Design And Implementation Of Intrusion Detection System For Smart Home

With the improvement of living condition,smart home devices are gradually favored by people.For those smart home devices with poor security and defense capabilities,they are often the focus of hacker attacks.Deploying IDS on the home gateway can identify the intrusion behavior,prevent the smart home device from being attacked arbitrarily,and then protect the security of the home network.Traditional machine learning methods are widely used in intrusion detection,which are simple to train,but lack in performance.In recent years,the deep learning method has developed rapidly in intrusion detection,and the detection rate is also getting higher and higher,but it also requires a stronger processor,so it is usually deployed in the cloud.If all home traffic is sent to the cloud for detection,it will not only waste the home bandwidth,but also put great pressure on the cloud server.In view of the above problems,based on the Three-way Decision theory,this paper adopts the strategy of combining gateway and cloud to implement an IDS for smart home.The main work includes:(1)Build a Random Forest Three-way Decision intrusion detection model deployed in the gateway to preliminarily identify the traffic.Use SMOTE-Tomek algorithm to solve the data imbalance problem and improve the accuracy of the model;Based on the idea of Three-way Decision,modify the combination mechanism of Random Forest,and construct the Random Forest Three-way Decision model(RF-TWD);Classify the network behavior into normal behavior,different types of attack behavior and suspicious behavior by using RF-TWD model.(2)Build a BiLSTM intrusion detection model deployed in the cloud to further identify suspicious behaviors.BiLSTM is used to build a deep neural network model,Batch normalization is introduced to accelerate the model training speed,Dropout is introduced to improve the detection performance of the model,and Softmax is used to achieve multi classification;At the same time,transfer learning is introduced to avoid the problem of too little suspicious behaviors.(3)Build an IDS for smart home with the RF-TWD model and BiLSTM model as the detection engine.Collect and analyze the network traffic in real time,and extract the required features.;Identify the extracted features with RF-TWD model deployed in the gateway and BiLSTM model deployed in the cloud.