高级持续性威胁检测与溯源研究进展

doi:10.3969/j.issn.1671-1122.2026.06.001

摘要/Abstract

摘要：

近年来，高级持续性威胁（APT）攻击呈现爆发式增长，对政府机构和关键基础设施安全构成严峻挑战。基于溯源图的入侵检测系统（PIDS）通过捕获和分析系统层面的依赖关系，成为检测复杂、隐蔽的APT攻击的重要手段。首先，文章阐述了支撑PIDS研究的各类测试基准，并分析了其在规模、复杂性、攻击覆盖和标注质量等方面存在的局限。进一步以溯源图为基础，对现有的APT检测与攻击溯源方法进行分类，揭示了从早期规则匹配到机器学习，再到当前基于大语言模型（LLM）方法的技术演进路径，并总结了各类方法的优势与不足。文章对基于LLM的APT检测与溯源的范式进行了归纳，并逐步厘清了该领域面临的关键挑战。最后，对未来研究方向进行了展望。

关键词: 高级持续性威胁, 溯源图, 入侵检测系统, 大语言模型

Abstract:

In recent years, advanced persistent threat (APT) attacks have experienced explosive growth, posing severe challenges to government agencies and critical infrastructure. Provenance-based intrusion detection systems (PIDS), which capture and analyze system-level dependency relationships, provide a key means for detecting complex and stealthy APT attacks. This paper first analyzed various benchmark datasets supporting PIDS research, highlighting their limitations in scale, complexity, attack coverage, and annotation quality. Furthermore, based on provenance graphs, it classified existing APT detection and investigation methods, revealing the technological evolution from early rule-based matching to machine learning and current approaches utilizing Large Language Model (LLM). The advantages and disadvantages of each category were summarized. This work presented systematic review of the paradigm for LLM-based APT detection and investigation and clarified the key challenges faced in this field. Finally, it outlined prospects for future research directions.

Key words: advanced persistent threat, provenance graph, intrusion detection system, large language model

中图分类号:

TP309

孙钰, 张轩瑞, 刘新宇. 高级持续性威胁检测与溯源研究进展[J]. 信息网络安全, 2026, 26(6): 833-853.

SUN Yu, ZHANG Xuanrui, LIU Xinyu. Advances in Advanced Persistent Threat Detection and Provenance Research[J]. Netinfo Security, 2026, 26(6): 833-853.

图/表 18

图1

图2

表1

图3

图4

表2

表3

表4

表5

表6

图5

图6

图7

图8

图9

图10

图11

表7

参考文献 92

[1]	TANKARD C. Advanced Persistent Threats and how to Monitor and Deter Them[J]. Network Security, 2011, 2011(8): 16-19.
[2]	360 Digital Security Group. Global APT Threat Research Report[EB/OL]. [2026-02-14]. https://pub1-bjyt.s3.360.cn/bcms/2025%E5%B9%B4%E5%BA%A6%E5%85%A8%E7%90%83APT%E5%A8%81%E8%83%81%E7%A0%94%E7%A9%B6%E6%8A%A5%E5%91%8A.pdf.
	360 数字安全集团. 2025年度全球APT威胁研究报告[EB/OL]. [2026-02-14]. https://pub1-bjyt.s3.360.cn/bcms/2025%E5%B9%B4%E5%BA%A6%E5%85%A8%E7%90%83APT%E5%A8%81%E8%83%81%E7%A0%94%E7%A9%B6%E6%8A%A5%E5%91%8A.pdf.
[3]	LIU Fucheng, WEN Yu, ZHANG Dongxue, et al. Log2vec: A Heterogeneous Graph Embedding Based Approach for Detecting Cyber Threats within Enterprise[C]//ACM. The 2019 ACM SIGSAC Conference on Computer and Communications Security. New York: ACM, 2019: 1777-1794.
[4]	INAM M A, CHEN Yinfang, GOYAL A, et al. SoK: History Is a Vast Early Warning System: Auditing the Provenance of System Intrusions[C]//IEEE. 2023 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2023: 2620-2638.
[5]	KENT A D. Comprehensive, Multi-Source Cyber-Security Events[EB/OL]. [2026-02-14]. https://csr.lanl.gov/data/cyber1/.
[6]	MANZOOR E, MILAJERDI S M, AKOGLU L. Fast Memory-Efficient Anomaly Detection in Streaming Heterogeneous Graphs[C]//ACM. The 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2016: 1035-1044.
[7]	HAN Xueyuan, PASQUIER T, BATES A, et al. UNICORN: Runtime Provenance-Based Detector for Advanced Persistent Threats[EB/OL]. (2020-01-14)[2026-02-14]. https://arxiv.org/abs/2001.01525.
[8]	GitHub. Transparent Computing Engagement 3 Data Release[EB/OL]. [2026-02-14]. https://github.com/darpa-i2o/Transparent-Computing/blob/master/README-E3.md.
[9]	GitHub. Operationally Transparent Cyber (OpTC) Data Release[EB/OL]. [2026-02-14]. https://github.com/FiveDirections/OpTC-data.
[10]	ALSAHEEL A, NAN Yuhong, MA Shiqing. ATLAS: A Sequence-Based Learning Approach for Attack Investigation[EB/OL]. (2025-07-07)[2026-02-14]. https://www.usenix.org/conference/atc25.
[11]	RIDDLE A, WESTFALL K, BATES A. ATLASv2: ATLAS Attack Engagements, Version 2[EB/OL]. (2023-10-03)[2026-02-14]. https://arxiv.org/abs/2401.01341.
[12]	LI Shaofei, DONG Feng, XIAO Xusheng, et al. NODLINK: An Online System for Fine-Grained APT Attack Detection and Investigation[EB/OL]. (2023-11-09)[2026-02-14]. https://arxiv.org/abs/2311.02331.
[13]	LYU Mingqi, GAO Hongzhe, QIU Xuebo, et al. TREC: APT Tactic/ Technique Recognition via Few-Shot Provenance Subgraph Learning[C]//ACM. The 2024 on ACM SIGSAC Conference on Computer and Communications Security. New York: ACM, 2024: 139-152.
[14]	MILAJERDI S M, GJOMEMO R, ESHETE B, et al. HOLMES: Real-Time APT Detection through Correlation of Suspicious Information Flows[C]//IEEE. 2019 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2019: 1137-1152.
[15]	XIONG Chunlin, ZHU Tiantian, DONG Weihao, et al. Conan: A Practical Real-Time APT Detection System with High Accuracy and Efficiency[J]. IEEE Transactions on Dependable and Secure Computing, 2022, 19(1): 551-565.
[16]	JIA Zian, XIONG Yun, NAN Yuhong. MAGIC: Detecting Advanced Persistent Threats via Masked Graph Representation Learning[EB/OL]. [2026-02-14]. https://www.usenix.org/system/files/usenixsecurity24-jia-zian.pdf.
[17]	CHENG Zijun, LYU Qiujian, LIANG Jinyuan, et al. Kairos: Practical Intrusion Detection and Investigation Using Whole-System Provenance[C]//IEEE. 2024 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2024: 3533-3551.
[18]	ZENGY J, WANG Xiang, LIU Jiahao, et al. SHADEWATCHER: Recommendation-Guided Cyber Threat Analysis Using System Audit Records[C]//IEEE. 2022 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2022: 489-506.
[19]	XU Zhiqiang, FANG Pengcheng, LIU Changlin, et al. DEPCOMM: Graph Summarization on System Audit Logs for Attack Investigation[C]// IEEE. 2022 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2022: 540-557.
[20]	CHENG Wenrui, ZHU Tiantian, XIONG Chunlin, et al. SoK: Knowledge Is All You Need: Accelerating Last Mile Delivery for Automated Provenance-based Intrusion Detection with LLMs[EB/OL]. (2025-07-22)[2026-02-14]. https://arxiv.org/abs/2503.03108v2.
[21]	ALY A, MANSOUR E, YOUSSEF A. OCR-APT: Reconstructing APT Stories from Audit Logs Using Subgraph Anomaly Detection and LLMs[EB/OL]. (2025-10-20)[2026-02-14]. https://arxiv.org/abs/2510.15188.
[22]	BENABDERRAHMANE S, VALTCHEV P, CHENEY J, et al. APT-LLM: Embedding-Based Anomaly Detection of Cyber Advanced Persistent Threats Using Large Language Models[EB/OL]. (2025-02-13)[2026-02-14]. https://arxiv.org/abs/2502.09385.
[23]	GANDHI P A, WUDALI P N, AMARU Y, et al. Shield: Apt Detection and Intelligent Explanation Using LLM[EB/OL]. (2025-02-04)[2026-02-14]. https://arxiv.org/abs/2502.02342.
[24]	MUKHERJEE K, KANTARCIOGLU M. LLM-Driven Provenance Forensics for Threat Investigation and Detection[EB/OL]. (2025-11-09)[2026-02-14]. https://arxiv.org/abs/2508.21323.
[25]	Liu Xinyu, Sun Yu, Xiong Gaojian, et al. SAGE: Self-Reflective End-to-End Framework for Automated APT Investigation in 5G Networks[C]// IEEE. IEEE INFOCOM 2026-IEEE Conference on Computer Communications. New York: IEEE, 2026: 1-10.
[26]	ZIPPERLE M, GOTTWALT F, CHANG E, et al. Provenance-Based Intrusion Detection Systems: A Survey[J]. ACM Computing Surveys, 2023, 55(7): 1-36.
[27]	ZHANG Bo, GAO Yansong, KUANG Boyu, et al. A Survey on Advanced Persistent Threat Detection: A Unified Framework, Challenges, and Countermeasures[J]. ACM Computing Surveys, 2025, 57(3): 1-36.
[28]	WANG Zhiwei, HE Xijie, YI Xin, et al. Survey of Attack and Detection Based on the Full Life Cycle of APT[J]. Journal of Communications, 2024, 45(9): 206-228.
	王郅伟, 何睎杰, 易鑫, 等. 基于APT活动全生命周期的攻击与检测综述[J]. 通信学报, 2024, 45(9): 206-228.
[29]	YANG Xiuzhang, PENG Guojun, LIU Side, et al. Survey on Attribution and Inference Research for APT Attacks[J]. Journal of Software, 2024, 36(1): 203-252.
	杨秀璋, 彭国军, 刘思德. 面向APT攻击的溯源和推理研究综述[J]. 软件学报, 2024, 36(1): 203-252.
[30]	QIU Jing, CHEN Rongrong, ZHU Haojin, et al. A Survey of Network Attack Investigation Based on Provenance Graph[J]. Acta Electronica Sinica, 2024, 52(7): 2529-2556.
	仇晶, 陈荣融, 朱浩瑾, 等. 基于溯源图的网络攻击调查研究综述[J]. 电子学报, 2024, 52(7): 2529-2556.
[31]	HUTCHINS E M, CLOPPERT M J, AMIN M R. Intelligence-Driven Computer Network Defense Informed by Analysis of Adversary Campaigns and Intrusion Kill Chains[EB/OL]. [2026-02-14]. https://www.lockheedmartin.com/content/dam/lockheed-martin/rms/documents/cyber/LM-White-Paper-Intel-Driven-Defense.pdf.
[32]	STROM B E, APPLEBAUM A, MILLER D P, et al. Mitre ATT&CK: Design and Philosophy[EB/OL]. [2026-02-14]. https://www.mitre.org/sites/default/files/2021-11/prs-19-01075-28-mitre-attack-design-and-philosophy.pdf.
[33]	CALTAGIRONE S, PENDERGAST A, BETZ C. The Diamond Model of Intrusion Analysis[EB/OL]. [2026-02-14]. https://www.activeresponse.org/wp-content/uploads/2013/07/diamond.pdf.
[34]	SHOSTACK A. Threat Modeling: Designing for Security[EB/OL]. [2026-02-14]. https://www.usenix.org/system/files/login/articles/14_books.pdf.
[35]	ALBERTS C, DOROFEE A, STEVENS J, et al. Introduction to the OCTAVE Approach[EB/OL]. [2026-02-14]. https://fitxers.oriolrius.cat/1296/octave.pdf.
[36]	ANDREW J, ERIC I, ARNOTH D R, et al. High-Level Overview[EB/OL]. [2026-02-14]. https://fight.mitre.org/FiGHT_High-Level_Overview_PRS-23-2698.pdf.
[37]	KING S T, CHEN P M. Backtracking Intrusions[C]// ACM. The Nineteenth ACM Symposium on Operating Systems Principles. New York: ACM, 2003: 223-236.
[38]	JIANG Baoxiang, MADHOUN N E, AGHA K A, et al. Orthrus: Achieving High Quality of Attribution in Provenance-based Intrusion Detection Systems[EB/OL]. [2026-02-14]. https://www.usenix.org/conference/usenixsecurity25/presentation/jiang-baoxiang.
[39]	WANG Su, WANG Zhiliang, ZHOU Tao, et al. THREATRACE: Detecting and Tracing Host-Based Threats in Node Level through Provenance Graph Learning[J]. IEEE Transactions on Information Forensics and Security, 2022, 17: 3972-3987.
[40]	KARIM S S, AFZAL M, IQBAL W, et al. Advanced Persistent Threat (APT) and Intrusion Detection Evaluation Dataset for Linux Systems 2024[EB/OL]. (2024-03-05)[2026-02-14]. https://doi.org/10.1016/j.dib.2024.110290.
[41]	LIU Qi, BAO Kaibin, HAGENMEYER V. Aviator: A MITRE Emulation Plan-Derived Living Dataset for Advanced Persistent Threat Detection and Investigation[C]// IEEE. 2024 IEEE International Conference on Big Data (BigData). New York: IEEE, 2024: 5610-5619.
[42]	Microsoft. Microsoft System Moniter[EB/OL]. [2026-02-14]. https://learn.microsoft.com/zh-cn/sysinternals/downloads/sysmon.
[43]	Broadcom. Carbon Black Cloud[EB/OL]. [2026-02-14]. https://www.vmware.com/products/carbon-.
[44]	LENART L. S2-046[EB/OL]. [2026-02-14]. https://cwiki.apache.org/confluence/display/ww/S2-046.
[45]	GitHub. Atomic Red Team[EB/OL]. [2026-02-14]. https://github.com/redcanaryco/atomic-red-team.
[46]	Wazuh, Inc. The Open Source Security Platform[EB/OL]. [2026-02-14]. https://wazuh.com/.
[47]	ALTINISIK E, DENIZ F, SENCAR H T. ProvG-Searcher: A Graph Representation Learning Approach for Efficient Provenance Graph Search[C]// ACM. The 2023 ACM SIGSAC Conference on Computer and Communications Security. New York: ACM, 2023: 2247-2261.
[48]	ALY A, IQBAL S, YOUSSEF A, et al. MEGR-APT: A Memory-Efficient APT Hunting System Based on Attack Representation Learning[J]. IEEE Transactions on Information Forensics and Security, 2024, 19: 5257-5271.
[49]	ZHU Tiantian, YU Jinkai, XIONG Chunlin, et al. APTSHIELD: A Stable, Efficient and Real-Time APT Detection System for Linux Hosts[J]. IEEE Transactions on Dependable and Secure Computing, 2023, 20(6): 5247-5264.
[50]	WANG Lingzhi, SHEN Xiangmin, LI Weijian, et al. Incorporating Gradients to Rules: Towards Lightweight, Adaptive Provenance-Based Intrusion Detection[EB/OL]. (2024-09-20)[2026-02-14]. https://arxiv.org/abs/2404.14720.
[51]	LI Zhenyuan, WANG Lingzhi, WANG Zhengkai, et al. Incorporating Gradients to Rules: Toward Online, Adaptive Provenance-Based Intrusion Detection[J]. IEEE Transactions on Dependable and Secure Computing, 2026, 23(1): 782-798.
[52]	YANG Fan, XU Jiacen, XIONG Chunlin. PROGRAPHER: An Anomaly Detection System Based on Provenance Graph Embedding[EB/OL]. [2026-02-14]. https://www.usenix.org/conference/usenixsecurity23/presentation/yang-fan.
[53]	REHMAN M, AHMADI H, UL H W. Flash: A Comprehensive Approach to Intrusion Detection via Provenance Graph Representation Learning[C]// IEEE. 2024 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2024: 3552-3570.
[54]	XIANG Xutao, YANG Yanqing, QIAN Yurong, et al. Provenance Graph-Based Advanced Persistent Threats Detection via Self-Supervised Contrastive Learning[EB/OL]. [2026-02-14]. https://www.computer.org/csdl/proceedings-article/trustcom/2025/653200a352/2dMlD9RLreU.
[55]	XU Fan, ZHAO Qinxin, LIU Xiaoxiao, et al. Advanced Persistent Threat Detection via Mining Long-Term Features in Provenance Graphs[EB/OL]. (2025-01-28)[2026-02-14]. https://link.springer.com/article/10.1007/s11704-024-40610-8.
[56]	KHOURY J, KLISURA Đ, ZANDDIZARI H, et al. Jbeil: Temporal Graph-Based Inductive Learning to Infer Lateral Movement in Evolving Enterprise Networks[C]// IEEE. 2024 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2024: 3644-3660.
[57]	GOYAL A, WANG Gang, BATES A. R-CAID: Embedding Root Cause Analysis within Provenance-Based Intrusion Detection[C]// IEEE. 2024 IEEE Symposium on Security and Privacy (SP). New York: IEEE, 2024: 3515-3532.
[58]	BILOT T, PARIS-SACLAY U, JIANG B, et al. A Comprehensive Analysis of State-of-the-Art Provenance-Based Intrusion Detection Systems[EB/OL]. [2026-02-14]. https://www.usenix.org/conference/usenixsecurity25/presentation/bilot.
[59]	KAPOOR M, MELTON J, RIDENHOUR M, et al. PROV-GEM: Automated Provenance Analysis Framework Using Graph Embeddings[C]// IEEE. 2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA). New York: IEEE, 2021: 1720-1727.
[60]	CHEN Tieming, DONG Chengyu, Lyu Mingqi, et al. APT-KGL: An Intelligent APT Detection System Based on Threat Knowledge and Heterogeneous Provenance Graph Learning[EB/OL]. (2022-12-26)[2026-02-14]. https://ieeexplore.ieee.org/document/9999407.
[61]	SANG Anyuan, FAN Xuezheng, YANG Li, et al. STGAN: Detecting Host Threats via Fusion of Spatial-Temporal Features in Host Provenance Graphs[C]// ACM. The ACM on Web Conference 2025. New York:ACM. 2025: 1046-1057.
[62]	QIAO Wei, FENG Yebo, LI Teng. Slot: Provenance-Driven APT Detection through Graph Reinforcement Learning[EB/OL]. (2025-07-17)[2026-02-14]. https://doi.org/10.48550/arXiv.2410.17910.
[63]	STEVE G. The Linux Audit Daemon[EB/OL]. [2026-02-14]. https://linux.die.net/man/8/auditd/.
[64]	RAMAKI A A, GHAEMI-BAFGHI A, RASOOLZADEGAN A. CAPTAIN: Community-Based Advanced Persistent Threat Analysis in IT Networks[EB/OL]. (2023-07-29)[2026-02-14]. https://www.sciencedirect.com/science/article/abs/pii/S1874548223000331?via%3Dihub.
[65]	NARAYANAN A, CHANDRAMOHAN M, VENKATESAN R, et al. Graph2vec:Learning Distributed Representations of Graphs[EB/OL]. (2017-07-17)[2026-02-14]. https://arxiv.org/abs/1707.05005.
[66]	LAI Siwei, XU Liheng, LIU Kang, et al. Recurrent Convolutional Neural Networks for Text Classification[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2015, 29(1): 2267-2273.
[67]	HAMILTON W, YING R, LESKOVEC J. Inductive Representation Learning on Large Graphs[EB/OL]. (2017-12-04)[2026-02-14]. https://dl.acm.org/doi/10.5555/3294771.3294869.
[68]	CHURCH K W. Word2Vec[J]. Natural Language Engineering, 2017, 23(1): 155-162.
[69]	ZHOU Jie, CUI Ganqu, HU Shengding, et al. Graph Neural Networks: A Review of Methods and Applications[J]. AI Open, 2020, 1: 57-81.
[70]	VELIČKOVIĆ P, CUCURULL G, CASANOVA A, et al. Graph Attention Networks[EB/OL]. (2018-02-04)[2026-02-14]. https://arxiv.org/abs/1710.10903.
[71]	BENTLEY J L. Multidimensional Binary Search Trees Used for Associative Searching[J]. Communications of the ACM, 1975, 18(9): 509-517.
[72]	HWANG F K, RICHARDS D S. Steiner Tree Problems[J]. Networks, 1992, 22(1): 55-89.
[73]	SALTON G, BUCKLEY C. Term-Weighting Approaches in Automatic Text Retrieval[J]. Information Processing & Management, 1988, 24(5): 513-523.
[74]	CHICCO D. Siamese Neural Networks: An Overview[EB/OL]. (2020-08-18)[2026-02-14]. https://link.springer.com/protocol/10.1007/978-1-1016-0826-5_3.
[75]	HU Qingyong, LEE C K, LIU Qi, et al. Hierarchical Graph Transformer with Adaptive Node Sampling[EB/OL]. [2026-02-14]. https://www.proceedings.com/068431-1539.html.
[76]	CHENG Zhangyu, ZOU Chengming, DONG Jianwei. Outlier Detection Using Isolation Forest and Local Outlier Factor[C]// ACM. The Conference on Research in Adaptive and Convergent Systems. New York: ACM, 2019: 161-168.
[77]	SCHLICHTKRULL M, KIPF T N, BLOEM P, et al. Modeling Relational Data with Graph Convolutional Networks[EB/OL]. (2018-06-03)[2026-02-14]. https://link.springer.com/chapter/10.1007/978-3-319-93417-4_38.
[78]	BOUNSIAR A, MADDEN M G. One-Class Support Vector Machines Revisited[C]// IEEE. 2014 International Conference on Information Science & Applications (ICISA). New York: IEEE, 2014: 1-4.
[79]	DEVLIN J, CHANG Mingwei, LEE K, et al. BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding[C]//ACL. The 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies. Stroudsburg: ACL, 2019: 4171-4186.
[80]	LAN Zhenzhong, CHEN Mingda, GOODMAN S, et al. ALBERT: A Lite BERT for Self-Supervised Learning of Language Representations[EB/OL]. (2020-02-09)[2026-02-14]. https://arxiv.org/abs/1909.11942.
[81]	LIU Yinhan, OTT M, GOYAL N, et al. RoBERTa: A Robustly Optimized BERT Pretraining Approach[EB/OL]. (2019-07-26)[2026-02-14]. https://arxiv.org/abs/1907.11692.
[82]	SANH V, DEBUT L, CHAUMOND J, et al. DistilBERT, a Distilled Version of BERT: Smaller, Faster, Cheaper and Lighter[EB/OL]. (2020-03-01)[2026-02-14]. https://arxiv.org/abs/1910.01108.
[83]	WANG Wenhui, WEI F, DONG Li, et al. Minilm: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers[EB/OL]. [2026-02-14]. https://proceedings.neurips.cc/paper/2020/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf.
[84]	HINTON G E, SALAKHUTDINOV R R. Reducing the Dimensionality of Data with Neural Networks[J]. Science, 2006, 313(5786): 504-507.
[85]	KINGMA D P, WELLING M. Auto-Encoding Variational Bayes[EB/OL]. (2022-12-10)[2026-02-14]. https://arxiv.org/abs/1312.6114.
[86]	VINCENT P, LAROCHELLE H, BENGIO Y, et al. Extracting and Composing Robust Features with Denoising Autoencoders[C]//ACM. The 25th International Conference on Machine Learning(ICML ’08). New York: ACM, 2008: 1096-1103.
[87]	LEWIS P, PEREZ E, PIKTUS A, et al. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks[C]// ACM. The 34th International Conference on Neural Information Processing Systems. New York: ACM, 2020: 9459-9474.
[88]	HE Kaiming, CHEN Xinlei, XIE Saining, et al. Masked Autoencoders Are Scalable Vision Learners[C]// IEEE. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE, 2022: 15979-15988.
[89]	SHARMA S, GOSAIN A, JAIN S. A Review of the Oversampling Techniques in Class Imbalance Problem[EB/OL]. (2021-08-18)[2026-02-14]. https://link.springer.com/chapter/10.1007/978-981-16-2594-7_38.
[90]	BAHADIR C D, WANG A Q, DALCA A V, et al. Deep-Learning-Based Optimization of the Under-Sampling Pattern in MRI[J]. IEEE Transactions on Computational Imaging, 2020, 6: 1139-1152.
[91]	YU Yong, SI Xiaosheng, HU Changhua, et al. A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures[J]. Neural Computation, 2019, 31(7): 1235-1270.
[92]	FORTUNATO S. Community Detection in Graphs[J]. Physics Reports, 2010, 486(3/4/5): 75-174.

维度	传统攻击	APT攻击
攻击者	个体或小组织网络犯罪分子	全球性、有组织、有纪律的黑客组织、敌对势力
攻击目标	随机性选择攻击，通常以个体为主，以达到获取金钱、盗窃身份、欺诈等目的	特定攻击目标，通常针对国家安全信息、重要技术设施、重要行业商业机密等
攻击手段	攻击手段单一，常基于已有的恶意软件展开攻击	攻击手段复杂隐蔽、形式多样，结合特制工具、社会工程学等展开攻击
攻击时间	攻击时间较短，以一次性、大范围攻击为主	攻击时间较长、长期潜伏、多次渗透、低频回传
攻击影响	通常造成短期的系统中断或数据泄露，对个人用户和企业的运营产生一定的干扰和损失	可对关键设施造成长期的破坏，在国际政治和军事冲突中发挥重要作用，甚至影响全球的安全形势
追溯难度	攻击特性强，容易在短时间内被检测和捕获	攻击特征弱，隐蔽性强，缺少样本数据，很难被检测和捕获

数据集	年份	主机操作系统	数据格式	大小
LANL^[5]	2015	Windows	.txt	160 GB
Streamspot^[6]	2016	Linux	.tsv	2.8 GB
Unicorn SC^[7]	2020	Linux	.txt	147 GB
DARPA TC E3^[8]	2018	Windows, Linux, FreeBSD, Android	CDM v18	116 GB
DARPA TC E5^[8]	2019	Windows, Linux, FreeBSD, Android	CDM v20	340 GB
DARPA OpTC^[9]	2020	Windows 10	eCar	1TB
ATLAS^[10]	2021	Windows	.txt	6.5 GB
ATLASv2^[11]	2023	Windows	.txt	154 GB
Nodlink SD^[12]	2024	Windows, Linux	.json	15 GB
Kellect4APT^[13]	2024	Windows	.csv	18.5 GB
Linux-APT^[40]	2024	Linux	.csv	205 MB

数据集	真实世界模拟程度
数据集	良性数据来源	攻击面	主机数 /个	总事件 /件	恶意事件 /件	时间跨度
LANL^[5]	真实	DNS	17684	10亿	0.3亿	56天
Streamspot^[6]	模拟	Firefox	1	0.6亿	0.45亿	1天
Unicorn SC^[7]	模拟	CI服务器	1	2.8亿	480万	3天
DARPA TC E3^[8]	模拟	Firefox SMB Nginx	7	20亿	0.5亿	10天
DARPA TC E5^[8]	模拟	Firefox SMB Nginx	18	120亿	2.5亿	9天
DARPA OpTC^[9]	模拟	Powershell	500	170亿	3亿	7天
ATLAS^[10]	模拟	Flash Word	2	249万	17万	1天
ATLASv2^[11]	模拟	Flash Word	2	27637	18146	5天
Nodlink SD^[12]	模拟	Apache Struts2 MSSQL	300	4116	190	5天
Kellect4APT^[13]	模拟	Firefox SMB Nginx Flash Word	1	5100万	49万	1天
Linux-APT^[40]	模拟	Flash Word	2	10万	2.5万	6周
数据集	APT攻击生命周期覆盖度	标注粒度与质量	文档完整性	—	—	—
LANL^[5]	完整	实体	×	—	—	—
Streamspot^[6]	部分	图	×	—	—	—
Unicorn SC^[7]	部分	图	×	—	—	—
DARPA TC E3^[8]	部分	实体	×	—	—	—
DARPA TC E5^[8]	部分	实体	×	—	—	—
DARPA OpTC^[9]	部分	实体	×	—	—	—
ATLAS^[10]	完整	实体	×	—	—	—
ATLASv2^[11]	完整	实体	×	—	—	—
Nodlink SD^[12]	完整	实体	×	—	—	—
Kellect4APT^[13]	完整	TTP	×	—	—	—
Linux-APT^[40]	完整	实体	×	—	—	—

数据集	覆盖技术数	复杂度	标注字段	标注方法
LANL^[5]	中等	高	登录事件、进程ID、网络流	红队演练日志映射
Streamspot^[6]	简单	低	仅标记恶意/良性图结构	脚本执行自动生成
Unicorn SC^[7]	中等	中	CamFlow审计日志	基于攻击阶段人工批次划分
DARPA TC E3^[8]	中等	中	PID、文件路径、Socket、系统调用	红队日志和专家手工映射
DARPA TC E5^[8]	中等	中	完整CDM字段（进程、文件、网络等）	专家根据攻击方案手工标记
DARPA OpTC^[9]	完整	高	eCar格式（进程、文件、注册表、网络等）	基于红队操作日志的自动映射
ATLAS^[10]	完整	中	核心受损实体（进程命令行、文件路径）	专家人工选择性标注
ATLASv2^[11]	完整	高	多源日志（Sysmon 树、Carbon Black 审计）	真实操作过程嵌入人工标记
Nodlink SD^[12]	中等	中	事件语义（Web Socket连接、文件写入）	基于脚本触发时间人工校验
Kellect4APT^[13]	完整	中	TTP ID、应用调用栈、内核级事件	基于原子红队工具半自动标注
Linux-APT^[40]	完整	低	告警信息(源IP、文件、规则描述)	基于安全预警规则后验标记

检测方法	系统	检测新型攻击	假阳率	及时性	攻击还原	可解释性	告警粒度
图匹配方法	Provg-Searcher^[47]	○	◑	○	×	○	图
图匹配方法	MEGR-APT^[48]	○	◑	○	×	○	图
基于规则匹配的方法	APTSHIELD^[49]	○	◑	●	√	●	节点
	CAPTAIN^[50]	○	◑	◑	×	◑	节点
	CAPTAIN+^[51]	○	◑	●	×	◑	节点
基于重建的图学习方法	ProGrapher^[52]	●	◑	●	×	○	节点
	ThreaTrace^[39]	●	◑	●	×	○	节点
	FLASH^[53]	●	◑	●	√	◑	节点
	MAGIC^[16]	●	○	◑	×	○	节点
	APT-SSC^[54]	●	○	●	×	○	节点
	LT-ProveGD^[55]	●	○	●	×	○	节点
	KAIROS^[17]	●	◑	●	√	◑	图
	JBEIL^[56]	●	◑	●	×	○	节点
	Nodlink^[12]	●	○	●	√	◑	节点
	R-CAID^[57]	●	◑	●	×	○	节点
	ShadeWatcher^[18]	●	◑	◑	×	◑	边
	VELOX^[58]	●	○	●	×	◑	边
	ORTHRUS^[38]	●	○	●	√	◑	边
基于分类的图学习方法	Prov-gem^[59]	●	◑	●	×	○	节点
	ATLAS^[10]	◑	○	◑	×	◑	路径
	TREC^[13]	●	○	●	√	●	攻击手段
	APT-KGL^[60]	◑	◑	◑	×	○	节点
	OCR-APT^[21]	●	○	●	√	◑	节点
	STGAN^[61]	◑	○	●	√	○	节点
	Slot^[62]	◑	○	●	√	◑	节点
基于LLM的方法	APT-LLM^[22]	●	○	◑	×	○	节点
	OMNISEC^[20]	◑	○	◑	×	◑	节点
	SHIELD^[23]	◑	○	◑	√	●	节点
	SAGE^[25]	●	○	●	√	●	节点