et al. Space-air-ground integrated wireless networks for 6G: Basics, key technologies, and future trends[J]. IEEE Journal on Selected Areas in Communications, 2024, 42(12): 3327–3354. doi: 10.1109/JSAC.2024.3492720.
[7]	ZHANG Da, GAO Junyu, and LI Xuelong. Learning long-range relationships for temporal aircraft anomaly detection[J]. IEEE Transactions on Aerospace and Electronic Systems, 2024, 60(5): 6385–6395. doi: 10.1109/TAES.2024.3404360.
[8]	WANG Xuhui, LIN W, ZHAO Jun, et al. AdaFuse: Memory-augmented adaptive fusion for few-shot fraud detection in heterogeneous graphs[J]. Annals of Operations Research, 2026: 1–27. doi: 10.1007/s10479-026-07043-x.
[9]	杨宏宇, 宋成瑜, 王朋, 等. 洋葱路由器网站指纹攻击与防御研究综述[J]. 电子与信息学报, 2024, 46(9): 3474–3489. doi: 10.11999/JEIT240091. YANG Hongyu, SONG Chengyu, WANG Peng, et al. Website fingerprinting attacks and defenses on tor: A survey[J]. Journal of Electronics & Information Technology, 2024, 46(9): 3474–3489. doi: 10.11999/JEIT240091.
[10]	KHAN H A, KHAN H, GHAFOOR S, et al. A survey on security of automatic dependent surveillance-broadcast (ADS-B) protocol: Challenges, potential solutions, and future directions[J]. IEEE Communications Surveys & Tutorials, 2025, 27(5): 3199–3226. doi: 10.1109/COMST.2024.3513213.
[11]	WANG Qianyu, TSAI W T, SHI Tianyu, et al. Catch me if you can: A multi-agent synthetic fraud detection framework for complex networks[C]. 2025 IEEE 41st International Conference on Data Engineering (ICDE), Hong Kong, China, 2025: 3629–3641. doi: 10.1109/ICDE65448.2025.00271.
[12]	胡杨林, 张天魁, 李博, 等. 无人机使能的通信感知一体化组网与技术研究综述[J]. 电子与信息学报, 2025, 47(4): 859–875. doi: 10.11999/JEIT241116. HU Yanglin, ZHANG Tiankui, LI Bo, et al. A survey on UAV-enabled integrated sensing and communication networking and technologies[J]. Journal of Electronics & Information Technology, 2025, 47(4): 859–875. doi: 10.11999/JEIT241116.
[13]	王赞, 陈晓, 路辉, 等. 面向有限数据和多任务场景的无人机故障诊断[J/OL]. https://doi.org/10.13976/j.cnki.xk.2025.1311, 2025. WANG Zan, CHEN Xiao, LU Hui, et al. UAV fault diagnosis for limited data and multi task scenarios[J/OL]. https://doi.org/10.13976/j.cnki.xk.2025.1311, 2025.
[14]	ZHENG Yunfei, ZHANG Xuejun, TAN Yuanhao, et al. Transformer-based identification for ADS-B transmitters in open–time sets[J]. Chinese Journal of Aeronautics, 2025, 38(8): 103418. doi: 10.1016/j.cja.2025.103418.
[15]	WU Yuhe, CHEN Yuran, LIU Zhuang, et al. Enhancing financial decision-making under cyber threats: A dual-branch framework integrating Bayesian deep learning and explainable AI[J]. Annals of Operations Research, 2025: 1–33. doi: 10.1007/s10479-025-06973-2.
[16]	HE Wenchong, JIANG Zhe, XIAO Tingsong, et al. A survey on uncertainty quantification methods for deep learning[J]. ACM Computing Surveys, 2026, 58(7): 179. doi: 10.1145/3786319.
[17]	BASORA L, VIENS A, CHAO M A, et al. A benchmark on uncertainty quantification for deep learning prognostics[J]. Reliability Engineering & System Safety, 2025, 253: 110513. doi: 10.1016/j.ress.2024.110513.
[18]	CHEN Rentong, WANG Shaoping, ZHANG Chao, et al. Component uncertainty importance measure in complex multi-state system considering epistemic uncertainties[J]. Chinese Journal of Aeronautics, 2024, 37(12): 31–54. doi: 10.1016/j.cja.2024.05.024.
[19]	RAZAVI M, MAVADDATI S, and KOOHI H. ResNet deep models and transfer learning technique for classification and quality detection of rice cultivars[J]. Expert Systems with Applications, 2024, 247: 123276. doi: 10.1016/j.eswa.2024.123276.
[20]	XUE Lanqing, LI Xiaopeng, and ZHANG N L. Not all attention is needed: Gated attention network for sequence data[C]. The Thirty-Fourth AAAI Conference on Artificial Intelligence, New York Hilton Midtown, USA, 2020: 6550–6557. doi: 10.1609/aaai.v34i04.6129.
[21]	HADI H J and HADI H J. UAV-GCS intrusion detection dataset (UAV-GCS-IDS)[EB/OL]. https://doi.org/10.6084/m9.figshare.29608541.v1, 2025.
[22]	WU Chuanwen, ZHANG Shumei, BAO Xiaoli, et al. Risk assessment approach of electronic component selection in equipment R&D Using the XGBoost algorithm and domain knowledge[J]. Processes, 2024, 12(8): 1716. doi: 10.3390/pr12081716.
[23]	DONG Yutong, JIANG Hongkai, MU Mingzhe, et al. Multi-sensor data fusion-enabled lightweight convolutional double regularization contrast transformer for aerospace bearing small samples fault diagnosis[J]. Advanced Engineering Informatics, 2024, 62: 102573. doi: 10.1016/j.aei.2024.102573.
[24]	ARIK S Ö and PFISTER T. TabNet: Attentive interpretable tabular learning[C]. The Thirty-Fifth AAAI Conference on Artificial Intelligence, 2021: 6679–6687. doi: 10.1609/aaai.v35i8.16826.
[25]	PEBRIANTI D, KHALANI Z, and RUSDAH. Predictive maintenance in aerospace industry using convolutional neural network[C]. 2024 9th International Conference on Mechatronics Engineering (ICOM), Kuala Lumpur, Malaysia, 2024: 157–162. doi: 10.1109/ICOM61675.2024.10652505.