Efficient Security Solutions for IoT Devices

Abstract

The Internet of Things (IoT) is a technological innovation that has revolutionized society. It has affected all aspects of life, from our homes and our cars, to our workplaces, where it has helped us to become more productive. The IoT will forever change the way we use simple things that do very little things to smart, fully capable things. IoT devices are able to process and automate everyday household and workplace tasks through simple sensors. Yet despite the benefits of these devices, there is the problem of these devices tends to use most of their resources and computing power for the application’s key features; thus, ensuring protection and privacy at a lower cost is a must, not to mention they are vulnerable to violations such as privacy issues and security breaches. The purpose of this thesis is to provide a solution to protect IoT devices from security breaches while requiring minimal computing resources that do not compromise the key features of the application. The current threats are investigated, as well as the reasons why IoT devices are vulnerable. Moreover, the technologies used in the IoT are examined, as well as the different communication layers of the IoT and their functioning. The findings reveal that IoT devices are prone to many software and hardware vulnerabilities. Solutions to these challenges are proposed, notably through the use of anomaly-based intrusion detection systems and machine learning (ML), Many proposed anomaly-based detection systems use different ML algorithms and techniques. However, this research aim to enhance the efficiencty of anomaly-based detection systems to be suitable for IoT environment. there is no standard benchmark to compare the power consumption. A benchmark that measures both accuracy and power consumption to calculate and evaluate each algorithm’s implementation is proposed. The benchmark was used to compare six different ML algorithms on the UNSW NB15 dataset. Feature selection was proposed to be used as an efficient solution for IoT. the proposed solution allowed achieving 98.49% accuracy while only taking 0.06s to finish the test stage which considers 4 times faster if we didn't use the proposed solution.