Proactive microwave link anomaly detection in cellular data networks

Jianfeng Zhang
Jianfeng Zhang
Marcus Kalander
Marcus Kalander
Junjian Ye
Junjian Ye

pp. 1069692020.

Cited by: 1|Bibtex|Views102|
EI
Keywords:
network topologicalRandom Forestproactive microwave link anomaly detectionAnomaly detectionDecision TreeMore(20+)
Weibo:
We show via evaluation that the features related to the network topology are crucial for anomaly detection and enable PMADS to achieve both high precision and high recall

Abstract:

Microwave links are widely used in cellular networks for large-scale data transmission. From the network operators’ perspective, it is critical to quickly and accurately detect microwave link failures before they actually happen, thereby maintaining the robustness of the data transmissions. We present PMADS, a machine-learning-based proac...More

Code:

Data:

0
Introduction
  • Microwave links are widely used in cellular data networks for high-speed Internet access [1, 7, 12].
  • Microwave link failures in cellular data networks are prevalent in practice
  • Various factors, such as large obstacles, background noise, or environmental factors, can degrade the performance of microwave transmissions [1].
  • The basic idea of active learning is to find the set of the most informative or representative samples for the learning task and use these samples to retrain the model [34].
Highlights
  • Microwave links are widely used in cellular data networks for high-speed Internet access [1, 7, 12]
  • We show that PMADS achieves a precision of 94.4% and a recall of 87.1% when using the topological features encoded via network embedding
  • Note that we have verified that for other training-to-testing ratios, PMADS still achieves accuracy gain compared to the baseline that uses only statistical features without adding the topological information
  • PMADS is a proactive microwave link anomaly detection system that is currently deployed in a production cellular data network
  • We show via evaluation that the features related to the network topology are crucial for anomaly detection and enable PMADS to achieve both high precision and high recall
  • We show that Anomaly Detection by Active Learning (ADAL) obtains comparable results to popular active learning algorithms, while having significantly lower running time
Results
  • The authors evaluate PMADS’s accuracy in two aspects: (i) effectiveness of the network features and different classifiers; and (ii) effectiveness of the proposed active learning algorithm ADAL, including parameter sensitivity and model updating capability.

    The authors use the dataset described in Section 2 for the evaluation.
  • The authors evaluate PMADS’s accuracy in two aspects: (i) effectiveness of the network features and different classifiers; and (ii) effectiveness of the proposed active learning algorithm ADAL, including parameter sensitivity and model updating capability.
  • The data from September 27, 2017 to October 16, 2017 is used as the training set, denoted as Etrain.
  • It contains 1,625 one-day link samples identified as anomalies and 40,446 nominal samples, denoted as positive and negative samples respectively.
  • The test set contains 495 positive samples and 12,075 negative samples
Conclusion
  • PMADS is a proactive microwave link anomaly detection system that is currently deployed in a production cellular data network.
  • It extracts statistical features from KPIs and learns network features from network topological information.
  • It applies a novel active learning algorithm ADAL, which selects samples for model updates and keeps the model adaptable at low cost.
  • The authors' findings shed light on how PMADS provides proactive fault tolerance for microwave links in cellular data networks
Summary
  • Introduction:

    Microwave links are widely used in cellular data networks for high-speed Internet access [1, 7, 12].
  • Microwave link failures in cellular data networks are prevalent in practice
  • Various factors, such as large obstacles, background noise, or environmental factors, can degrade the performance of microwave transmissions [1].
  • The basic idea of active learning is to find the set of the most informative or representative samples for the learning task and use these samples to retrain the model [34].
  • Objectives:

    The authors' goal is to design a system for anomaly detection to monitor all microwave links and find those with degradations.
  • The authors' goal is to label each sample xi as either positive or negative
  • Results:

    The authors evaluate PMADS’s accuracy in two aspects: (i) effectiveness of the network features and different classifiers; and (ii) effectiveness of the proposed active learning algorithm ADAL, including parameter sensitivity and model updating capability.

    The authors use the dataset described in Section 2 for the evaluation.
  • The authors evaluate PMADS’s accuracy in two aspects: (i) effectiveness of the network features and different classifiers; and (ii) effectiveness of the proposed active learning algorithm ADAL, including parameter sensitivity and model updating capability.
  • The data from September 27, 2017 to October 16, 2017 is used as the training set, denoted as Etrain.
  • It contains 1,625 one-day link samples identified as anomalies and 40,446 nominal samples, denoted as positive and negative samples respectively.
  • The test set contains 495 positive samples and 12,075 negative samples
  • Conclusion:

    PMADS is a proactive microwave link anomaly detection system that is currently deployed in a production cellular data network.
  • It extracts statistical features from KPIs and learns network features from network topological information.
  • It applies a novel active learning algorithm ADAL, which selects samples for model updates and keeps the model adaptable at low cost.
  • The authors' findings shed light on how PMADS provides proactive fault tolerance for microwave links in cellular data networks
Tables
  • Table1: Types of Key Performance Indicators (KPIs)
  • Table2: Definitions of the 19 statistical features for a microwave link attached to NE1 and NE2. We use the KPIs of one day in the calculations. Note that Features 13 to 19 include values from both NEs
  • Table3: Evaluation with different feature sets and classifiers. (a) Recall
  • Table4: The running time (seconds) of each algorithm
Download tables as Excel
Related work
  • Proactive maintenance: Some studies pay special attention to proactive maintenance in operational networks [33, 43]. Kimura et al [22] propose a log analysis system for proactive detection of failures and show that the abnormality depends not only on the keywords in messages but also generation patterns. Opprentice [27] is a novel anomaly detection approach based on supervised machine learning by collecting performance data in social networks. Hora [32] employs architectural knowledge to predict how a failure can propagate and affect other components. Our work addresses failure prediction in cellular networks, with specific focus on the degradation of microwave links.

    Analysis of cellular networks: Failure prediction and anomaly detection have been widely applied in cellular networks. Some studies analyze the logs of network devices [13, 46], while others focus on using performance data (e.g., KPIs) for failure detection or prediction purposes [3, 6, 38, 44, 49]. Our work specifically considers the topological information (i.e., the dependencies among microwave links) in failure prediction.
Reference
  • Microwave Link. http://www.microwave-link.com.
    Findings
  • N. Abe, B. Zadrozny, and J. Langford. Outlier detection by active learning. In Proceedings of the 12th ACM SIGKDD International Cconference on Knowledge Discovery and Data Mining, pages 504–509. ACM, 2006.
    Google ScholarLocate open access versionFindings
  • F. Ahmed, J. Erman, Z. Ge, A. X. Liu, J. Wang, and H. Yan. Detecting and localizing end-to-end performance degradation for cellular data services. In Proceedings of the 35th Annual IEEE International Conference on Computer Communications (INFOCOM), pages 1–9. IEEE, 2016.
    Google ScholarLocate open access versionFindings
  • P. Casas, P. Fiadino, and A. D’Alconzo. Machine-learning based approaches for anomaly detection and classification in cellular networks. In Proceedings of IFIP Traffic Monitoring and Analysis workshop (TMA), 2016.
    Google ScholarLocate open access versionFindings
  • B. Cheung, G. Kumar, and S. A. Rao. Statistical algorithms in fault detection and prediction: Toward a healthier network. Bell Labs Technical Journal, 9(4):171– 185, 2005.
    Google ScholarLocate open access versionFindings
  • G. F. Ciocarlie, U. Lindqvist, S. Novaczki, and H. Sanneck. Detecting anomalies in cellular networks using an ensemble method. In Proceedings of the 9th international conference on network and service management (CNSM), pages 171–174. IEEE, 2013.
    Google ScholarLocate open access versionFindings
  • Cisco. Microwave Adaptive Bandwidth Feature: Make Better Use of Available Bandwidth. https://www.cisco.com/c/en/us/solutions/collateral/service-provider/unified-ran-backhaul/white_paper_c11-728355.pdf.
    Findings
  • P. Cui, X. Wang, J. Pei, and W. Zhu. A survey on network embedding. IEEE Transactions on Knowledge and Data Engineering (TKDE), 31(5):833–852, May 2019.
    Google ScholarLocate open access versionFindings
  • S. Das, W.-K. Wong, T. Dietterich, A. Fern, and A. Emmott. Incorporating expert feedback into active anomaly discovery. In Proceedings of the 16th International Conference on Data Mining (ICDM), pages 853–858. IEEE, 2016.
    Google ScholarLocate open access versionFindings
  • S. Dasgupta and D. Hsu. Hierarchical sampling for active learning. In Proceedings of the 25th International Conference on Machine learning (ICML), pages 208–215. ACM, 2008.
    Google ScholarLocate open access versionFindings
  • C. Elkan and K. Noto. Learning classifiers from only positive and unlabeled data. In Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 213–220. ACM, 2008. Microwave Towards 2020.
    Google ScholarLocate open access versionFindings
  • [13] Y.-Y. Fanjiang and C.-P. Wu. Automatic data logging and quality analysis system for mobile devices. Mobile Information Systems, 2017, 2017.
    Google ScholarLocate open access versionFindings
  • [14] K. Fujii and H. Kashima. Budgeted stream-based active learning via adaptive submodular maximization. In Advances in Neural Information Processing Systems, pages 514–522, 2016.
    Google ScholarLocate open access versionFindings
  • [15] A. Grover and J. Leskovec. node2vec: Scalable feature learning for networks. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 855–864. ACM, 2016.
    Google ScholarLocate open access versionFindings
  • [16] Q. Guan, Z. Zhang, and S. Fu. Ensemble of bayesian predictors and decision trees for proactive failure management in cloud computing systems. Journal of Communications, 7(1):52–61, 2012.
    Google ScholarLocate open access versionFindings
  • [17] C. Haochen, P. Bryan, R. Al-Rfou, and S. Steven. A tutorial on network embeddings. CoRR, abs/1808.02590, 2018.
    Findings
  • [18] S. C. Hoi, R. Jin, J. Zhu, and M. R. Lyu. Batch mode active learning and its application to medical image classification. In Proceedings of the 23rd International Conference on Machine Learning (ICML), pages 417–424. ACM, 2006.
    Google ScholarLocate open access versionFindings
  • [19] S.-J. Huang, R. Jin, and Z.-H. Zhou. Active learning by querying informative and representative examples. In Advances in neural information processing systems, pages 892–900, 2010.
    Google ScholarLocate open access versionFindings
  • [20] ITU. ITU-T Recommendation G.826-Error performance parameters and objectives for international, constant bit rate digital paths at or above the primary rate. https://www.itu.int/rec/T-REC-G.826.
    Findings
  • [21] R. M. Khanafer, B. Solana, J. Triola, R. Barco, L. Moltsen, Z. Altman, and P. Lazaro. Automated diagnosis for umts networks using bayesian network approach. IEEE Transactions on vehicular technology, 57(4):2451–2461, 2008.
    Google ScholarLocate open access versionFindings
  • [22] T. Kimura, A. Watanabe, T. Toyono, and K. Ishibashi. Proactive failure detection learning generation patterns of large-scale network logs. IEICE Transactions on Communications, 2018.
    Google ScholarLocate open access versionFindings
  • [23] K. Konyushkova, R. Sznitman, and P. Fua. Learning active learning from data. In Advances in Neural Information Processing Systems, pages 4225–4235, 2017.
    Google ScholarLocate open access versionFindings
  • [24] D. D. Lewis and J. Catlett. Heterogenous uncertainty sampling for supervised learning. In Machine learning proceedings, pages 148–156.
    Google ScholarLocate open access versionFindings
  • [25] D. D. Lewis and W. A. Gale. A sequential algorithm for training text classifiers. In Proceedings of the 17th Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval (SIGIR), pages 3–12.
    Google ScholarLocate open access versionFindings
  • [26] B. Liu, Y. Dai, X. Li, W. S. Lee, and P. S. Yu. Building text classifiers using positive and unlabeled examples. In Proceedings of the 3th International Conference on Data Mining (ICDM), pages 179–186. IEEE, 2003.
    Google ScholarLocate open access versionFindings
  • [27] D. Liu, Y. Zhao, H. Xu, Y. Sun, D. Pei, J. Luo, X. Jing, and M. Feng. Opprentice: Towards practical and automatic anomaly detection through machine learning. In Proceedings of the 2015 Internet Measurement Conference (IMC), pages 211– 224. ACM, 2015.
    Google ScholarLocate open access versionFindings
  • [28] F. T. Liu, K. M. Ting, and Z.-H. Zhou. Isolation forest. In Proceedings of the 8th International Conference on Data Mining (ICDM), pages 413–422. IEEE, 2008.
    Google ScholarLocate open access versionFindings
  • [29] H. B. Mann and D. R. Whitney. On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics, 18(1):50–60, 1947.
    Google ScholarLocate open access versionFindings
  • [30] L. Pan, J. Zhang, P. P. Lee, H. Cheng, C. He, C. He, and K. Zhang. An intelligent customer care assistant system for large-scale cellular network diagnosis. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge discovery and data mining, pages 1951–1959. ACM, 2017.
    Google ScholarLocate open access versionFindings
  • [31] B. Perozzi, R. Al-Rfou, and S. Skiena. Deepwalk: Online learning of social representations. In Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 701–710. ACM, 2014.
    Google ScholarLocate open access versionFindings
  • [32] T. Pitakrat, D. Okanovic, A. van Hoorn, and L. Grunske. Hora: Architectureaware online failure prediction. Journal of Systems and Software, 137:669–685, 2018.
    Google ScholarLocate open access versionFindings
  • [33] F. Salfner, M. Lenk, and M. Malek. A survey of online failure prediction methods. ACM Computing Surveys (CSUR), 42(3):10, 2010.
    Google ScholarLocate open access versionFindings
  • [34] B. Settles. Active learning. Synthesis Lectures on Artificial Intelligence and Machine Learning, 6(1):1–114, 2012.
    Google ScholarLocate open access versionFindings
  • [35] H. S. Seung, M. Opper, and H. Sompolinsky. Query by committee. In Proceedings of the 5th Annual Workshop on Computational Learning Theory, pages 287–294. ACM, 1992.
    Google ScholarLocate open access versionFindings
  • [36] L. H. Shuan, T. Y. Fei, S. W. King, G. Xiaoning, and L. Z. Mein. Network equipment failure prediction with big data analytics. International Journal of Advances in Soft Computing & Its Applications, 8(3):59–69, 2016.
    Google ScholarLocate open access versionFindings
  • [37] M. A. Siddiqui, A. Fern, T. G. Dietterich, R. Wright, A. Theriault, and D. W. Archer. Feedback-guided anomaly discovery via online optimization. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge discovery and data mining, pages 2200–2209. ACM, 2018.
    Google ScholarLocate open access versionFindings
  • [38] P. Szilagyi and S. Novaczki. An automatic detection and diagnosis framework for mobile communication systems. IEEE Transactions on Network and Service Management, 9(2):184–197, 2012.
    Google ScholarLocate open access versionFindings
  • [39] J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei. Line: Large-scale information network embedding. In Proceedings of the 24th International Conference on World Wide Web (WWW), pages 1067–1077. International World Wide Web Conferences Steering Committee, 2015.
    Google ScholarLocate open access versionFindings
  • [40] K. Veeramachaneni, I. Arnaldo, V. Korrapati, C. Bassias, and K. Li. AI2: Training a big data machine to defend. In Proceedings of the 2nd International Conference on Big Data Security on Cloud (BigDataSecurity), pages 49–54. IEEE, 2016.
    Google ScholarLocate open access versionFindings
  • [41] D. Wang, P. Cui, and W. Zhu. Structural deep network embedding. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, pages 1225–1234. ACM, 2016.
    Google ScholarLocate open access versionFindings
  • [42] Z. Wang and J. Ye. Querying discriminative and representative samples for batch mode active learning. ACM Transactions on Knowledge Discovery from Data (TKDD), 9(3):17, 2015.
    Google ScholarLocate open access versionFindings
  • [43] Z. Wang, M. Zhang, D. Wang, C. Song, M. Liu, J. Li, L. Lou, and Z. Liu. Failure prediction using machine learning and time series in optical network. Optics Express, 25(16):18553–18565, 2017.
    Google ScholarLocate open access versionFindings
  • [44] J. Wu, P. P. Lee, Q. Li, L. Pan, and J. Zhang. CellPAD: Detecting performance anomalies in cellular networks via regression analysis. In Proceedings of the IFIP Networking Conference (IFIP Networking), 2018.
    Google ScholarLocate open access versionFindings
  • [45] Y.-Y. Yang, S.-C. Lee, Y.-A. Chung, T.-E. Wu, S.-A. Chen, and H.-T. Lin. libact: Pool-based active learning in python. Technical report, National Taiwan University, Oct. 2017. available as arXiv preprint https://arxiv.org/abs/1710.00379.
    Findings
  • [46] Z. Yuan, Y. Li, C. Peng, S. Lu, H. Deng, Z. Tan, and T. Raza. A machine learning based approach to mobile network analysis. In Proceedings of the 27th International Conference on Computer Communication and Networks (ICCCN), pages 1–9. IEEE, 2018.
    Google ScholarLocate open access versionFindings
  • [47] S. Zhang, Y. Liu, W. Meng, Z. Luo, J. Bu, S. Yang, P. Liang, D. Pei, J. Xu, Y. Zhang, Y. Chen, H. Dong, X. Qu, and L. Song. PreFix: Switch failure prediction in datacenter networks. Proceedings of the ACM on Measurement and Analysis of Computing Systems (POMACS), 2(1):2, 2018.
    Google ScholarLocate open access versionFindings
  • [48] Y. Zhang, P. Zhao, J. Cao, W. Ma, J. Huang, Q. Wu, and M. Tan. Online adaptive asymmetric active learning for budgeted imbalanced data. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 2768–2777. ACM, 2018.
    Google ScholarLocate open access versionFindings
  • [49] X. Zhao, X. Luo, T. Wu, and D. Xiao. The prediction mathematical model for HO performance in LTE networks. In Proceedings of 8th International Conference on Networking, Architecture and Storage (NAS), pages 191–197. IEEE, 2013.
    Google ScholarLocate open access versionFindings
  • [50] I. Zliobaite, A. Bifet, B. Pfahringer, and G. Holmes. Active learning with drifting streaming data. IEEE Transactions on Neural Networks and Learning Systems, 25(1):27–39, 2013. Patrick P. C. Lee received the B.Eng. degree (first-class honors) in Information Engineering from the Chinese University of Hong Kong in 2001, the M.Phil. degree in Computer Science and Engineering from the Chinese University of Hong Kong in 2003, and the Ph.D. degree in Computer Science from Columbia University in 2008. He is now an Associate Professor of the Department of Computer Science and Engineering at the Chinese University of Hong Kong. His research interests are in storage systems, distributed systems and networks, and cloud computing.
    Google ScholarLocate open access versionFindings
  • Pinghui Wang received the B.S. degree in information engineering and the Ph.D. degree in automatic control from Xi'an Jiaotong University, China, in 2006 and 2012 respectively. He is currently an associate professor in MOE Key Laboratory for Intelligent Networks and Network Security at Xi'an Jiaotong University, China. His research interests include Internet traffic measurement and modeling, traffic classification, abnormal detection, and online social network measurement.
    Google ScholarFindings
Your rating :
0

 

Tags
Comments