Anomaly Detection in IOT
Abstract
One of the newest and most popular technologies is the Internet of Things (IoT). IoT has an impact on a number of industries, such as smart cities, healthcare, logistics tracking, and the automobile sector. Networks containing IoT devices are being the target of an increasing number of cyberattacks and breaches. By employing machine learning to improve anomaly detection, this paper seeks to strengthen security in IoT networks. This revealed the difficulties and weaknesses in protecting Internet of Things networks. The complexity of IoT networks, the human element, the quantity of devices, and network size are the obstacles. The identified limitations include the dearth of modeling input parameters necessary for anomaly identification in IoT networks, as well as the paucity of research on signature-based intrusion detection systems utilized for anomaly detection. Additionally, the performance of machine learning algorithms on real and standard IoT datasets is not compared. In order to evaluate the anomaly binary classification capabilities of the machine learning methods Neural Networks, Gaussian Naive Bayes, Support Vector Machines, and Decision Trees, this paper generates a dataset and contrasts its outcomes with the KDDCUP99 dataset. The outcomes demonstrate that on the generated IoT dataset, Support Vector Machine and Gaussian Naive Bayes outperform the other models. In this paper, the average execution time was lowered by 58% by reducing the number of characteristics needed by machine learning algorithms for anomaly detection in IoT networks to just five features. This paper examines the ability of CNNwGFC, an improved Convolutional Neural Network model, to identify and categorize irregularities in Internet of Things networks. Compared to the traditional Convolutional Neural Network, our model achieves a 15.34% increase in accuracy for IoT anomaly classification in the UNSW-NB15. Compared to the best accuracy found in the literature, the CNNwGFC multi-classification accuracy (96.24%) is 7.16 percent higher.
