CN115801451B - File protection cloud perception management platform double-network isolation method - Google Patents

File protection cloud perception management platform double-network isolation method Download PDF

Info

Publication number
CN115801451B
CN115801451B CN202310043031.4A CN202310043031A CN115801451B CN 115801451 B CN115801451 B CN 115801451B CN 202310043031 A CN202310043031 A CN 202310043031A CN 115801451 B CN115801451 B CN 115801451B
Authority
CN
China
Prior art keywords
information
network
representing
frog
management platform
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202310043031.4A
Other languages
Chinese (zh)
Other versions
CN115801451A (en
Inventor
李燕强
齐少华
马国伟
张泽宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hebei Indoor Technology Co ltd
Original Assignee
Hebei Indoor Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hebei Indoor Technology Co ltd filed Critical Hebei Indoor Technology Co ltd
Priority to CN202310043031.4A priority Critical patent/CN115801451B/en
Publication of CN115801451A publication Critical patent/CN115801451A/en
Application granted granted Critical
Publication of CN115801451B publication Critical patent/CN115801451B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

A dual-network isolation method of a file protection cloud perception management platform relates to the technical field of digital information transmission, and firstly, a file identification system is adopted to identify and collect file information; secondly, controlling network connection by adopting a network control system, and realizing internal and external network connection through an interface connection program; thirdly, an external network management system is adopted to realize the connection between the management platform and the Internet through a second transmitting antenna and a second receiving antenna; thirdly, the intranet management system realizes the connection between the management platform and the internal network through the first transmitting antenna and the first receiving antenna; thirdly, the information processing system realizes network information optimization classification through a neural network-frog-leaping algorithm; finally, the network safety monitoring system monitors the network safety state in real time in the network connection process. The invention greatly improves the archive protection cloud sensing capability and the data communication capability.

Description

File protection cloud perception management platform double-network isolation method
Technical Field
The invention relates to the technical field of digital information transmission, in particular to a dual-network isolation method of an archive protection cloud perception management platform.
Background
The dual-network isolation means that the external network and the internal network cannot interfere with each other, namely, the internal network cannot be called when the external network is called, and the external network cannot be influenced, and the internal network is called; the existing double-network isolation technology can be divided into three types, namely double-machine isolation, a double-network isolation card and a double-network switcher, wherein the double-machine isolation is to configure two computers, one is in charge of internal network connection, and the other is in charge of external network connection; the dual-network isolation card is characterized in that a hard disk is additionally configured, an internal network hard disk is in a power-off state when an external network is connected, and the external network hard disk is in the power-off state when the internal network is connected; the dual-network switcher is a network switching device.
In the prior art, a dual-network isolation technology is gradually advanced, and a dual-network isolation implementation method is proposed in a patent CN201510060188.3, wherein a mode switching instruction sent by an embedded controller is used for controlling a logic chip connected with a network, the first logic chip is used for operating a first internal and external network switching state, the second logic chip is used for operating a second internal and external network switching state, the network connection is controlled to be in a dual-network mode, a dual-operating system mode and a conventional mode, the switching instruction is in a high level, the dual-operating system instruction is in a low level, and the network mode is in the conventional mode; the switching instruction is low level, the double operation instruction is high level, and the network mode is a double operation system mode; in the patent CN201120217067.2, a dual-network isolation system is proposed, which is connected to a network through a dual-network isolation hardware switch, wherein a TCM chip stores network configuration data, a dual-network isolation switch is used as a network selector and connected to the TCM chip and integrated in a computer motherboard, the TCM chip is used for storing gateway data and performing encryption processing on the data, a network connection system is selected, and a BIOS chip obtains the gateway configuration data starting system to implement network selection. Both techniques can realize dual-network isolation, but do not have information processing capability, the isolation effect of the dual-network is not obvious, and the network signal processing efficiency is low.
Disclosure of Invention
Aiming at the problems, the invention discloses a double-network isolation method of an archive protection cloud perception management platform, which can realize effective isolation of internal and external networks, efficiently process network data, optimize information and greatly improve network security.
In order to achieve the technical effects, the invention adopts the following technical scheme:
a dual-network isolation method of a file protection cloud perception management platform is characterized by comprising the following steps of: comprising the following steps:
s1: a PDA classifier is adopted to scan two-dimension code labels of the file bags, basic information in the file bags is identified, and the file protection cloud perception management platform automatically synchronizes the basic information in a database through a GPRS network communication unit arranged inside;
s2: the network control system controls the network connection of the archive protection cloud perception management platform, an operator inputs an internal network connection command at a network connection interface, and the management platform realizes internal work network connection through a special data circuit; an operator inputs an external network connection command at a network connection interface, and the management platform is connected with an external network through an optical fiber internet module to realize network information transmission search;
s3: the management platform inquires file information, inputs a secret key inquiry command, a signal generator generates the frequency of an inquiry signal, a first transmitting antenna transmits the frequency signal, a first receiving antenna receives the frequency signal, the frequency signal is converted into an electric signal through a signal conversion circuit, and a signal processing module calibrates and processes the inquiry command of the electric signal and dynamically invokes file information in a database according to the inquiry command;
s4: the management platform is connected with the internet to inquire information, the main server establishes wireless connection with external subnet units through a second transmitting antenna, communication channels are established among the subnet units with the same communication parameters, the second receiving antenna receives returned information of each subnet unit, and the information processing module classifies and processes the returned information; the information processing module adopts a neural network-frog-leaping algorithm to realize the classification processing of the information, and the neural network-frog-leaping algorithm comprises an information optimization model, an improved frog-leaping algorithm model, a frog-leaping classifier and a data transmission relay station.
S5: the network safety monitoring system carries out safety supervision on the working states of the internal network and the external network of the management platform, the network speed measuring unit calculates the network speed according to the response time, the network data statistics unit calculates the data information in the network nodes and divides the network nodes into stable type, fluctuation type and mutation type according to the characteristics of the data information, the safety management unit carries out safety value statistics on the fluctuation type and mutation type network nodes based on a neural network deep learning algorithm, the safety value is lower than a lowest threshold value, and the network hazard warning unit starts an alarm mode; the neural network deep learning algorithm comprises a BP training model, a function set and a deep statistics module.
As a further technical scheme of the invention, the file bag two-dimension code comprises file numbers, generation time, belonging addresses, file attributes, author information and a two-dimension code history scanning record of the file.
As a further technical scheme of the invention, the file database is stored by adopting a distributed ring topology structure, and files are stored in the distributed ring topology structure in the form of stacks with the same attribute and the same group.
As a further technical scheme of the invention, the archive protection cloud perception management platform queries information through an external network, the information received by the second receiving antenna is classified through a neural network-frog-leaping algorithm, and the method for optimizing information data by the neural network algorithm comprises the following steps:
s41: calculating information energy function
Defining hidden layer information values in a neural network algorithm as
Figure SMS_1
Wherein->
Figure SMS_2
Representing the amount of hidden layer information->
Figure SMS_3
Representing the information value as a vector, the visible layer information value as +.>
Figure SMS_4
Wherein->
Figure SMS_5
The information energy function in the neural network is shown as a formula (1) and represents the information quantity of the visible layers:
Figure SMS_6
(1)
in the formula (1), the components are as follows,
Figure SMS_19
representing the number of bits of the hidden layer information value->
Figure SMS_10
Representing implicit coefficients in the neural network, < >>
Figure SMS_23
Representing display coefficients in a neural network, < >>
Figure SMS_13
Representing the number of bits of the information value of the visible layer, +.>
Figure SMS_18
Bias value representing hidden layer information, +.>
Figure SMS_15
Representing the%>
Figure SMS_16
Bias value of bit information value +.>
Figure SMS_11
Representing the%>
Figure SMS_21
Bit information value->
Figure SMS_7
Bias value representing visible layer information, +.>
Figure SMS_8
Representing the +.>
Figure SMS_12
Bias value of bit information value +.>
Figure SMS_20
Representing the +.>
Figure SMS_14
Bit information value->
Figure SMS_22
Information connection matrix representing hidden layer and visible layer, < >>
Figure SMS_9
Representing the%>
Figure SMS_17
Bit information value and +.>
Figure SMS_24
A connection matrix of bit information values;
s42: calculating probability distribution of information values of an implicit layer and a visible layer according to an energy function, wherein the probability distribution is shown as a formula (2):
Figure SMS_25
(2)/>
in the formula (2), the amino acid sequence of the compound,
Figure SMS_27
representing the partitioning factor,/->
Figure SMS_32
Representing the arithmetic constant->
Figure SMS_33
Representing an information energy function representing information values in the hidden layer +.>
Figure SMS_28
And visible layer->
Figure SMS_29
Information energy function, +.>
Figure SMS_30
Representing probability distribution->
Figure SMS_34
Representing hidden layer->
Figure SMS_26
And visible layer->
Figure SMS_31
Probability distribution of information values; fitting the distribution condition of the received network return information in the neural network through the information value distribution probability conditions of the hidden layer and the visible layer;
s43: calculating the relative entropy of information values of an implicit layer and an input layer
Defining hidden layer information value probability distribution as
Figure SMS_35
The probability distribution of the information values of the visible layer is +.>
Figure SMS_36
Figure SMS_37
And->
Figure SMS_38
The relative entropy calculation formula of (2) is shown in formula (3):
Figure SMS_39
(3)
in the formula (3), the amino acid sequence of the compound,
Figure SMS_40
representing sample space, ++>
Figure SMS_41
Indicating relative distance>
Figure SMS_42
Representing hidden layer->
Figure SMS_43
And visible layer->
Figure SMS_44
Distance of information value distribution->
Figure SMS_45
Representing a spatial relationship, wherein when the relative distance is minimum, the implicit layer information value is closest to the visible layer information value distribution;
s44: reconstructing an information error function
The received information enters an hidden layer for processing, a certain amount of hidden layer information values are selected as training samples, and the calculation formula of the logarithmic loss function of the hidden layer information values is shown as formula (4):
Figure SMS_46
(4)
in the formula (4), the amino acid sequence of the compound,
Figure SMS_48
representing the weight connection matrix from visible layer to hidden layer, < ->
Figure SMS_53
Representing loss factor,/->
Figure SMS_55
Representing the%>
Figure SMS_49
Bit information value +.>
Figure SMS_51
Probability distribution (S)/(S)>
Figure SMS_54
Representing a logarithmic loss function, +.>
Figure SMS_56
Representing implicit layer offset value +.>
Figure SMS_47
Visible layer bias value->
Figure SMS_50
Weight connection matrix from visible layer to hidden layer>
Figure SMS_52
A logarithmic loss function of (2);
and continuously optimizing the received information value by a small batch characteristic gradient descent method of the information momentum, and minimizing a reconstruction error according to a logarithmic loss function until the information value error reaches an ideal minimum value.
As a further technical scheme of the invention, the information classification is realized by improving the frog-leaping algorithm according to the neural network algorithm, and the realization steps of the improved frog-leaping algorithm are as follows:
randomly selecting optimized information value composition
Figure SMS_59
Individual information frog individuals, the information frog population is
Figure SMS_63
Defining the%>
Figure SMS_65
Only information frog->
Figure SMS_58
Wherein->
Figure SMS_60
Representing information frog>
Figure SMS_64
Characteristic elements representing information frog, +.>
Figure SMS_66
Representing the number of characteristic elements +.>
Figure SMS_57
Representing>
Figure SMS_61
Information about frog->
Figure SMS_62
Characteristic elements for information frog in populationThe step length of the frog exercise with the row information is updated, and the updated calculation formula is shown as the formula (5):
Figure SMS_67
(5)
in the formula (5), the amino acid sequence of the compound,
Figure SMS_69
information frog indicating best fitness in the population, < ->
Figure SMS_73
Figure SMS_76
Representing +.>
Figure SMS_70
Figure SMS_71
Information frog, ->
Figure SMS_72
Representing random real numbers, < >>
Figure SMS_75
Information frog indicating worst fitness in population, < ->
Figure SMS_68
Representing information frog exercise factor->
Figure SMS_74
Step length update of the information frog is represented;
calculating updated information frog according to the updated information frog step length, as shown in a formula (6):
Figure SMS_77
(6)/>
in the formula (6), the amino acid sequence of the compound,
Figure SMS_78
representing the optimized information frog, ++>
Figure SMS_79
Figure SMS_80
Information frog indicating worst fitness
Figure SMS_81
A lower limit and an upper limit of the step size allowed to be changed;
through chaotic mapping, the frog-leaping mixing optimization efficiency of the information frog is improved, and the mapping calculation is shown as a formula (7):
Figure SMS_82
(7)
in the formula (7), the amino acid sequence of the compound,
Figure SMS_83
representing the mapping coefficient->
Figure SMS_84
Representing the sequence->
Figure SMS_85
The mapped sequence of the information frog is represented, the mapped relational expression and the updated information frog form a new frog sequence, the information frog with the largest sequence number is selected from the frog sequence to be used as a lead frog, the characteristic elements contained in the lead frog are main characteristic elements of the frog group, an information characteristic classification model is established, and the characteristic elements are used as classification standards to realize information value classification.
As a further technical scheme of the invention, the data transmission relay station is used for temporarily storing data received by an external network, and comprises a data collector, a registering unit and a characteristic storage group; the data acquisition device adopts a low-power consumption singlechip to acquire data to be optimized, the registering unit adopts a shift register to temporarily store the optimized data, and the characteristic storage group adopts a capacitance storage to realize classified storage of the classified data.
As a further technical solution of the present invention, a network isolation method includes: the 001-number network channel is adopted to realize the connection between the management platform and the internal network, and the 002-number network channel is adopted to realize the connection between the management platform and the external network; the 001 and 002 network channels adopt parallel circuits to control the on or off of the network channels; the management platform opens the network channel based on the control current and does not communicate with the network channel to realize network isolation.
As a further technical scheme of the invention, the invention comprises the following steps: the depth statistics module is used for realizing sum calculation of the safety values and comprises a data normalization processing module, a DEEP model and an auxiliary module.
The beneficial effects of the invention are as follows:
different from the conventional technology, the invention provides a dual-network isolation method of a file protection cloud perception management platform, a file identification system is used for identifying file information, a network control system controls network connection, internal and external network connection is realized through an interface connection program, an external network management system realizes connection of a management platform and the Internet through a second transmitting antenna and a second receiving antenna, an internal network management system realizes connection of the management platform and an internal network through a first transmitting antenna and a first receiving antenna, an information processing system realizes network information optimization classification through a neural network-frog-leaping algorithm, and a network security monitoring system monitors network security states in real time in the network connection process; compared with other double-network isolation technologies, the double-network isolation method of the archive protection cloud perception management platform can better realize effective isolation of the internal network and the external network, efficiently process network data, optimize information and greatly improve network safety.
Drawings
For a clearer description of embodiments of the invention or of solutions in the prior art, the drawings that are necessary for the description of the embodiments or of the prior art will be briefly described, it being apparent that the drawings in the description below are only some embodiments of the invention, from which, without inventive faculty, other drawings can be obtained for a person skilled in the art, in which:
FIG. 1 shows a step diagram of a dual-network isolation method of an archive protection cloud sensing management platform;
FIG. 2 shows a dual-network isolation system diagram of an archive protection cloud sensing management platform of the present invention;
FIG. 3 illustrates a signal optimization classification flow chart of the present invention;
FIG. 4 shows the results of the algorithm classification experiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, it being understood that the embodiments described herein are for illustration and explanation of the present invention only, and are not intended to limit the present invention;
in a specific embodiment, a dual-network isolation method and implementation process of an archive protection cloud sensing management platform are shown in fig. 1 to fig. 4.
A dual-network isolation method of a file protection cloud perception management platform is characterized by comprising the following steps of: comprising the following steps:
s1: the palm computer (Personal Digital Assistant, PDA) classifier has certain operating system, can scan the file pocket in its 50cm scope, there is software development (Software Development Kit, SDK) system in the interior, have man-machine interface, can realize display and input function, scan the file pocket two-dimensional code label, discern the basic information in the file pocket, the file protects the cloud and perceives the management platform to synchronize the basic information in the database automatically through the internal GPRS network communication unit; wherein the database is integrated in a unit computer hard disk.
S2: the network control system controls the network connection of the archive protection cloud perception management platform by implementing a closed loop circuit set for control and communication through a network, an operator inputs an internal network connection command at a network connection interface, and the management platform realizes internal work network connection through a special data circuit; an operator inputs an external network connection command at a network connection interface, and the management platform is connected with an external network through an optical fiber internet module to realize network searching;
s3: the management platform inquires file information, the network security management platform verifies that a secret key inquiry command needs to be input for identity, the signal generator generates the frequency of an inquiry signal when the command is correct, the first transmitting antenna transmits the frequency signal, the first receiving antenna receives the frequency signal and converts the frequency signal into an electric signal through the signal conversion circuit, and the signal processing module calibrates and processes the inquiry command of the electric signal and dynamically invokes file information in the database according to the inquiry command; the signal processing module adopts DSP and FPGA parallel integration, selects ADSP series multiple DSP coupling digital signal processing module and Xilinx company Spartan3AN series chip, wherein Spartan3AN series chip is as core, multiple DSP coupling digital signal processing module auxiliary integration around the chip, constitutes signal processing module through net gape and circuit.
S4: the management platform queries information on the internet, the main server establishes wireless connection with external subnet units through a second transmitting antenna, communication channels are established among the subnet units with the same communication parameters, the second receiving antenna receives returned information of each subnet unit, the information processing module classifies and processes the returned information, and the subnet units correspond to datagram mechanisms of the packet switched network. The network connection first establishes two new datagrams with a maximum capacity of 64 bits, the first datagram length is set to insert data, the more-flag is set to 1, the offset is unchanged, the second datagram length is set to insert data, the more-flag is set to 0, and the offset is set to one eighth of the first datagram length. The information processing module adopts a neural network-frog-leaping algorithm to realize the classification processing of the information, and the neural network-frog-leaping algorithm comprises an information optimization model, an improved frog-leaping algorithm model, a frog-leaping classifier and a data transmission relay station.
S5: the network safety monitoring system carries out safety supervision on the working states of the internal network and the external network of the management platform, the network speed measuring unit calculates network speed according to response time, the network data counting unit calculates data information in network nodes and divides the network nodes into stable type, fluctuation type and mutation type according to the characteristics of the data information, the safety management unit carries out safety value statistics on the fluctuation type and mutation type network nodes based on a neural network deep learning algorithm, the safety value is lower than a minimum threshold value, and the network hazard warning unit starts an alarm mode. The neural network deep learning algorithm comprises a BP training model, a function set and a deep statistics module.
In a specific embodiment, the two-dimensional code of the file bag includes a file number, a generation time, an address, a file attribute, author information and a history record of the two-dimensional code.
In a specific embodiment, the archive database storage adopts a distributed ring topology structure, and the archives are stored in the distributed ring topology structure in the form of stacks with the same attribute and the same group.
In a specific embodiment, the archive protection cloud perception management platform queries information through an external network, information received by the second receiving antenna is classified through a neural network-frog-leaping algorithm, and the method for optimizing information data by the neural network algorithm comprises the following steps:
s41: calculating information energy function
Defining hidden layer information values in a neural network algorithm as
Figure SMS_86
Wherein->
Figure SMS_87
Representing the amount of hidden layer information->
Figure SMS_88
Representing the information value as a vector, the visible layer information value as +.>
Figure SMS_89
Wherein->
Figure SMS_90
The information energy function in the neural network is shown as a formula (1) and represents the information quantity of the visible layers:
Figure SMS_91
(1)
in the formula (1), the components are as follows,
Figure SMS_105
representing the number of bits of the hidden layer information value->
Figure SMS_95
Representing implicit coefficients in the neural network, < >>
Figure SMS_109
Representing display coefficients in a neural network, < >>
Figure SMS_93
Representing the number of bits of the information value of the visible layer, +.>
Figure SMS_100
Bias value representing hidden layer information, +.>
Figure SMS_98
Representing the%>
Figure SMS_106
Bias value of bit information value +.>
Figure SMS_97
Representing the%>
Figure SMS_101
Bit information value->
Figure SMS_92
Bias value representing visible layer information, +.>
Figure SMS_104
Representing the +.>
Figure SMS_94
Bias value of bit information value +.>
Figure SMS_102
Representing the +.>
Figure SMS_99
Bit information value->
Figure SMS_103
Information connection matrix representing hidden layer and visible layer, < >>
Figure SMS_96
Representing the%>
Figure SMS_108
Bit information value and +.>
Figure SMS_107
A connection matrix of bit information values;
among the parameters mentioned above, the above-mentioned parameters,
Figure SMS_110
implicit layer->
Figure SMS_111
Bias value of bit information value for adjusting given value of hidden layer,
Figure SMS_112
Visible layer->
Figure SMS_113
The offset value of the bit information value is used to adjust the visible layer set point,/->
Figure SMS_114
The information connection matrix is used for adjusting the sensing accuracy in the system, < + >>
Figure SMS_115
The larger the implicit coefficient is, the higher the information content is, the higher the possibility that the information value is utilized in the optimization process is, and the received information value can be standardized and unified through the parameter calculation, so that the following processing is facilitated.
S42: calculating probability distribution of information values of an implicit layer and a visible layer according to an energy function, wherein the probability distribution is shown as a formula (2):
Figure SMS_116
(2)
in the formula (2), the amino acid sequence of the compound,
Figure SMS_118
representing the partitioning factor,/->
Figure SMS_121
Representing the arithmetic constant->
Figure SMS_123
Representing an information energy function representing information values in the hidden layer +.>
Figure SMS_119
And visible layer->
Figure SMS_122
Information energy function, +.>
Figure SMS_124
Representing probability distribution->
Figure SMS_125
Representing hidden layer->
Figure SMS_117
And visible layer->
Figure SMS_120
Probability distribution of information values; fitting the distribution condition of the received network return information in the neural network through the information value distribution probability conditions of the hidden layer and the visible layer;
among the parameters mentioned above, the above-mentioned parameters,
Figure SMS_126
the distribution factors carry out statistical calculation on information values of different layers, so that the operation efficiency is improved, and the user is added with the information values of the different layers>
Figure SMS_127
The value range of the operation constant is [0,1 ]]And through the parameter calculation, the information value is subjected to unsupervised model training, so that the following processing is facilitated.
S43: calculating the relative entropy of information values of an implicit layer and an input layer
Defining hidden layer information value probability distribution as
Figure SMS_128
The probability distribution of the information values of the visible layer is +.>
Figure SMS_129
Figure SMS_130
And->
Figure SMS_131
The relative entropy calculation formula of (2) is shown in formula (3):
Figure SMS_132
(3)
in the formula (3), the amino acid sequence of the compound,
Figure SMS_133
representing sample space, ++>
Figure SMS_134
Indicating relative distance>
Figure SMS_135
Representing hidden layer->
Figure SMS_136
And visible layer->
Figure SMS_137
Distance of information value distribution->
Figure SMS_138
Representing a spatial relationship, wherein the relative distance is minimum, and the implicit layer information value is closest to the visible layer information value distribution;
among the parameters mentioned above, the above-mentioned parameters,
Figure SMS_139
the sample space comprises an input layer, an implicit layer, a visible layer and an output layer;
Figure SMS_140
The relative distance is expressed in the sample space, and the probability distribution of the information value of the hidden layer is +.>
Figure SMS_141
Probability distribution with visible layer information value is +.>
Figure SMS_142
And the distribution distance is calculated through the parameters, and the approximate value of the information value is solved, so that the following processing is facilitated.
S44: reconstructing an information error function
The received information enters an hidden layer for processing, a certain amount of hidden layer information values are selected as training samples, and the calculation formula of the logarithmic loss function of the hidden layer information values is shown as formula (4):
Figure SMS_143
(4)
in the formula (4), the amino acid sequence of the compound,
Figure SMS_146
representing the weight connection matrix from visible layer to hidden layer, < ->
Figure SMS_148
Representing loss factor,/->
Figure SMS_152
Representing the%>
Figure SMS_145
Bit information value +.>
Figure SMS_147
Probability distribution (S)/(S)>
Figure SMS_150
Representing a logarithmic loss function, +.>
Figure SMS_153
Representing implicit layer offset value +.>
Figure SMS_144
Visible layer bias value->
Figure SMS_149
Weight connection matrix from visible layer to hidden layer>
Figure SMS_151
A logarithmic loss function of (2);
in the above parameters, the visible layer-to-hidden layer weight connection matrix
Figure SMS_154
Representation->
Figure SMS_155
And (3) connecting the matrixes of the connection matrixes, continuously optimizing the received information value by a small batch characteristic gradient descent method of the information momentum, and minimizing a reconstruction error according to a logarithmic loss function until the information value error reaches an ideal minimum value.
In a specific embodiment, the information classification is realized by improving the frog-leaping algorithm according to the neural network algorithm, and the implementation steps of the improved frog-leaping algorithm are as follows:
randomly selecting optimized information value composition
Figure SMS_157
Individual information frog individuals, the information frog population is
Figure SMS_161
Defining the%>
Figure SMS_164
Only information frog->
Figure SMS_158
Wherein->
Figure SMS_160
Representing information frog>
Figure SMS_163
Characteristic elements representing information frog, +.>
Figure SMS_165
Representing the number of characteristic elements +.>
Figure SMS_156
Representing>
Figure SMS_159
Information about frog->
Figure SMS_162
And (3) carrying out information frog movement step length updating on the information frog in the population by the characteristic elements, wherein an updating calculation formula is shown in a formula (5):
Figure SMS_166
(5)
in the formula (5), the amino acid sequence of the compound,
Figure SMS_168
information frog indicating best fitness in the population, < ->
Figure SMS_173
Figure SMS_174
Representing +.>
Figure SMS_169
Figure SMS_170
Information frog, ->
Figure SMS_172
Representing random real numbers, < >>
Figure SMS_175
Information frog indicating worst fitness in population, < ->
Figure SMS_167
Representing information frog exercise factor->
Figure SMS_171
Step length update of the information frog is represented;
among the parameters mentioned above, the above-mentioned parameters,
Figure SMS_176
the random real numbers are generated by uniformly distributing (0, 1), when the fitness of the information frog is fuzzy, namely +.>
Figure SMS_177
Information frog and +.>
Figure SMS_178
And when the information frog with the worst adaptability in the population approaches, reselecting the frog population. Through the parameter calculation, the input information frog can be screened, and the following processing is facilitated.
Calculating updated information frog according to the updated information frog step length, as shown in a formula (6):
Figure SMS_179
(6)
in the formula (6), the amino acid sequence of the compound,
Figure SMS_180
representing the optimized information frog, ++>
Figure SMS_181
Figure SMS_182
Information frog indicating worst fitness
Figure SMS_183
A lower limit and an upper limit of the step size allowed to be changed; />
Among the parameters mentioned above, the above-mentioned parameters,
Figure SMS_184
the optimized information frog is better than the information frog with worst adaptability in the original population
Figure SMS_185
Replacing information frog in the population, otherwise, reselecting frog population, and re-optimizing information frog with worst adaptability in the population>
Figure SMS_186
Continuously optimizing updating information frog->
Figure SMS_187
Until the global convergence reaches the maximum number of times.
Through chaotic mapping, the frog-leaping mixing optimization efficiency of the information frog is improved, and the mapping calculation is shown as a formula (7):
Figure SMS_188
(7)
in the formula (7), the amino acid sequence of the compound,
Figure SMS_189
representing the mapping coefficient->
Figure SMS_190
Representing the sequence->
Figure SMS_191
The mapped sequence of the information frog is represented, the mapped relational expression and the updated information frog form a new frog sequence, the information frog with the largest sequence number is selected from the frog sequence to be used as a lead frog, the characteristic elements contained in the lead frog are main characteristic elements of the frog group, an information characteristic classification model is established, and the characteristic elements are used as classification standards to realize information value classification.
In a specific embodiment, the data transmission relay station is used for temporarily storing data received by the external network, and comprises a data collector, a register unit and a feature storage group; the data acquisition device adopts a low-power consumption singlechip to acquire data to be optimized, the registering unit adopts a shift register to temporarily store the optimized data, and the characteristic storage group adopts a capacitance storage to realize classified storage of the classified data.
In a specific embodiment, the network isolation method includes: the 001-number network channel is adopted to realize the connection between the management platform and the internal network, and the 002-number network channel is adopted to realize the connection between the management platform and the external network; the 001 and 002 network channels adopt parallel circuits to control the on or off of the network channels; the management platform opens the network channel based on the control current and does not communicate with the network channel to realize network isolation.
In a specific embodiment, the depth statistics module is used for realizing sum calculation of the security values, and the depth statistics module comprises a data normalization processing module, a DEEP model and an auxiliary module.
In a specific embodiment, a dual-network isolation system of an archive protection cloud sensing management platform is characterized in that: comprising the following steps:
the display management platform realizes matrix addressing control of the liquid crystal light valve through a TET-LCD liquid crystal display screen, a backlight LCD and a large-scale integrated circuit and is used for displaying file information, instructions, network connection information and abnormal information, and the display management platform displays information based on an ARM image processor;
the file identification system is used for reading the basic information of files and comprises a palm computer classifier and a GPRS network communication unit, wherein the palm computer classifier is connected with the GPRS network communication unit;
the network safety monitoring system is used for detecting the safety of a network in the connection and use process and comprises a network speed measuring unit, a network data statistics unit, a network hazard warning unit and a safety management unit, wherein the network speed measuring unit is connected with the network data statistics unit, the network data statistics unit is connected with the network hazard warning unit, and the network hazard warning unit is connected with the safety management unit;
the network speed measuring unit is realized through Eclipse environment codes, and the display management platform displays an interface after network speed measurement. The network data statistics is based on a resting brain function network model to process and count the data in the network, and the processing statistics comprise data format conversion, time point removal, time layer correction, automatic correction space standardization, linear drift removal, covariate removal and the like.
The double-network control system is used for controlling the archive protection cloud perception management platform to be connected with a network and comprises a network selection unit and a connection unit; the network selection unit introduces a software-defined network controller, and the controller controls reporting of network parameter information and analysis of the received network information, and an operator selects or the controller selects a network preferentially.
The signal processing system adopts digital frequency conversion to realize mode conversion, data compression, extraction, transformation operation and filtering matching of received information, and adopts a parallel capturing method to search the phase and frequency of a signal code in the time domain, wherein the signal mixer extracts the center frequency of the signal, and frequency offset in the signal is eliminated by frequency conversion processing, so that the searching and processing of the signal are realized.
The external network management system is used for realizing the connection between the management platform and the Internet and comprises a second transmitting antenna, an external network and a second receiving antenna; the external network connection adopts an MPLS VPN mode network deployment model to bear network connection.
The intranet management system is used for realizing the connection between a management unit and a unit network to which the management unit belongs, and comprises a first transmitting antenna, an external network and a first receiving antenna; and an intranet GRE tunnel is established between the intranet line and each unit exit router, so that the intranet GRE tunnel is suitable for various protocols, and a connection secret key is established to protect the privacy of the intranet.
The display management platform is respectively connected with the signal processing system, the file identification system, the network security monitoring system and the double-network control system, and the signal processing system is respectively connected with the external network management system and the internal network management system.
In a specific embodiment, the signal processing system adopts a ZU9EG series FPGA chip, and a processor of the chip is integrated with the ASIC in a low power consumption manner, so that dual channel processing of internal signal data and external signals is realized.
In a specific embodiment, the network connection unit adopts a JDBC database to realize network connection, and the JDBC database realizes connection of internal and external networks by calling different interface connection programs.
In a specific embodiment, the analysis and verification of the data optimization classification effect are carried out on the neural network-frog-leaping algorithm, and the specific experiment is as follows:
in order to verify the optimized classification effect of the neural network-frog-leaping algorithm on the received network information, a whale-optimized convolution algorithm and a dragonfly-improved data mining algorithm are used for carrying out a comparison experiment with the algorithm in the invention, the neural network-frog-leaping algorithm is set to be in an English mode, NELA is set to be in an English mode, WOCA is set to be in the whale-optimized convolution algorithm, and the dragonfly-improved data mining algorithm is set to be in an English mode and DDMA is set to be in the dragonfly-improved data mining algorithm.
Experimental environment: the Simulink algorithm modeling platform, the signal generator and the data display screen.
Initial parameters of three algorithm models are set, the training frequency is 50, the verification frequency is 10, the iteration frequency is 200, the signal generator sends out signals with 5 periods, and the signal values processed by the algorithms are shown in table 1.
Table 1 algorithm optimization rate
Figure SMS_192
As can be seen from table 1, the optimizing rate of NELA on the signal is the best, and next, the optimizing rate of DDMA and WOCA on the signal is the worst, and the classification results of the three algorithms on the signal are shown in fig. 4.
As can be seen from fig. 4, as the number of iterations increases, the information feature set of each algorithm increases gradually, where the analysis speed of the information feature set of NELA is the fastest, the analyzed information feature set is higher than the analysis results of other algorithms, as the number of iterations increases, the number of information feature sets also increases continuously, and finally the increasing speed tends to be gentle; wherein NELA has an information profiling capability greater than DDMA, which has an information profiling capability greater than WOCA.
Experiments show that the neural network-frog-leaping algorithm adopted by the invention has higher optimizing capability and classifying capability on signals, and the neural network-frog-leaping algorithm adopted by the invention can be used for more efficiently and accurately optimizing and classifying the signals.
While specific embodiments of the present invention have been described above, it will be understood by those skilled in the art that the foregoing detailed description is given by way of example only, and that various omissions, substitutions and changes in the form of the details of the method and system illustrated may be made by those skilled in the art without departing from the spirit and scope of the invention; for example, it is within the scope of the present invention to combine the above-described method steps to perform substantially the same function in substantially the same way to achieve substantially the same result; accordingly, the scope of the invention is limited only by the following claims.

Claims (5)

1. A dual-network isolation method of a file protection cloud perception management platform is characterized by comprising the following steps of: comprising the following steps:
s1: a PDA classifier is adopted to scan two-dimension code labels of the file bags, basic information in the file bags is identified, and the file protection cloud perception management platform automatically synchronizes the basic information in a database through a GPRS network communication unit arranged inside;
s2: the network control system controls the network connection of the archive protection cloud perception management platform, an operator inputs an internal network connection command at a network connection interface, and the management platform realizes internal work network connection through a special data circuit; an operator inputs an external network connection command at a network connection interface, and the management platform is connected with an external network through an optical fiber internet module to realize network information transmission search;
s3: the management platform inquires file information, inputs a secret key inquiry command, a signal generator generates the frequency of an inquiry signal, a first transmitting antenna transmits the frequency signal, a first receiving antenna receives the frequency signal, the frequency signal is converted into an electric signal through a signal conversion circuit, and a signal processing module calibrates and processes the inquiry command of the electric signal and dynamically invokes file information in a database according to the inquiry command;
s4: the management platform is connected with the internet to inquire information, the main server establishes wireless connection with external subnet units through a second transmitting antenna, communication channels are established among the subnet units with the same communication parameters, the second receiving antenna receives returned information of each subnet unit, and the information processing module classifies and processes the returned information; the information processing module adopts a neural network-frog-leaping algorithm to realize the classification processing of the information, and the neural network-frog-leaping algorithm comprises an information optimization model, an improved frog-leaping algorithm model, a frog-leaping classifier and a data transmission relay station;
s5: the network safety monitoring system carries out safety supervision on the working states of the internal network and the external network of the management platform, the network speed measuring unit calculates the network speed according to the response time, the network data statistics unit calculates the data information in the network nodes and divides the network nodes into stable type, fluctuation type and mutation type according to the characteristics of the data information, the safety management unit carries out safety value statistics on the fluctuation type and mutation type network nodes based on a neural network deep learning algorithm, the safety value is lower than a lowest threshold value, and the network hazard warning unit starts an alarm mode; the neural network deep learning algorithm comprises a BP training model, a function set and a deep statistics module;
the archive protection cloud perception management platform queries information through an external network, information received by the second receiving antenna is optimized and classified through a neural network-frog-leaping algorithm, and the method for optimizing information data through the neural network algorithm comprises the following steps:
s41: calculating an information energy function;
defining hidden layer information values in a neural network algorithm as
Figure QLYQS_1
Wherein->
Figure QLYQS_2
Representing the amount of hidden layer information->
Figure QLYQS_3
Representing the information value as a vector, the visible layer information value as +.>
Figure QLYQS_4
Wherein->
Figure QLYQS_5
The information energy function in the neural network is shown as a formula (1) and represents the information quantity of the visible layers:
Figure QLYQS_6
(1)
in the formula (1), the components are as follows,
Figure QLYQS_14
representing the number of bits of the hidden layer information value->
Figure QLYQS_9
Representing implicit coefficients in the neural network, < >>
Figure QLYQS_11
Representing display coefficients in a neural network, < >>
Figure QLYQS_10
Representing the number of bits of the information value of the visible layer, +.>
Figure QLYQS_15
Bias value representing hidden layer information, +.>
Figure QLYQS_19
Representing the%>
Figure QLYQS_23
Bias value of bit information value +.>
Figure QLYQS_16
Representing the%>
Figure QLYQS_21
Bit information value->
Figure QLYQS_7
Bias value representing visible layer information, +.>
Figure QLYQS_12
Representing the +.>
Figure QLYQS_18
Bias value of bit information value +.>
Figure QLYQS_22
Representing the +.>
Figure QLYQS_20
Bit information value->
Figure QLYQS_24
An information connection matrix representing the hidden layer and the visible layer,
Figure QLYQS_8
representing the%>
Figure QLYQS_13
Bit information value and +.>
Figure QLYQS_17
A connection matrix of bit information values;
s42: calculating probability distribution of information values of an implicit layer and a visible layer according to an energy function, wherein the probability distribution is shown as a formula (2):
Figure QLYQS_25
(2)
in the formula (2), the amino acid sequence of the compound,
Figure QLYQS_28
representing the partitioning factor,/->
Figure QLYQS_29
Representing the arithmetic constant->
Figure QLYQS_31
Representing information energy functions, representing information values at hidden layers
Figure QLYQS_26
And visible layer->
Figure QLYQS_32
Information energy function, +.>
Figure QLYQS_33
Representing probability distribution->
Figure QLYQS_34
Representing hidden layer->
Figure QLYQS_27
And visible layer->
Figure QLYQS_30
Probability distribution of information values; fitting the distribution condition of the received network return information in the neural network through the information value distribution probability conditions of the hidden layer and the visible layer;
s43: calculating the relative entropy of the information values of the hidden layer and the input layer;
defining hidden layer information value probability distribution as
Figure QLYQS_35
The probability distribution of the information values of the visible layer is +.>
Figure QLYQS_36
Figure QLYQS_37
And->
Figure QLYQS_38
The relative entropy calculation formula of (2) is shown in formula (3):
Figure QLYQS_39
(3)
in the formula (3), the amino acid sequence of the compound,
Figure QLYQS_40
representing sample space, ++>
Figure QLYQS_41
Indicating relative distance>
Figure QLYQS_42
Representing hidden layer->
Figure QLYQS_43
And visible layer->
Figure QLYQS_44
Distance of information value distribution->
Figure QLYQS_45
Representing spatial relationships; the relative distance is minimum, and the implicit layer information value is closest to the visible layer information value distribution;
s44: reconstructing an information error function;
the received information enters an hidden layer for processing, a certain amount of hidden layer information values are selected as training samples, and the calculation formula of the logarithmic loss function of the hidden layer information values is shown as formula (4):
Figure QLYQS_46
(4)
in the formula (4), the amino acid sequence of the compound,
Figure QLYQS_48
representing the weight connection matrix from visible layer to hidden layer, < ->
Figure QLYQS_52
Representing loss factor,/->
Figure QLYQS_55
Representing the%>
Figure QLYQS_49
Bit information value +.>
Figure QLYQS_50
Probability distribution (S)/(S)>
Figure QLYQS_53
Representing a logarithmic loss function, +.>
Figure QLYQS_56
Representing implicit layer offset value +.>
Figure QLYQS_47
Visible layer bias value->
Figure QLYQS_51
Weight connection matrix from visible layer to hidden layer>
Figure QLYQS_54
A logarithmic loss function of (2);
continuously optimizing the received information value by a small batch characteristic gradient descent method of the information momentum, and minimizing a reconstruction error according to a logarithmic loss function until the information value error reaches an ideal minimum value;
the neural network algorithm realizes information classification by improving the frog-leaping algorithm, and the method comprises the following steps of:
randomly selecting optimized information value composition
Figure QLYQS_58
Individual information frog individuals, the information frog population is
Figure QLYQS_60
Defining the%>
Figure QLYQS_65
Only information frog->
Figure QLYQS_59
Wherein->
Figure QLYQS_61
Representing information frog>
Figure QLYQS_63
Characteristic elements representing information frog, +.>
Figure QLYQS_66
Representing the number of characteristic elements +.>
Figure QLYQS_57
Representing>
Figure QLYQS_62
Information about frog->
Figure QLYQS_64
And (3) carrying out information frog movement step length updating on the information frog in the population by the characteristic elements, wherein an updating calculation formula is shown in a formula (5):
Figure QLYQS_67
(5)
in the formula (5), the amino acid sequence of the compound,
Figure QLYQS_69
information frog indicating best fitness in the population, < ->
Figure QLYQS_72
Figure QLYQS_74
Representing +.>
Figure QLYQS_70
Figure QLYQS_73
Information frog, ->
Figure QLYQS_75
Representing random real numbers, < >>
Figure QLYQS_76
Information frog indicating worst fitness in population, < ->
Figure QLYQS_68
Representing information frog exercise factor->
Figure QLYQS_71
Step length update of the information frog is represented;
calculating updated information frog according to the updated information frog step length, as shown in a formula (6):
Figure QLYQS_77
(6)/>
in the formula (6), the amino acid sequence of the compound,
Figure QLYQS_78
representing the optimized information frog, ++>
Figure QLYQS_79
Figure QLYQS_80
Information frog representing worst fitness +.>
Figure QLYQS_81
A lower limit and an upper limit of the step size allowed to be changed;
through chaotic mapping, the frog-leaping mixing optimization efficiency of the information frog is improved, and the mapping calculation is shown as a formula (7):
Figure QLYQS_82
(7)
in the formula (7), the amino acid sequence of the compound,
Figure QLYQS_83
representing the mapping coefficient->
Figure QLYQS_84
Representing the sequence->
Figure QLYQS_85
Representing the mapped sequence of the information frog, forming a new frog sequence by the mapped relational expression and the updated information frog, selecting the information frog with the largest sequence number from the frog sequence as a lead frog, wherein characteristic elements contained in the lead frog are main characteristic elements of frog groups, establishing an information characteristic classification model, and realizing information value classification by taking the characteristic elements as classification standards;
the network isolation method comprises the following steps: the 001-number network channel is adopted to realize the connection between the management platform and the internal network, and the 002-number network channel is adopted to realize the connection between the management platform and the external network; the 001 and 002 network channels adopt parallel circuits to control the on or off of the network channels; the management platform opens the 001 # network channel and the 002 # network channel based on the control current and does not communicate with each other to realize network isolation.
2. The archive protection cloud perception management platform dual-network isolation method according to claim 1, wherein the archive protection cloud perception management platform dual-network isolation method is characterized by comprising the following steps: the file bag two-dimension code comprises a file number, a generation time, an address, a file attribute, author information of the file and a two-dimension code history scanning record of the file.
3. The archive protection cloud perception management platform dual-network isolation method according to claim 1, wherein the archive protection cloud perception management platform dual-network isolation method is characterized by comprising the following steps: the archives are stored in the distributed ring topology structure in the form of stacks with the same attribute and the same group.
4. The archive protection cloud perception management platform dual-network isolation method according to claim 1, wherein the archive protection cloud perception management platform dual-network isolation method is characterized by comprising the following steps: the data transmission relay station is used for temporarily storing data received by the external network, and comprises a data collector, a registering unit and a characteristic storage group; the data acquisition device adopts a low-power consumption singlechip to acquire data to be optimized, the registering unit adopts a shift register to temporarily store the optimized data, and the characteristic storage group adopts a capacitance storage to realize classified storage of the classified data.
5. The archive protection cloud perception management platform dual-network isolation method according to claim 1, wherein the archive protection cloud perception management platform dual-network isolation method is characterized by comprising the following steps: the depth statistics module is used for realizing sum calculation of the safety values and comprises a data normalization processing module, a DEEP model and an auxiliary module.
CN202310043031.4A 2023-01-29 2023-01-29 File protection cloud perception management platform double-network isolation method Active CN115801451B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310043031.4A CN115801451B (en) 2023-01-29 2023-01-29 File protection cloud perception management platform double-network isolation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310043031.4A CN115801451B (en) 2023-01-29 2023-01-29 File protection cloud perception management platform double-network isolation method

Publications (2)

Publication Number Publication Date
CN115801451A CN115801451A (en) 2023-03-14
CN115801451B true CN115801451B (en) 2023-06-02

Family

ID=85428969

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310043031.4A Active CN115801451B (en) 2023-01-29 2023-01-29 File protection cloud perception management platform double-network isolation method

Country Status (1)

Country Link
CN (1) CN115801451B (en)

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9405786B2 (en) * 2013-10-22 2016-08-02 Nice-Systems Ltd. System and method for database flow management
CN109309680A (en) * 2018-10-09 2019-02-05 山西警察学院 Network security detection method and guard system based on neural network algorithm
CN111008216A (en) * 2019-12-23 2020-04-14 上海中信信息发展股份有限公司 Method for transferring file retrieval information across physical isolation network
CN113627471A (en) * 2021-07-03 2021-11-09 西安电子科技大学 Data classification method, system, equipment and information data processing terminal
CN113971367B (en) * 2021-08-27 2024-07-12 天津大学 Automatic convolutional neural network framework design method based on shuffled frog-leaping algorithm

Also Published As

Publication number Publication date
CN115801451A (en) 2023-03-14

Similar Documents

Publication Publication Date Title
CN105069703B (en) A kind of electrical network mass data management method
CN102163230B (en) Method for implementing personalized information retrieval system for customizing privacy protection
CN113329023A (en) Encrypted flow malice detection model establishing and detecting method and system
CN110929806A (en) Picture processing method and device based on artificial intelligence and electronic equipment
CN110990498A (en) Data fusion method based on FCM algorithm
CN108710644A (en) One kind is about government affairs big data processing method
CN115278737B (en) Data acquisition method of 5G network
CN111242276A (en) One-dimensional convolution neural network construction method for load current signal identification
CN115580450A (en) Method and device for detecting flow, electronic equipment and computer readable storage medium
WO2019062404A1 (en) Application program processing method and apparatus, storage medium, and electronic device
CN113676357B (en) Decision method for edge data processing in power internet of things and application thereof
CN113687610B (en) Method for protecting terminal information of GAN-CNN power monitoring system
CN115801451B (en) File protection cloud perception management platform double-network isolation method
US20230409422A1 (en) Systems and Methods for Anomaly Detection in Multi-Modal Data Streams
Qiu et al. Abnormal traffic detection method of internet of things based on deep learning in edge computing environment
CN115660538B (en) Cargo transportation method and system
CN116805039A (en) Feature screening method, device, computer equipment and data disturbance method
CN115379308B (en) Internet of things equipment data acquisition system based on satellite remote communication
CN116957585A (en) Data processing method, device, equipment and storage medium
CN114205238B (en) Network resource optimization and model training method and device, storage medium and equipment
CN113409096B (en) Target object identification method and device, computer equipment and storage medium
Cai et al. Cluster-based federated learning framework for intrusion detection
CN114971504A (en) Entity type determination method and related device
CN116189706A (en) Data transmission method, device, electronic equipment and computer readable storage medium
CN111241054A (en) Virtual database-based power communication network remote heterogeneous data source integration method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant