Outlier detection technique using ct-ocsvm and fuzzy rule-based system in wireless sensor networks

الملخص EN

The development of Wireless Sensor Networks (WSNs) has been attained in the past few years due to its important using in wide range of application.

The readings of data derived from WSN nodes are not always accurate and may contain abnormal data.

This paper proposed an anomaly detection and classification algorithm in WSNs.

At first, an integration of Contourlet Transform (CT) algorithm and One Class Support Vector Machine (OCSVM) algorithm (CT-OCSVM) is utilized to detect outliers then Fuzzy Inference System (FIS) is used to identify the source of these outliers.

The underlying aim of this paper focuses on treating the collected streams of data as raw datum of an image, which is then passed through some filters using CT to get compressed size of directional subbands coefficients.

The coefficients of CT are examined by OCSVM algorithm to detect anomalies.

Finally the source of anomalies is identified based on using FIS and by exploiting the spatial temporal correlation existing between the sensed data.

The integrated algorithm is tested using different types of filters.

Real datasets collected from a small WSN constructed in a local lab are used for testing the integrated algorithms.

The simulation results have shown a high rate of accurate classification with high detection rate and low false alarm has been attained in the past few years due to its important using in wide range of application.

The readings of data derived from WSN nodes are not always accurate and may contain abnormal data.

This paper proposed an anomaly detection and classification algorithm in WSNs.

At first, an integration of Contourlet Transform (CT) algorithm and One Class Support Vector Machine (OCSVM) algorithm (CT-OCSVM) is utilized to detect outliers then Fuzzy Inference System (FIS) is used to identify the source of these outliers.

The underlying aim of this paper focuses on treating the collected streams of data as raw datum of an image, which is then passed through some filters using CT to get compressed size of directional subbands coefficients.

The coefficients of CT are examined by OCSVM algorithm to detect anomalies.

Finally the source of anomalies is identified based on using FIS and by exploiting the spatial temporal correlation existing between the sensed data.

The integrated algorithm is tested using different types of filters.

Real datasets collected from a small WSN constructed in a local lab are used for testing the integrated algorithms.

The simulation results have shown a high rate of accurate classification with high detection rate and low false alarm rate.