Anomaly Detection in IoT Frameworks Using Machine Learning

Anomaly Detection in IoT Frameworks Using Machine Learning

Phidahunlang Chyne (North-Eastern Hill University, India), Parag Chatterjee (National Technological University, Argentina & University of the Republic, Uruguay), Sugata Sanyal (Tata Institute of Fundamental Research, India) and Debdatta Kandar (North-Eastern Hill University, India)
DOI: 10.4018/978-1-7998-2444-2.ch004
OnDemand PDF Download:
Available
$29.50
No Current Special Offers
TOTAL SAVINGS: $29.50

Abstract

Rapid advancements in hardware programming and communication innovations have encouraged the development of internet-associated sensory devices that give perceptions and information measurements from the physical world. According to the internet of things (IoT) analytics, more than 100 IoT devices across the world connect to the internet every second, which in the coming years will sharply increase the number of IoT devices by billions. This number of IoT devices incorporates new dynamic associations and does not totally replace the devices that were purchased before yet are not utilized any longer. As an increasing number of IoT devices advance into the world, conveyed in uncontrolled, complex, and frequently hostile conditions, securing IoT frameworks displays various challenges. As per the Eclipse IoT Working Group's 2017 IoT engineer overview, security is the top worry for IoT designers. To approach the challenges in securing IoT devices, the authors propose using unsupervised machine learning model at the network/transport level for anomaly detection.
Chapter Preview
Top

Introduction

The escalated advancement of system innovations and the quick development of the number of heterogeneous networks that are interconnected and which consistently trade vast information volumes have made a prolific ground for the advancement of adaptable cyber-attack classification and in the event of different types of network anomalies. It is estimated that by 2020, the complete number of Internet-associated gadgets being utilized will be 25 to 50 billion. As the numbers develop and advances become more developed, the volume of information distributed will increment. Internet associated gadgets innovation alluded to as the Internet of Things (IoT), keeps on expanding the current Internet by giving network and connection between the physical and digital universes. Notwithstanding expanded volume, the IoT creates Big Data portrayed by speed in terms of location dependency and time, with an assortment of various modalities and fluctuating information quality. Intelligent analysis and processing of this Big Data is the way to creating brilliant IoT applications. IoT and comparative conditions are at a very fascinating momentum of both the network of inventions and their developments. Tan and Wang(2010), and Wu et al. (2010) discussed in their papers that IoT alludes to an ongoing worldview that has quickly made progress in the present day wireless communications. IoT is then another innovative trend joining new computing and communication standards. Within this new trend, there are smart gadgets that have an advanced substance and are universally interconnected on a system and to the worldwide Internet (Armentano et. al., 2018). Regular items may incorporate knowledge and the capacity to sense, translate and respond to their condition, joining the Internet with rising innovations, for example, the used of Radio Frequency Identification (RFID) as mentioned by Sharma and Siddiqui(2010) as well as Ziegler and Urbas (2011), real-time locations and installed sensors. The IoT idea depends on the thought of a widespread nearness of 'things' or 'articles, for example, RFID labels, sensors, actuators, cell phones, and so on, with computerized recognizable proof and tending to plans that empower them to coordinate with neighbours so as to accomplish a few shared objectives. In the business area, the most evident results of IoT may emerge in modern computerization and manufacturing, in coordination, in business or procedure management and in astute plans for intelligent transportation. Let us explore IoT in the next subsections.

IoT Architecture

IoT is a blend of embedded technologies including wired and wireless communications, sensor and actuator gadgets, and the physical articles associated with the Internet as mentioned by Atzori et al.(2010) and Cecchinel et al. (2014). One of the long-standing targets of computing is to streamline and enhance human activities and experiences. IoT requires data to either represent better services to users or improve the IoT system execution to achieve this intelligently. Thusly, frameworks ought to have the capacity to access raw information from various assets over the system and investigate this data in request to extricate learning.

IoT requires intelligent processing and reliability inside the system. To give this, system engineering contains three layers: the sensing layer, the transport layer, and the application layer. The detecting/sensing layer contains edge devices that are made out of an assortment of sensors and actuators that gather information and send it through the transportation layer to the application layer for analysis. The transport layer comprises of system communications including Wi-Fi, Bluetooth, ZigBee, and 802.15.4. The transport layer contains the gateways/passages that process the data also, hand-off the data over the system. The application layer contains the logical connection between the user and the Internet through smart applications. Figure 1 is a pictorial representation of the IoT Architecture.

Figure 1.

IoT Architecture1

978-1-7998-2444-2.ch004.f01

Key Terms in this Chapter

Denial of Service Attacks: A denial-of-service (DoS) attack is an attack intended to close down a machine or system, making it out of reach to its destination. DoS assaults achieve this by flooding the target with traffic or sending it data that prompts an accident.

Markov Chain Model: A Markov chain is a numerical framework that encounters advances starting with one state then onto the next as indicated by certain probabilistic principles.

Inliers: Inliers are characterized as a perception that is clarified by fundamental probability density function.

Data Pre-Processing: Data Pre-processing is a strategy that is utilized to change over the raw information into a clean data collection. At whatever point the information is assembled from various sources it is gathered in a crude configuration which feasible for the analysis. Hence, data pre-processing is necessary.

Botnet: A botnet is a group of nodes associated in a planned manner for vindictive purposes. Every node in a botnet is known as a bot. These bots structure a system of trading off nodes, which is constrained by an outsider and used to transmit malware or spam or to dispatch attacks.

AI/Machine Learning: AI is the logical investigation of algorithms and factual models that computer frameworks use to play out a particular role successfully without utilizing explicit instructions, depending on examples and inference rules. It is viewed as a subset of fabricated brainpower.

Anomaly detection: Anomaly detection (additionally known as outlier discovery) is the identification of uncommon things, occasions, or perceptions, which raise doubts by varying fundamentally from most of the data.

Complete Chapter List

Search this Book:
Reset