|
Abstract
|
Among security protection methods, the presence of IDS intrusion detection systems enables the protection of integrity, privacy and security of data sent through the network, and therefore, play an important role in dealing with security attacks. The tasks of preventing partial or complete attacks on the IoT network or preventing, detecting, reacting and reporting malicious activities are performed by IDS [3]. In previous years, it was challenging to implement intrusion detection systems that used traditional algorithms based on statistical methods on big data, because these systems were generated through behavior-based statistical methods and summaries based on previous rules. used to model behaviors in the network. But these methods are not practical and accurate to a large extent, and they are replaced by machine learning (ML) methods, which are the most common methods for intrusion detection [4]. Machine learning methods can enhance experiential learning and decision-making processes of various systems by improving skills and capacity without open programming. Continuing the development of machine learning methods, intelligent sub-branches are produced using deep learning methods, in which advanced and updated datasets are preferred to perform a greater number of intrusion detections. Intrusion detection systems based on deep learning are usually trained using the latest datasets produced for intrusion detection. In addition, there are different types of attacks and allocations in the updated data set. The accurate intrusion detection rate of deep learning architectures usually increases as the number of features in a dataset increases [5].
In this research, we focus on the problem of attack detection in Internet of Things networks to achieve joint communication, energy efficiency and maximize network capacity using Graph Neural Networks (GNN). This research develops a hybrid intrusion detection system by combining feature selection and graph neural network.
|