A SMART AI-DISTRIBUTION MODEL FOR IOT DATASETS

Abstract

A shortage of distributed networks has created a mass of heterogeneous sources of data that encompass the actual activity of real-world Internet of Things (IoT) surroundings and convoluted conditions of threat as to evaluate the authenticity of the newer technologies. Although the need to examine the context of cyber risk to the IoT network infrastructure and development of Artificial Intelligence (AI) based safeguards has been increasing. This paper would present a new IoT testbed infrastructure that would be utilized in evaluating security systems that take advantage of Intelligence. It has been simplified to implement Network Function Virtual (NFV), Software-Defined Networks (SDN) and Service Orchestration that offer customizable test-bed systems that facilitate interaction among edges, fog, and clouds tier using the framework NSX vCloud NFV. Normal and malicious threat scenarios are conducted to collect tagged data sources as a framework is being implemented. The developed data is referred to as "TonIOT" because they encompass various methods of data collection such as internet traffic data, Windows OS and Linux-based data sources which comprise IoT application telematics data sources. Some machine learning-based attack detection methods are Gradient Boosting Machines, random forest, Naive bayes, and Neural Networks, which are applied to analyze the database of TonIOT net, which demonstrates a good quality of accuracy rate when applying the set of training and testing data. The diversity of the legal and abnormal behavior of TonIOT networking dataset further validates the analysis of AI-based security mechanisms with numerous other similar networked datasets.