Research Of Intrusion Detection System Of The Internet Of Things Based On Neural Network

With the arrival of Internet of Things times, the safety of Transport Layer of the Internet ofThings has got more and more attention. The loss caused by the safety problems in the Internetof Things is more serious than that of the traditional network, not only in the aspect ofinformation materials, but also involved with the entity in the real world. Traditional intrusiondetection technology can’t adapt the complex network environment of the Internet of Things.Therefore, it’s rather urgent to study the intrusion detection system corresponding to the currentInternet of Things safety.The performance of traditional intrusion detection system is rather low in the complexnetwork environment of the Internet of Things, unable to deal with some intrusion accidentsintellectually. Considering these reasons, the thesis decides to study a kind of intrusion detectionsystem model which can work well in the complex environment. The model is designed bycombining the advantages of BP neural network and AGENT technology. Based on thedisadvantages of BP neural network, the thesis decides to optimize weight and learning speed toimprove the performance of the whole system model in the environment of the Internet ofThings.Firstly, the thesis introduces the basic concepts and theory knowledge of intrusion detectionfrom its definition, classification, generalized model and so on, and then describes the relatedconcepts of the AGENT technology and the basic theory knowledge of the neural network ingreater detail.Secondly, the thesis introduces the basic safety problems existing in the Internet of Things,and aiming at the realistic complex network environment proposes the way of combining BPneural network and AGENT technology to solve the problem of low performance of the intrusiondetection system in the complex Internet of Things environment. Because of the defects like lowconvergence rate and local optimum of BP neural network, the thesis proposes applying theGenetic programming algorithm to optimize link weight and learning speed, and then applyingthe optimized BP algorithm to the intrusion detection system. According to the study of thegeneralized model, the thesis presents the intrusion detection system model based on neural network, and introduces the workings of each AGENT unit component and the operating mannerof the whole system in detail.Finally, the study tests and evaluates the model algorithm, and describes the KDDCUP99data set adopted in the test simulation, and considering the actual network environment choosesand primarily disposes the data set. At last, according to the experimental result and its analysis,compare the several kinds of algorithms in the experiment and the model algorithm proposed inthe thesis. The analysis results prove that the model algorithm proposed in the thesis has thebetter performance and can be used in the security defense of the Internet of Things.