Data-Filtering System to Avoid Total Data Distortion in IoT Networking

• Main Authors: Dae-Young Kim, Young-Sik Jeong, Seokhoon Kim
• Format: Article
• Language: English
• Published: MDPI AG 2017-01-01
• Series: Symmetry
• Subjects: data-filtering system; data distortion; naïve Bayesian classifier; Internet of Things (IoT)
• Online Access: http://www.mdpi.com/2073-8994/9/1/16

In the Internet of Things (IoT) networking, numerous objects are connected to a network. They sense events and deliver the sensed information to the cloud. A lot of data is generated in the IoT network, and servers in the cloud gather the sensed data from the objects. Then, the servers analyze the collected data and provide proper intelligent services to users through the results of the analysis. When the server analyzes the collected data, if there exists malfunctioning data, distortional results of the analysis will be generated. The distortional results lead to misdirection of the intelligent services, leading to poor user experience. In the analysis for intelligent services in IoT, malfunctioning data should be avoided because integrity of the collected data is crucial. Therefore, this paper proposes a data-filtering system for the server in the cloud. The proposed data-filtering system is placed in front of the server and firstly receives the sensed data from the objects. It employs the naïve Bayesian classifier and, by learning, classifies the malfunctioning data from among the collected data. Data with integrity is delivered to the server for analysis. Because the proposed system filters the malfunctioning data, the server can obtain accurate analysis results and reduce computing load. The performance of the proposed data-filtering system is evaluated through computer simulation. Through the simulation results, the efficiency of the proposed data-filtering system is shown.