Hybrid Approach to Improve Accuracy of Intrusion Detection System in IoT Environment

Abstract

Internet of Things (IoT) is an interconnected system of devices that encourage continuous exchange of data between devices. Experts predict that by now, wireless network traffic size accounts for two third of total Internet traffics to be generated by 50 billion Wi-Fi and cellular connected devices [1]. These devices could be agricultural, medical, healthcare, smart city, driverless vehicles, industrial robots or wearables that can be remotely controlled and configured [1]. IoT devices are expected to become more popular than mobile devices and will have access to the most vital data [3]. This will cause in rise attack surface area and probabilities of attacks. Internet of Things (IoT) is recent technology and system that has the capacity to change our way of life. As IoT becomes more pervasive every day, probabilities of an attack against it increase. Having such a boom in the number of these devices by 2020, marks the question of what sort of sensitive data these devices will be able to communicate. Such data can be environmental, financial, medical, or any other crucial data. Security attacks against IoT devices are increasing every year. September 14, 2019, about 2.9 billion security attacks were measured [4]. Hence, some form of intrusion detection can be used to alert security attack. This work focuses on developing intrusion detection and prevention system for iot network.IT involves combining anomaly based and signature based modules. Signature based module is implemented using iptables firewall and anomaly based module is implemented using deep learning algorithm. Raw data collected from iot LAN containing 2 Raspberry Pi, 1 smart phone, 1 CCTV camera and one laptop. The collected data is preprocessed and trained on Google collaborator. The trained model is tested and evaluated using different performance measurement techniques and got 99.96% accuracy. Finally, the deep neural model is integrated with anomaly based module to maximize the accuracy of intrusion detection. This works focus on detection and preventing of doss attacks in IOT network.