Design and application of a safe and energy efficient new method at the network layer in the internet of things

Over the past decade, attacks targeting the Internet of Things (IoT) have increased to a worrying extent with the widespread use of mobile devices and applications connected to the Internet. Routing Protocol for Low-Power and Lossy Network (RPL) allows packets to be routed between nodes in compliance with the desired purpose for the Wireless Sensor Network at the network layer. RPL, which has an essential role and is difficult to secure, is vulnerable to several attacks. These attacks have a negative impact on data transmission between nodes and, by consuming resources, cause great destruction of the topology. Hello Flooding (HF) attacks against RPL result in the consumption of limited resources (memory, processing and energy) in nodes. For this reason, there is a need for an effective method to detect and prevent HF attacks on RPL. Deep learning methods can be successfully applied to intrusion detection systems. In this thesis, Gated Recurrent Unit deep learning method, an improved version of Recurrent Neural Networks, has been proposed to predict and prevent HF attacks on RPL in IoT networks. The proposed approach has been compared with the two most commonly used classification algorithms, the Support Vector Machine and Logistic Regression, and it has been presented in this thesis that has been given more promising results in detecting HF attacks from RPL flood attacks which consume limited energy in nodes. In addition, detection and prevention of HF attacks has been performed with a much lower failure rate than literature research.