Article Preview
Top1. Introduction
The connection of the internet with an embedded computer system coupled with sensors and actuators brought about the notion of the internet of things (IoT). The name ‘Internet of Things’ developed from a presentation in the Procter & Gamble corporation. The presenter showed that the company could use the internet to gather information about clients without any human assistance and called it the internet of things (Ibarra-Esquer et al., 2017). The popularity developed from there and confined until now when people are linking it to virtual cyberspace and believe that it can make human life easier. The internet of things now has a close connection to location and tracking, sensing, actuation, and processes. This technology avails an ability to detect items through the use of a digital tag or serial numbers. They also check the physical location of people and items in the world through various Geo-location devices and offer communication capabilities. Apart from all these functions, IoT allows the actuation of devices and the ability to modify the external physical environment. In the same way the system can process commands sent through the internet and lead to real benefits for the user. The internet of things possesses similar challenges of security and privacy (Ouaddah et al., 2017). The growing market for IoT also attracts malicious individuals trying to gain access to the marketplace (Bertino & Islam, 2017). This new technology comes with difficulties in protecting the privacy and safety of data. Organizations with interest in the IoT technology should train their staff about safety measures and ways of detecting cyber-attacks.
Software defined networking is a new network paradigm that improves security in IoT networks. The centralized controller in SDN manages the network and controls the data flow in the network elements. It has received significant attention from industry and researchers, and it has been deployed in different scenarios and environments (Lee et al., 2020). The network traffic analysis process consists of packet recording and analyzing the behavior of network traffic communication to detect attacks in the IoT network (Banerjee & Samantaray, 2019; Sultana et al., 2019). The threat landscape evolution, in coincidence with the increasing dependence on modern technology, has occasioned a structural transformation for the development of the incident retort framework, while the IT would continuously play an essential part in response action, its responsibility has reformed for a caring one. Nowadays, with the upsurge in the utilization of technologies within the workroom, cybercriminals can get access to company networks. (Shea et al., 2019). As a consequence, the whole cyber battleground has changed and turned out to be far more multifaceted hence the need for an incident response framework.
Artificial intelligence improved system security against cyber, reactive defences are not enough to halt attackers from accessing even the top security architectures. (Desouza et al., 2020) Industrial IoT has special communication requirements including high reliability, low latency, flexibility, and security as provided by the 5G mobile technology and growth is the foundation of development in the modern age. (Varga et al., 2020). Ensuring the safety entails shielding both IoT services and devices from illegal access from external and in the devices, confidentiality is a significant security aspect of the IoT network (Siboni et al., 2019).
Deep learning and machine learning are a great addition to the IoT era. Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbour (KNN), Logistic Regression (LR) are the prominent machine learning approaches used in attack and anomaly detection in IoT network (Hasan et al., 2019). RF algorithm combines bagging) and random feature to create a series of controlled variation decision trees. In K-NN classification, a classifier returns a class membership. SVM seeks a hyperplane, which best splits, the various classes (Fenanir et al., 2019). SVM trains data to find multiple support vectors, which define the hyperplane. The prediction only relies on the support vectors. Besides linear classification. SVM is considered to be more robust in performance (Sciancalepore et al., 2020).