pith. sign in

arxiv: 2606.29909 · v1 · pith:7HYYHO6Dnew · submitted 2026-06-29 · 💻 cs.CV

Traffic-CBM: A Structurally Interpretable Multimodal Framework for Encrypted Traffic Classification

Honglei Jin , Wenshuo Chen , Shaofeng Liang , Haozhe Jia , Menshuo Zhao , Shuxu Jin , Songning Lai , Yutao Yue This is my paper

Pith reviewed 2026-06-30 06:17 UTC · model grok-4.3

classification 💻 cs.CV
keywords encrypted traffic classificationmultimodal learningconcept representationsinterpretable modelsnetwork traffic analysisstructural interpretability
0
0 comments X

The pith

Traffic-CBM organizes encrypted traffic signals into explicit concept summaries drawn from predefined evidence groups.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a framework that replaces direct fusion of flow statistics, packet sequences, and byte data with a hierarchical concept space. Predefined groups turn each type of traffic evidence into scalar summaries that remain visible during classification. A reader would care because this structure lets analysts trace which evidence level influences a prediction without sacrificing accuracy on standard benchmarks.

Core claim

Traffic-CBM maps grouped flow statistics to statistical concepts, applies dedicated temporal encoders to disjoint subspaces for temporal concepts, and decomposes byte-level evidence into packet-level and cross-packet concepts, forming a unified hierarchical concept space that supports structural analysis of multimodal traffic evidence.

What carries the argument

The hierarchical concept space, where each concept is a scalar evidence summary constrained by a predefined traffic evidence group.

If this is right

  • The model reaches competitive and balanced accuracy on multiple encrypted traffic benchmarks.
  • The concept space is actively used by the classifier rather than ignored.
  • Different levels of traffic evidence become directly comparable through their concept activations.
  • Structural explanations of predictions become available without post-hoc attribution methods.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same grouping approach could be tested on other multimodal classification tasks where data sources differ in granularity.
  • Concept-level interventions, such as clamping a statistical concept, might allow controlled tests of decision sensitivity.
  • If the predefined groups prove insufficient for new traffic types, the framework would require an automatic group-discovery step.

Load-bearing premise

Predefined traffic evidence groups can produce scalar concept summaries that capture the relevant multimodal signals without requiring manual semantic annotations.

What would settle it

If randomizing the values of the learned concepts leaves the model's predictions on the encrypted traffic benchmarks unchanged, the claim that the concept space drives classification would be falsified.

Figures

Figures reproduced from arXiv: 2606.29909 by Haozhe Jia, Honglei Jin, Menshuo Zhao, Shaofeng Liang, Shuxu Jin, Songning Lai, Wenshuo Chen, Yutao Yue.

Figure 1
Figure 1. Figure 1: Overall performance versus cross-dataset stability [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of Traffic-CBM. Traffic-CBM organizes statistical, sequential, and byte-level traffic evidence into a hierarchical [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: GA2M main effects of learned concepts on Cipher [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Representative first-order shape functions of selected learned concepts on CipherSpectrum(AES128). The x-axis [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Ranking of the strongest first-order and second [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
read the original abstract

Encrypted traffic classification has achieved strong performance, but its decision process remains difficult to interpret. Existing methods usually combine flow statistics, packet sequences, and byte-level representations into opaque latent features, making it unclear which type of evidence actually drives the prediction. In this paper, we propose Traffic-CBM, a structurally interpretable multimodal framework for encrypted traffic classification. Instead of directly fusing heterogeneous traffic signals into a black-box representation, Traffic-CBM organizes them into a unified hierarchical concept space. These concepts are not manually annotated semantic attributes; rather, they are scalar evidence summaries constrained by predefined traffic evidence groups. More specifically, grouped flow statistics are mapped to statistical concepts, dedicated temporal encoders learn temporal concepts from disjoint feature subspaces, and byte-level evidence is further organized into packet-level and cross-packet concepts. This design turns heterogeneous traffic evidence into an explicit concept representation and makes different levels of traffic evidence easier to analyze. We evaluate Traffic-CBM on multiple encrypted traffic benchmarks. Results show that it achieves competitive and balanced classification performance while providing a clearer structural interpretation interface than conventional end-to-end fusion models. Further analyses suggest that the learned concept space is actively used in the prediction process and provides a clearer structural explanation of multimodal traffic evidence.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes Traffic-CBM, a structurally interpretable multimodal framework for encrypted traffic classification. It organizes heterogeneous traffic signals (flow statistics, temporal sequences, byte-level data) into a unified hierarchical concept space using predefined evidence groups: statistical concepts from grouped flow statistics, temporal concepts from disjoint subspaces via dedicated encoders, and packet-level/cross-packet concepts from byte-level evidence. The framework claims competitive and balanced classification performance on multiple encrypted traffic benchmarks while providing clearer structural interpretation than conventional end-to-end fusion models, with further analyses indicating the concept space is actively used in predictions.

Significance. If the experimental claims hold with proper baselines and ablations, the work could advance interpretable multimodal learning in network security by replacing opaque latent fusion with explicit, analyzable concept representations derived without manual semantic annotations. This addresses a key limitation in encrypted traffic classification where decision processes are hard to interpret.

major comments (2)
  1. [Abstract] Abstract: the claim of 'competitive and balanced classification performance' is asserted without any quantitative results, baselines, error bars, ablation studies, or dataset details, so the central performance claim cannot be evaluated.
  2. [Abstract] Abstract: the construction of scalar concept summaries from 'predefined traffic evidence groups' is described at a high level but no procedure, validation, or constraints are provided to confirm these groups capture relevant multimodal signals without introducing bias or requiring manual tuning.
minor comments (1)
  1. [Abstract] Abstract: the phrase 'hierarchical concept space' is used without a reference to a figure, equation, or formal definition of the hierarchy levels.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'competitive and balanced classification performance' is asserted without any quantitative results, baselines, error bars, ablation studies, or dataset details, so the central performance claim cannot be evaluated.

    Authors: We agree that the abstract would be strengthened by including key quantitative results to support the performance claim. In the revised version, we will add specific metrics such as accuracy and macro-F1 scores on the primary benchmarks, along with references to the main baselines and datasets used. The full experimental details, including error bars, ablation studies, and dataset descriptions, are already provided in Sections 4 and 5 of the manuscript. revision: yes

  2. Referee: [Abstract] Abstract: the construction of scalar concept summaries from 'predefined traffic evidence groups' is described at a high level but no procedure, validation, or constraints are provided to confirm these groups capture relevant multimodal signals without introducing bias or requiring manual tuning.

    Authors: The abstract is intentionally concise. The full procedure for constructing the scalar concept summaries from predefined traffic evidence groups—using standard groupings of flow statistics, dedicated encoders on disjoint temporal subspaces, and byte-level packet features—is detailed in Section 3, along with the rationale that these groups derive from conventional traffic feature categories without manual semantic annotations. This design inherently avoids per-sample manual tuning. We will revise the abstract to include a brief clause referencing that the groups follow established multimodal traffic partitioning practices. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The abstract and provided context describe a multimodal framework that organizes traffic signals into hierarchical concepts via predefined groups and encoders, but contain no equations, fitting procedures, self-citations, or derivation steps that reduce predictions or concepts to their own inputs by construction. No load-bearing claims are shown to be equivalent to fitted parameters or prior self-referential results. The evaluation claims competitive performance without detailing any circular reduction in the concept construction or prediction process. This is the expected outcome for a high-level architectural description lacking mathematical derivations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 4 invented entities

The framework rests on the domain assumption that predefined evidence groups yield useful concepts; it introduces the hierarchical concept space as a new organizational structure without independent evidence for its effectiveness beyond the abstract claim.

axioms (1)
  • domain assumption Predefined traffic evidence groups produce scalar summaries that capture relevant heterogeneous signals
    The mapping from flow statistics, temporal subspaces, and byte data to concepts depends on these groups being effective without manual annotation.
invented entities (4)
  • hierarchical concept space no independent evidence
    purpose: To unify multimodal traffic signals into explicit, analyzable concepts
    Central new structure introduced to replace opaque latent features.
  • statistical concepts no independent evidence
    purpose: To summarize grouped flow statistics
    Derived directly from the framework's design for statistical evidence.
  • temporal concepts no independent evidence
    purpose: To capture patterns from disjoint feature subspaces via encoders
    Part of the temporal evidence organization.
  • packet-level and cross-packet concepts no independent evidence
    purpose: To organize byte-level evidence at multiple scales
    Introduced for byte-level multimodal handling.

pith-pipeline@v0.9.1-grok · 5774 in / 1277 out tokens · 37356 ms · 2026-06-30T06:17:35.694533+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

54 extracted references · 37 canonical work pages

  1. [1]

    Rishabh Agarwal, Levi Melnick, Nicholas Frosst, Xuezhou Zhang, Ben Lengerich, Rich Caruana, and Geoffrey Hinton. 2021. Neural Additive Models: Interpretable Machine Learning with Neural Nets. InAdvances in Neural Information Processing Systems (NeurIPS)

    2021

  2. [2]

    Khaled Al-Naami, Swarup Chandra, Ahmad Mustafa, Latifur Khan, Zhiqiang Lin, Kevin Hamlen, and Bhavani Thuraisingham. 2016. Adaptive Encrypted Traffic Fingerprinting with Bi-Directional Dependence. InProceedings of the Annual Computer Security Applications Conference (ACSAC). 177–188

    2016

  3. [3]

    Motaharul Islam, and Mohammad Nurul Huda

    Zahedi Azam, Md. Motaharul Islam, and Mohammad Nurul Huda. 2023. Com- parative Analysis of Intrusion Detection Systems and Machine Learning-Based Model Analysis Through Decision Tree.IEEE Access11 (2023), 80348–80391. doi:10.1109/ACCESS.2023.3296444

    work page doi:10.1109/access.2023.3296444 2023

  4. [4]

    Asmaa Benchama and Khalid Zebbara. 2023. Novel Approach to Intrusion Detection: Introducing GAN-MSCNN-BILSTM with LIME Predictions.Data and Metadata2 (2023). doi:10.56294/dm2023202

    work page doi:10.56294/dm2023202 2023

  5. [5]

    Rich Caruana, Yin Lou, Johannes Gehrke, Paul Koch, Marc Sturm, and Noemie Elhadad. 2015. Intelligible Models for HealthCare: Predicting Pneumonia Risk and Hospital 30-day Readmission. InProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD). 1721–1730. doi:10. 1145/2783258.2788613

    work page arXiv 2015

  6. [6]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the North American Chapter of the Association for Computational Linguistics (NAACL)

    2019

  7. [7]

    Jay DeYoung, Sarthak Jain, Nazneen Fatema Rajani, Eric Lehman, Caiming Xiong, Richard Socher, and Byron C. Wallace. 2020. ERASER: A Benchmark to Evaluate Rationalized NLP Models. InProceedings of the Annual Meeting of the Association for Computational Linguistics (ACL). 4443–4458. doi:10.18653/v1/2020.acl-main. 408

    work page doi:10.18653/v1/2020.acl-main 2020

  8. [8]

    Wenqi Dong, Jing Yu, Xinjie Lin, Gaopeng Gou, and Gang Xiong. 2025. Deep learning and pre-training technology for encrypted traffic classification: A com- prehensive review.Neurocomputing617 (2025), 128444. doi:10.1016/j.neucom. 2024.128444

    work page doi:10.1016/j.neucom 2025

  9. [9]

    Ghorbani

    Gerard Draper-Gil, Arash Habibi Lashkari, Mohammad Saiful Islam Mamun, and Ali A. Ghorbani. 2016. Characterization of Encrypted and VPN Traffic Using Time-Related Features. InProceedings of the International Conference on Information Systems Security and Privacy (ICISSP)

    2016

  10. [10]

    Viet Duong, Qiong Wu, Zhengyi Zhou, Hongjue Zhao, Chenxiang Luo, Eric Zavesky, Huaxiu Yao, and Huajie Shao. 2024. CAT: Interpretable Concept-based Taylor Additive Models. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD). 723–734. doi:10.1145/3637528. 3672020

    work page doi:10.1145/3637528 2024

  11. [11]

    Zijun Hang, Yuliang Lu, Yongjie Wang, and Yi Xie. 2023. Flow-MAE: Leveraging Masked AutoEncoder for Accurate, Efficient and Robust Malicious Traffic Classi- fication. InProceedings of the International Symposium on Research in Attacks, Intrusions and Defenses (RAID). 297–314. doi:10.1145/3607199.3607206

    work page doi:10.1145/3607199.3607206 2023

  12. [12]

    Jamie Hayes and George Danezis. 2016. k-fingerprinting: a robust scalable web- site fingerprinting technique. InProceedings of the USENIX Security Symposium. 1187–1203

    2016

  13. [13]

    Hong Ye He, Zhi Guo Yang, and Xiang Ning Chen. 2020. PERT: Payload Encoding Representation from Transformer for Encrypted Traffic Classification. In2020 ITU Kaleidoscope: Industry-Driven Digital Transformation (ITU K). 1–8. doi:10. 23919/ITUK50268.2020.9303204

    work page arXiv 2020

  14. [14]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick

  15. [15]

    Rope3D: The Roadside Perception Dataset for Autonomous Driving and Monocular 3D Object Detection Task,

    Masked Autoencoders Are Scalable Vision Learners. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 15979– 15988. doi:10.1109/CVPR52688.2022.01553

    work page doi:10.1109/cvpr52688.2022.01553 2022

  16. [16]

    Pamela Hermosilla, Sebastián Berríos, and Héctor Allende-Cid. 2025. Explainable AI for Forensic Analysis: A Comparative Study of SHAP and LIME in Intrusion Detection Models.Applied Sciences15, 13 (2025). doi:10.3390/app15137329

    work page doi:10.3390/app15137329 2025

  17. [17]

    Yang Ji, Ying Sun, Yuting Zhang, Zhigaoyuan Wang, Yuanxin Zhuang, Zheng Gong, Dazhong Shen, Chuan Qin, Hengshu Zhu, and Hui Xiong. 2025. A Com- prehensive Survey on Self-Interpretable Neural Networks.Proc. IEEE113 (2025), 783–813. doi:10.1109/JPROC.2025.3635153

    work page doi:10.1109/jproc.2025.3635153 2025

  18. [18]

    Haozhe Jia, Wenshuo Chen, Zhihui Huang, Lei Wang, Hongru Xiao, Nanqian Jia, Keming Wu, Songning Lai, Bowen Tian, and Yutao Yue. 2025. Physics-Informed Representation Alignment for Sparse Radio-Map Reconstruction. InProceedings of the ACM International Conference on Multimedia (ACM MM). 12352–12360. doi:10.1145/3746027.3758161

    work page doi:10.1145/3746027.3758161 2025

  19. [19]

    Pang Wei Koh, Thao Nguyen, Yew Siang Tang, Stephen Mussmann, Emma Pierson, Been Kim, and Percy Liang. 2020. Concept Bottleneck Models. In Proceedings of the International Conference on Machine Learning (ICML). 5338– 5348

    2020

  20. [20]

    Kunda Lin, Xiaolong Xu, and Honghao Gao. 2021. TSCRNN: A Novel Clas- sification Scheme of Encrypted Traffic Based on Flow Spatiotemporal Fea- tures for Efficient Management of IIoT.Computer Networks190 (2021), 107974. doi:10.1016/j.comnet.2021.107974

    work page doi:10.1016/j.comnet.2021.107974 2021

  21. [21]

    Xinjie Lin, Gang Xiong, Gaopeng Gou, Zhen Li, Junzheng Shi, and Jing Yu

  22. [22]

    InProceedings of the ACM Web Conference (WWW)

    ET-BERT: A Contextualized Datagram Representation with Pre-training Transformers for Encrypted Traffic Classification. InProceedings of the ACM Web Conference (WWW). 633–642. doi:10.1145/3485447.3512217

    work page doi:10.1145/3485447.3512217

  23. [23]

    Chang Liu, Longtao He, Gang Xiong, Zigang Cao, and Zhen Li. 2019. FS-Net: A Flow Sequence Network for Encrypted Traffic Classification. InProceedings of the IEEE Conference on Computer Communications (INFOCOM). 1171–1179. doi:10.1109/INFOCOM.2019.8737507

    work page doi:10.1109/infocom.2019.8737507 2019

  24. [24]

    Yang Liu, Tianwei Zhang, and Shi Gu. 2025. Hybrid Concept Bottleneck Mod- els. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 20179–20189. doi:10.1109/CVPR52734.2025.01879

    work page doi:10.1109/cvpr52734.2025.01879 2025

  25. [25]

    Ilya Loshchilov and Frank Hutter. 2019. Decoupled Weight Decay Regularization. InInternational Conference on Learning Representations (ICLR)

    2019

  26. [26]

    Mohammad Lotfollahi, Mahdi Jafari Siavoshani, Ramin Shirali Hossein Zade, and Mohammdsadegh Saberian. 2020. Deep Packet: A Novel Approach for Encrypted Traffic Classification Using Deep Learning.Soft Computing24, 3 (2020), 1999–2012. doi:10.1007/s00500-019-04030-2

    work page doi:10.1007/s00500-019-04030-2 2020

  27. [27]

    Yin Lou, Rich Caruana, Johannes Gehrke, and Giles Hooker. 2013. Accurate Intelligible Models with Pairwise Interactions. InProceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD). 623–631. doi:10.1145/2487575.2487579

    work page doi:10.1145/2487575.2487579 2013

  28. [28]

    Lundberg and Su-In Lee

    Scott M. Lundberg and Su-In Lee. 2017. A Unified Approach to Interpreting Model Predictions. InAdvances in Neural Information Processing Systems (NeurIPS). 4768– 4777

    2017

  29. [29]

    Alfredo Nascita, Francesco Cerasuolo, Giuseppe Aceto, Domenico Ciuonzo, Va- lerio Persico, and Antonio Pescapé. 2023. Explainable Mobile Traffic Classifi- cation: The Case of Incremental Learning. InProceedings of the 2023 Workshop on Explainable and Safety Bounded Machine Learning for Networking. 25–31. doi:10.1145/3630050.3630178

    work page doi:10.1145/3630050.3630178 2023

  30. [30]

    Alfredo Nascita, Antonio Montieri, Giuseppe Aceto, Domenico Ciuonzo, Va- lerio Persico, and Antonio Pescapé. 2023. Improving Performance, Reliabil- ity, and Feasibility in Multimodal Multitask Traffic Classification with XAI. IEEE Transactions on Network and Service Management20 (2023), 1267–1289. doi:10.1109/TNSM.2023.3246794

    work page doi:10.1109/tnsm.2023.3246794 2023

  31. [31]

    Alfredo Nascita, Antonio Montieri, Giuseppe Aceto, Domenico Ciuonzo, Valerio Persico, and Antonio Pescapé. 2021. XAI Meets Mobile Traffic Classification: Understanding and Improving Multimodal Deep Learning Architectures.IEEE Transactions on Network and Service Management18, 4 (2021), 4225–4246. doi:10. 1109/TNSM.2021.3098157

    work page arXiv 2021

  32. [32]

    Nguyen, and Tsui-Wei Weng

    Tuomas Oikarinen, Subhro Das, Lam M. Nguyen, and Tsui-Wei Weng. 2023. Label-free Concept Bottleneck Models. InThe Eleventh International Conference on Learning Representations

    2023

  33. [33]

    Eleonora Poeta, Gabriele Ciravegna, Eliana Pastor, Tania Cerquitelli, and Elena Baralis. 2025. Concept-based Explainable Artificial Intelligence: A Survey.Com- put. Surveys(2025). doi:10.1145/3774643

    work page doi:10.1145/3774643 2025

  34. [34]

    Why Should I Trust You?

    Marco Tulio Ribeiro, Sameer Singh, and Carlos Guestrin. 2016. "Why Should I Trust You?": Explaining the Predictions of Any Classifier. InProceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD). 1135–1144. doi:10.1145/2939672.2939778

    work page doi:10.1145/2939672.2939778 2016

  35. [35]

    Ghorbani

    Iman Sharafaldin, Arash Habibi Lashkari, and Ali A. Ghorbani. 2018. Toward Generating a New Intrusion Detection Dataset and Intrusion Traffic Character- ization. InProceedings of the International Conference on Information Systems Security and Privacy (ICISSP)

    2018

  36. [36]

    Adit Sharma and Arash Habibi Lashkari. 2025. A Survey on Encrypted Network Traffic: Identification and Classification Techniques, Challenges, and Future Directions.Computer Networks257 (2025), 110984. doi:10.1016/j.comnet.2024. 110984

    work page doi:10.1016/j.comnet.2024 2025

  37. [37]

    Meng Shen, Jinpeng Zhang, Liehuang Zhu, Ke Xu, and Xiaojiang Du. 2021. Accurate Decentralized Application Identification via Encrypted Traffic Analysis Using Graph Neural Networks.IEEE Transactions on Information Forensics and Security16 (2021), 2367–2380. doi:10.1109/TIFS.2021.3050608

    work page doi:10.1109/tifs.2021.3050608 2021

  38. [38]

    Payap Sirinam, Mohsen Imani, Marc Juarez, and Matthew Wright. 2018. Deep Fin- gerprinting: Undermining Website Fingerprinting Defenses with Deep Learning. InProceedings of the ACM SIGSAC Conference on Computer and Communications Security (CCS). 1928–1943. doi:10.1145/3243734.3243768

    work page doi:10.1145/3243734.3243768 2018

  39. [39]

    Jiwon Suh, Juwon Hong, Mose Gu, and Jaehoon Paul Jeong. 2025. Interpretable Detection of Encrypted Traffic Using SHAP-Based Feature Attribution. InProceed- ings of the International Conference on Information and Communication Technology Convergence (ICTC). 190–195. doi:10.1109/ICTC66702.2025.11388509

    work page doi:10.1109/ictc66702.2025.11388509 2025

  40. [40]

    Taylor, Riccardo Spolaor, Mauro Conti, and Ivan Martinovic

    Vincent F. Taylor, Riccardo Spolaor, Mauro Conti, and Ivan Martinovic. 2018. Robust Smartphone App Identification via Encrypted Network Traffic Analysis. IEEE Transactions on Information Forensics and Security13, 1 (2018), 63–78. doi:10. 1109/TIFS.2017.2737970

    work page arXiv 2018

  41. [41]

    Bowen Tian, Wenshuo Chen, Zexi Li, Songning Lai, Jiemin Wu, and Yutao Yue

  42. [42]

    InProceedings of the ACM International Conference on Multimedia (ACM MM)

    Text2Weight: Bridging Natural Language and Neural Network Weight Spaces. InProceedings of the ACM International Conference on Multimedia (ACM MM). 10152–10160. doi:10.1145/3746027.3755441 Traffic-CBM: A Structurally Interpretable Multimodal Framework for Encrypted Traffic Classification

    work page doi:10.1145/3746027.3755441

  43. [43]

    Dubois, Martina Lindorfer, David Choffnes, Maarten van Steen, and Andreas Peter

    Thijs Sebastiaan van Ede, Riccardo Bortolameotti, Andrea Continella, Jingjing Ren, Daniel J. Dubois, Martina Lindorfer, David Choffnes, Maarten van Steen, and Andreas Peter. 2020. FlowPrint: Semi-Supervised Mobile-App Fingerprint- ing on Encrypted Network Traffic. InNetwork and Distributed System Security Symposium (NDSS). doi:10.14722/ndss.2020.24412

    work page doi:10.14722/ndss.2020.24412 2020

  44. [44]

    Gomez, Łukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need. InAdvances in Neural Information Processing Systems (NeurIPS). 6000– 6010

    2017

  45. [45]

    Tongze Wang, Xiaohui Xie, Wenduo Wang, Chuyi Wang, Youjian Zhao, and Yong Cui. 2024. NetMamba: Efficient Network Traffic Classification via Pre-Training Unidirectional Mamba. InProceedings of the IEEE International Conference on Network Protocols (ICNP). 1–11. doi:10.1109/ICNP61940.2024.10858569

    work page doi:10.1109/icnp61940.2024.10858569 2024

  46. [46]

    Wei Wang, Ming Zhu, Xuewen Zeng, Xiaozhou Ye, and Yiqiang Sheng. 2017. Malware Traffic Classification Using Convolutional Neural Network for Repre- sentation Learning. InProceedings of the International Conference on Information Networking (ICOIN). 712–717. doi:10.1109/ICOIN.2017.7899588

    work page doi:10.1109/icoin.2017.7899588 2017

  47. [47]

    Nimesha Wickramasinghe, Arash Shaghaghi, Gene Tsudik, and Sanjay Jha

  48. [48]

    InProceedings of the IEEE Symposium on Security and Privacy (SP)

    SoK: Decoding the Enigma of Encrypted Network Traffic Classifiers. InProceedings of the IEEE Symposium on Security and Privacy (SP). 1825–1843. doi:10.1109/SP61157.2025.00165

    work page doi:10.1109/sp61157.2025.00165 2025

  49. [49]

    Tieqi Xi, Qiuhua Zheng, Chuanhui Cheng, Ting Wu, Guojie Xie, Xuebiao Qian, Haochen Ye, and Zhenyu Sun. 2025. SwiftSession: A Novel Incremental and Adaptive Approach to Rapid Traffic Classification by Leveraging Local Features. Future Internet17, 3 (2025), 114. doi:10.3390/fi17030114

    work page doi:10.3390/fi17030114 2025

  50. [50]

    Luming Yang, Lin Liu, JunJie Huang, Zhuotao Liu, Shiyu Liang, Shaojing Fu, and Yongjun Wang. 2025. MM4flow: A Pre-trained Multi-modal Model for Versatile Network Traffic Analysis. InProceedings of the ACM SIGSAC Conference on Computer and Communications Security (CCS). 1664–1678. doi:10.1145/3719027. 3744804

    work page doi:10.1145/3719027 2025

  51. [51]

    Mateo Espinosa Zarlenga, Zohreh Shams, Michael Edward Nelson, Been Kim, and Mateja Jamnik. 2023. TabCBM: Concept-based Interpretable Neural Networks for Tabular Data.Transactions on Machine Learning Research(2023)

    2023

  52. [52]

    Ruijie Zhao, Mingwei Zhan, Xianwen Deng, Yanhao Wang, Yijun Wang, Guan Gui, and Zhi Xue. 2023. Yet Another Traffic Classifier: A Masked Autoencoder Based Traffic Transformer with Multi-Level Flow Representation.Proceedings of the AAAI Conference on Artificial Intelligence37, 4 (2023), 5420–5427. doi:10. 1609/aaai.v37i4.25674

    2023

  53. [53]

    Wenbo Zheng, Chao Gou, Lan Yan, and Shaocong Mo. 2020. Learning to Classify: A Flow-Based Relation Network for Encrypted Traffic Classification. InProceed- ings of the ACM Web Conference (WWW). 13–22. doi:10.1145/3366423.3380090

    work page doi:10.1145/3366423.3380090 2020

  54. [54]

    Guangmeng Zhou, Xiongwen Guo, Zhuotao Liu, Tong Li, Qi Li, and Ke Xu. 2025. TrafficFormer: An Efficient Pre-trained Model for Traffic Data. InProceedings of the IEEE Symposium on Security and Privacy (SP). 1844–1860. doi:10.1109/ SP61157.2025.00102 Honglei Jin et al. A Dataset Details We use six public encrypted traffic benchmarks in this work: Cipher- Sp...

    work page arXiv 2025