Research On Intrusion Detection Methods For Smart Home Based On Multilayer Neural Network

Relying on Internet of things,cloud computing,big data,artificial intelligence and other technologies,smart home is rapidly developed.But smart home is facing to risks in information security such as denial of service,illegal access,or disclosure of privacy even though it brings people convenience.Research on intrusion detection for smart home is in its primary stage,the existing solutions still have difficulties in handling a large number of high-dimensional data and have questions such as low detection rates,and high rates of false positives.Based on the analysis of the security threats and characteristics of the network itself,This thesis establishes the goal of high testing accuracy and low latency,and explores the smart home intrusion detection methods.The main contributions of this thesis are include:1.Present a multi-layer neural network intrusion detection method that combines deep learning and fuzzy neural network,and builds a smart home intrusion detection model based on multi-layer neural network.The learning of data features is completed based on deep learning,high-dimensional data is mapped to low-dimensional data,and the category of low-dimensional data is judged based on fuzzy neural network analysis.2.The traditional method relies on the experience to determine the number of network layers,and provides a method for optimizing and determining the network depth.Based on the network reconfiguration error,the depth of the network can be judged and the network can be trained from self-organization.It can not only guarantee the detection accuracy,but also reduce the time complexity of the algorithm.3.Set up a smart home intrusion detection and verification system composed of four modules: data acquisition,data preprocessing,intrusion detection and local alarm.Winpcap captures and parses the packets of the server network card.Based on the statistical characteristics of the packet header and content information,the string is converted into a value and normalized data is obtained to be detected.The data is used as an input of the intrusion detection algorithm with multi-layer neural network to deduce certain results of judgement,and then the results are outputted and the certain response is triggered based on any attack results.The simulation test results show that the smart home intrusion detection scheme which introduces a new generation of artificial intelligence methods such as deep learning can effectively improve the detection accuracy and detection efficiency of the attack behavior.For example,the typical KDDCUP99 data set is tested in MATLAB simulation software for denial of service.The detection rate of attacks and remote illegal access can reach 94%,and the detection rate of new types of attacks in the network exceeds 60%.Based on the built-in smart home intrusion detection system,the verification test results show that the average delay for Distributed Denial of Service(DDoS)attacks detection is 6s,and the detection accuracy rate can reach 96%.