Research On The Detection And Discovery Of Malicious Nodes In The Internet Of Things Based On Deep Learning

In the information age,the Internet of things(Hereinafter abbreviated as IOT)has become the infrastructure of smart city.We are surrounded by the IOT.Some crucial data about the national economy and people's livelihood are facing the security risks of being leaked and misused.In recent years,data leakage occurs from time to time,and most attackers steal data by controlling IOT nodes.Research shows that Mirai,the malicious code,is the major malware that controls IOT nodes.Therefore,this paper proposes a scheme to detect the security state of IOT nodes based on deep learning algorithm and takes Mirai detection as an example.In addition,this paper also uses the trust value to define the security state of the IOT nodes,and evaluates the security state by calculating the trust value.It is found that there are differences in grayscale images after conversion of different types of traffic.The malicious traffic is preprocessed and the original data is transformed into gray image.Through the feature extraction and classification learning process of the internal structure of neural network,the normal traffic and abnormal traffic in the data set are effectively classified.In order to select the most efficient algorithm,this paper compares the activation function and loss function of 7 kinds of neural networks,and compares the processing results of the same data set.Finally,Resnet-18 is determined as the optimal application model.To ensure the integrity and reliability of the collected nodes data,the nodes state are evaluated by the calculated trust value.As a whole,the state of a node is determined by direct trust value,recommended trust value and statistical trust value,and the trust state of a node is measured comprehensively with subjective factors as the main factor and objective factors as the auxiliary factor.The data type of trust value is changed from floating-point type to integer type,which significantly reduces the amount of calculation and improves the speed of trust value calculation.In addition,by tracking the behavior of nodes in different states,the change process of each state is described in detail.At the same time,the above mechanism is verified by the change of trust value of 3 kinds of nodes.Finally,an autonomous resistance mechanism between nodes is proposed,which can resist external malicious attacks under certain conditions.In this project,the Netlogo simulation platform is used for simulation experiment,and the experimental results show that the work of this paper achieves the expected detection accuracy.