Analysis and detection of routing attacks in the internet of Things using Deep learning

Abstract

Internet of things applications is growing day by day, as they are being used in many areas and systems, and as their uses and modes of employment increase, there are many gaps with them, the most important problem is security. IOT has a large number of connected devices and therefore mobile data traffic is large and routing protocols are a key element. IOT has many routing protocols, the most widely used is RPL protocol, which takes into account limited power and the device’s capabilities, but it suffers from several weaknesses, the most important one is routing based attacks which targeting this protocol. In this work, we aim to solve the problem of Internet of Things exposure to RPL-based attacks as routing protocol. We built an anomaly intrusion detection system based on deep learning and an IoT attacks dataset (Minerva-IoT) containing the most important attacks built through Cooja simulator and implementation of different scenarios that allowed for the extraction of important features with the addition of new sensitive features such as nodes power and their geographical location, balancing the dataset by fix minority classes (rare attacks) to avoid fake performance using smart algorithms. The results were very satisfactory after the most important challenges in intrusion detection systems were achieved from a false alarm rate (false positive), accuracy and precision.