Research On Sparse Protocol Sample Parsing Technology Of Industrial Internet Of Things

With the rapid development of industrial digitalization and networking,Industrial Internet of things(IIoT)has arisen to build an interconnected and smart industry.As the nerve center of national infrastructure,IIoT needs high security communication protocol interconnection,so protocol security analysis is particularly important.The premise of protocol security analysis is to know the protocol format.However,a large number of unknown and private protocols exist in IIoT in order to optimize communication performance or provide personalized functions for IIoT equipment manufacturers.The existing methods of unknown protocol prasing method mainly include static analysis and dynamic analysis.Dynamic analysis method needs to obtain executable protocol processing program,but it is difficult to obtain or affect the operation of hardware equipment in IIoT environment.Static analysis method takes the network data flow as the analysis object,and the accuracy of protocol parsing depends on the number of samples.However,in the relatively private IIoT environment,it is difficult to capture a large number of protocol data,and there is no public data set,which makes the protocol samples that can be analyzed less,that is,the samples are sparse.Therefore,it is necessary to design an analysis method of IIoT sparse protocol samples.The research contents of this paper are as follows:(1)This paper proposes a sparse protocol parsing model based on binary particle swarm optimization algorithm(BPSO+HMM).Firstly,the model designs the fitness function of BPSO algorithm according to the range of protocol fields,and updates the individual optimal and global optimal by comparing the fitness values among individuals until the value boundary of each field is no longer changed,and outputs the expanded high-quality protocol samples.Then the protocol field structure is used as the hidden state of hidden Markov model(HMM),and the expanded protocol samples are used as observation sequences to train HMM parameters.Then the maximum likelihood probability of the hidden state is estimated by Viterbi algorithm based on HMM,and the best IIoT protocol field format is predicted.(2)This paper proposes a sparse protocol parsing model based on genetic algorithm(GA+HMM).Firstly,the fitness function of GA is designed based on the feedback data,and new samples are expanded by crossover and mutation operations.Then,high-quality samples are selected by roulette operation.When the average fitness value of the population reaches the preset range,the expanded high-quality protocol samples are output,and then the parameters of HMM are trained by the expanded protocol samples,which improves the accuracy of IIoT protocol analysis.(3)This paper designs and implements an automatic and high accuracy IIoT sparse protocol parsing system.Taking Modbus,S7 Comm,IEC104 and MQTT protocols as examples,through experiments,the expansion effects of the two protocol expansion models are compared,and then the parsing accuracy of BPSO+HMM,GA+HMM,GA+RNN and GA+LSTM are compared to verify the feasibility of the system.