A Skeleton Pattern Representation Method for Anomaly Detection in Wireless Sensor Networks

Anomaly detection of time series in wireless sensor networks has gained significant attention. Researchers employ representation techniques to reduce the dimensionality of time series. The Piecewise Aggregate Pattern Representation (PAPR) method vertically divides the time series and represents as patterns. However, the computational cost of the PAPR method for anomaly detection is high. To overcome this issue, we propose a time series representation method for anomaly detection called Skeleton Pattern Representation for anomaly detection (SPR-AD). An improved Perceptually Important Point (iPIP) identification algorithm is devised to transform the subsequence into a skeleton with PIPs. Then the skeleton is divided into several equal-sized subspaces in the amplitude domain. Statistical information of the PIPs located in subspaces is constructed to represent the skeleton as a pattern. Based on the two-phase representation scheme, anomaly is determined by anomaly scores which are based on the of similarity among pattern. The experimental results demonstrate the performance of the SPR-AD method is better in terms of computational cost and detection accuracy than the PAPR method.

[1]  Andreas Dengel,et al.  Pattern-Based Contextual Anomaly Detection in HVAC Systems , 2017, 2017 IEEE International Conference on Data Mining Workshops (ICDMW).

[2]  Yifeng Guo,et al.  Sensor placement for lifetime maximization in monitoring oil pipelines , 2010, ICCPS '10.

[3]  Pawel Kulakowski,et al.  Performance study of wireless sensor and actuator networks in forest fire scenarios , 2013, Int. J. Commun. Syst..

[4]  Chonghui Guo,et al.  An Improved Piecewise Aggregate Approximation Based on Statistical Features for Time Series Mining , 2010, KSEM.

[5]  Witold Pedrycz,et al.  A Piecewise Aggregate pattern representation approach for anomaly detection in time series , 2017, Knowl. Based Syst..

[6]  Seok-Ju Chun,et al.  Representation and clustering of time series by means of segmentation based on PIPs detection , 2010, 2010 The 2nd International Conference on Computer and Automation Engineering (ICCAE).

[7]  Eamonn J. Keogh,et al.  Dimensionality Reduction for Fast Similarity Search in Large Time Series Databases , 2001, Knowledge and Information Systems.

[8]  Hwang Soo Lee,et al.  Wireless sensor network design for tactical military applications : Remote large-scale environments , 2009, MILCOM 2009 - 2009 IEEE Military Communications Conference.

[9]  Michael E. Fitzpatrick,et al.  Anomaly detection in time series data using a combination of wavelets, neural networks and Hilbert transform , 2015, 2015 6th International Conference on Information, Intelligence, Systems and Applications (IISA).

[10]  Saeed Amizadeh,et al.  Generic and Scalable Framework for Automated Time-series Anomaly Detection , 2015, KDD.

[11]  Mohamed Abid,et al.  Outlier detection approaches for wireless sensor networks: A survey , 2017, Comput. Networks.

[12]  Eamonn J. Keogh,et al.  HOT SAX: efficiently finding the most unusual time series subsequence , 2005, Fifth IEEE International Conference on Data Mining (ICDM'05).

[13]  VARUN CHANDOLA,et al.  Anomaly detection: A survey , 2009, CSUR.