IoT Intrusion Detection Model Based On XGBoost Method

The Internet of Things(Io T)refers to the exchange of information and communication through various information sensing devices that enable any object to be interconnected with the network according to established network protocols.As the network of interconnected devices continues to expand,Io T devices have been widely used in smart homes,healthcare,blockchain,etc.In the Io T environment,sensory data has basic characteristics such as large quantity,complex type,and timeliness.The huge amount of data brings a series of security and privacy hazards to Io T,and establishing effective intelligent algorithms to identify abnormal behaviors in the data has become an important research trend in the security field.In order to safeguard data from tampering and privacy from theft in Io T,intrusion detection techniques are widely used in the Io T environment,but traditional detection means do not focus on the existence of imbalance in the amount of data for different attack types,which leads to the inability to effectively detect and classify attack types with small sample size.In addition traditional detection means are often feature-based or behavior-based and cannot be accurately classified to meet the needs of Io T intrusion detection.According to the problems mentioned above,this paper proposes an Io T intrusion detection model based on the XGBoost method to solve the problem of inaccurate classification caused by the imbalance of detection data.The main research contents are as follows:(1)The Io T intrusion detection system is divided into two phases: feature selection phase and feature classification phase.Feature selection and classification are performed by combining enhanced deep learning algorithms.To reduce the influence of irrelevant features and eliminate redundant features.(2)The first stage uses the XGBoost method to score the importance of the features in the data set,and finds the balance point between the number of features and the detection accuracy by setting a reasonable threshold,and finally selects an optimal set of features.(3)In the second stage,the traditional stochastic forest algorithm is improved and its weight is optimized to solve the problem of inaccurate classification caused by the imbalance.(4)The validity of the intrusion detection model discussed in this paper is verified by simulation experiments.The simulation results show that the model can effectively select optimal features and detect and classify reasonably.