Anomaly Based Intrusion Detection system In IoT Using Deep Learning Techniques

Abstract

The Internet of Things is the network of tiny objects. The main goal of IoT is to connect the objects to the internet for sharing and to communicate with each other without human interference to improve the quality of human life. IoT applications are widely applied in smart homes, smart healthcare, smart city, and smart logistics. The IoT network is simply faced with cyber security challenges. Since the devices of the IoT are tiny objects and they are resource-constrained to install advanced security mechanisms. The designers didn’t consider security prevention but their main goal is addressing the IoT system to the whole world. The anomaly-based intrusion detection system is a mechanism used to monitors the network activity and it alerts an alarm if any deviation is passed from the normal behaviors of the thresholds. However, when it is applied in the IoT network it requires huge computational processing, and battery power, as well as the false alarm rate, is high. The purpose of this study is to develop an anomaly-based intrusion detection system in IoT using deep learning techniques. Recent researchers proved that the combination of the intrusion detection system with a deep learning mechanism is efficient and accurate in countermeasures the limitations of the traditional IDS for IoT systems. To develop the models the used algorithm is a deep neural network(DNN) which creates multiple hidden layers. Deep neural networks can learn in multiple levels, corresponding to different levels of abstraction from the dataset. The dataset used for learning and testing is collected from the IoT network which is combined_IoT3. The combined_IoT3 dataset is comprised of both normal traffic, and DoS attack traffic. The dataset is splitting into training and testing. The new model is generated after learning and testing by the DNN algorithm which is anomaly-based IDS. The result indicates that the accuracy of the model is 99.99 percent and the false alarm rate is decreased to zero percent. The new study outperformed in all metrics from the existing study. According to the results, the model is novel inaccuracy and false alarm rate. Therefore the deep neural network algorithm by combining with IDS is robust and effective with the prominent accuracy for securing the IoT network environment.