Evaluation of Internet of Things computer network security and remote control technology

2.1 Network security

In the International Organization for Standardization, the security of a computer system is defined as the technology and management used to protect the data processing system to prevent the destruction, modification, and disclosure of computer hardware, software, and data. Fundamentally, network security is an application technology, involving many technical aspects. Among them, network information security and confidentiality is to protect the network information system from threats and accidents. Technically, the confidentiality, integrity, authenticity, availability, and non-repudiation of network information are the important purposes of its realization. The key to network security lies in the transmission of data, information, networks, and the processing capacity of terminals [9]. At the same time, it can only serve the right users. At the same time, due to the defects of the computer network itself, as well as its own defects, the hardware, communication, software, information resources, etc., of the computer network are subject to predictable or accidental or malicious damage, which leads to the failure of the information system and the paralysis of the system and thus leads to serious economic losses. In this case, the purpose network security based on preventing various attacks in the network is particularly important.

Computer network has changed people’s behavior, access to and use of information, and changed people’s life style completely. There are three main forms of network security: external intrusion, internal penetration, and improper behavior [10]. External intrusion is an attack on unauthorized computer system users. The so-called internal penetration means that authorized computer system users access unauthorized data. Because of its own defects and openness, as well as the intrusion of hackers, the network is not safe.

Since the main resource of computer network is its services to users and the information it holds, the security of computer network is mainly manifested in ensuring the availability of network services and the integrity of information [11,12,13]. The former needs to selectively provide each user with its own network services, while the latter needs to ensure its confidentiality, integrity, availability, and accuracy. It can be seen that the basic problem of building a secure network system is to properly control the types and scope of network services on the premise of ensuring network connectivity and availability, so as to ensure network availability and information integrity [14,15].

A secure computer network should have the characteristics as shown in Figure 1.

Figure 1

Features of secure computer network.

It can be seen from Figure 1 that (1) the reliability of the network system is its most fundamental requirement. Reliability refers to the software and hardware performance of the network system without failure; (2) availability refers to the network information that can be accessed by authorized users, i.e., when necessary, it can ensure that authorized users can access the information of the network; (3) confidentiality means not disclosing information in the network. Confidentiality is the key technology to ensure the reliability and availability of network information systems. Confidentiality means that even in the case of information disclosure, unauthorized users cannot verify their identity within a certain time limit; (4) integrity refers to the network information that would not change without permission, i.e., during storage and transmission, there would be no deletion, modification, forgery, disorder, replay, insertion, etc.; (5) non-repudiation refers to a kind of information exchange on the Internet, which can ensure that no party can deny or deny the transaction, just like signing and receiving documents when publishing or receiving documents.

Technically, the main content of network security has four parts, as shown in Figure 2.

Figure 2

Network security content.

In Figure 2, the network security content mainly includes network entity security, software security, network data security, and network security management. From this point, it can be seen that the security of computer network should not only ensure the security of computer network equipment, but also ensure the security of data. It is characterized by a security protection measure taken to solve the security problems of the computer network itself, so as to ensure its own security.

The goal of network security strategy is to determine how a network organization protects itself and the information it provides [16]. The overall strategy is used to describe the general concept of the company’s security policy, while the specific provisions are used to explain which behaviors can be done and which cannot be done.

The network security policy can be divided into four levels: there is no connection between the internal network and the external network, so both are prohibited; unless explicitly permitted, everything is prohibited; except for those explicitly prohibited, others can be used; everything is allowed, including prohibited things.

Users can choose their own security strategy according to the actual situation in four levels. When the system’s own conditions change, it needs to be adjusted appropriately.

A good network security strategy should include the following points: the security obligations of network users, the security responsibilities of system administrators, the rational use of Internet resources, and the measures to be taken when network security problems are found [17,18].

Network security management of network security policy refers to technical management combined with administrative management, including physical security policy, access control policy, information encryption policy, and network security management policy.

2.2 Remote control system

The development trend of IoT industry is changing with each passing day, among which smart agriculture and smart home are the most prominent. However, the current control system of smart agriculture is mainly concentrated on the Local Area Network (LAN), and the development requirements of smart agriculture are increasingly high. IoT technology is gradually developing on the basis of various cloud computing technologies and shows a trend of blooming [19,20].

Before designing the architecture of the system, a detailed functional analysis must be carried out to lay a good foundation for future design and implementation.

IoT remote control mainly includes remote device, cloud platform, and user end. The remote network communication mode is adopted between the three modules. The user end connects with the cloud platform to complete the transmission of commands. At the same time, the cloud platform connects the remote device with the remote device and transmits commands.

The structure of the system is shown in Figure 3.

Figure 3

System structure.

Cloud platform: Cloud platform refers to the use of the current mainstream cloud computing platform to connect user terminals with smart agricultural equipment and to forward, store, and analyze information about smart agriculture. At present, cloud computing technology has been widely used in mobile internet, data mining, and other aspects, and the remote control requirements of smart agriculture complement its own characteristics and advantages, so cloud computing technology is also increasingly applied in smart agriculture.

User mobile terminal: User mobile terminal is a smart phone application. Users can operate smart agricultural equipment locally or remotely through smart mobile terminals to facilitate operation. Now, with the rapid development of the network, the application of smart phones is becoming more and more widespread, and smart phones have become an important part of smart agriculture.

Remote equipment: Remote equipment can be divided into two types: LAN gateway and local smart agricultural equipment. The reason why the two are put together is that some local intelligent agricultural equipment can be connected to the cloud platform without the help of the LAN gateway, thus realizing the function of intelligent agriculture. Due to the low computing capacity of some devices, they must be uniformly managed by the LAN gateway and connected to the cloud platform through the gateway interface.

The research content of this article is a complete IoT remote device control system. In order to complete the basic functions of the system, the objectives and requirements of the system were analyzed, as shown in Figure 4.

Figure 4

System functions.

User management: The problems faced by the remote control system are the limited network resources and the guarantee of the user’s use security. Therefore, a user management module must be set to manage the user’s permissions.

Remote device management: Remote device management is the basic functional requirement of the device. Before it can be controlled, it must be managed, including storage, display, query, name change, etc. After leaving the LAN environment to which the device is connected, the user can connect to the cloud platform through mobile network or other ways to remotely manage the device and obtain the status of the device.

Remote device control: Remote device control is a key problem to be solved in this article. It can realize remote control, light adjustment, and other remote operations of the device. Similar to the previous feature, this feature is also used for non-local LAN.

Remote notification alarm: The so-called remote notification alarm is that when a device changes or triggers a special situation, the system would send the alarm information to the user’s terminal, so that the user can get it in time.

Remote rule engine: Rule engine refers to the behavior completed by the device by setting rules composed of activities in smart agriculture. This function is implemented by a gateway with high computing power. When running remotely, it is necessary to use the cloud communication function to complete the acquisition, modification, deletion, and configuration of rules.

The IoT remote control system studied in this article can be divided into several modules, so that the function of each module can be clearly divided, which is convenient for implementation and testing.

The main functional modules of the system include network communication module, message instruction processing module, device control management module, cloud platform software module, user management module, and interface module.

2.3 Improvement of BPNN algorithm using GA

At present, many scholars have combined GA with neural network and used GA’s global optimization ability to achieve the optimal prediction of neural network. The combination of the two can be determined from two aspects: the first is auxiliary combination method – it uses GA to preprocess the data and then uses BPNN to solve the problem, such as pattern recognition with GA for feature extraction, and then uses NN for classification. The second is cooperation. Under the fixed network topology of BPNN, GA and NN deal with the problem together, use GA to determine the link weight, or directly use GA to optimize the network structure and reuse, and BPNN trains the network.

The search of GA does not depend on gradient information, nor does it need the differentiability of the solution. It only needs the solvability of the solution under constrained conditions. Moreover, the feature of this method is global search. Therefore, GA is used to optimize the connection weight and network structure of the neural network, which can solve the problem of BPNN well and can also effectively improve the generalization performance of the neural network.

GA is an optimization method of BPNN. It adjusts the network weight by changing the gradient information on which the BPNN algorithm depends. At the same time, it can make full use of the global search characteristics of GA to start with the optimal network structure and network connection weight. Since the hidden layer, output layer, and input layer form a neural network, and the number of nodes in the input layer and output layer is determined by the model samples, the number of hidden nodes should be considered in the optimization process of BPNN.

The mathematical description of the optimal solution of genetic algorithm–backpropagation neural network’s (GA–BPNN’s) problem is as follows:

BPNN is used to determine the basic solution space of the aforementioned equation. First, the training times and training error η 1 of the network are set, and then, the training samples and the error η 2 of the detection samples are input. When the errors η 1 and η 2 are satisfactory, the minimum and maximum values of the connection weight are recorded as i min and i max , and the interval [ i max − ε 1 , i max + ε 2 ] (as the adjustment parameter) is used as the basic solution space of the connection weight.

u , b , ϑ , r are the network inputs; both y l ( t ) , y ˆ l ( t ) are the network outputs. Since the GA regards the maximum value of the objective function as its fitness function during optimization, the fitness function is defined as

then formula (2) would become:

In this algorithm, the adaptability of each calculated individual is obtained by this code. The algorithm can obtain the number of hidden nodes of the network and obtain the connection weight of the network from the weight value. Then, the algorithm calculates the number of hidden nodes in the network.

The most adaptive individuals in the population are retained, and they are not involved in the cross-variation calculation, but are copied to the next generation. Other individuals are determined by roulette.

When the neuron is deleted by mutation operation, the corresponding weight coefficient would be reset to 0. When the variable operation increments a neuron, the code of the node would be initialized randomly. Since the weighting coefficients are generally encoded by floating point numbers, new mutation operations and crossover operations must be designed.

It is assumed that its crossover operator would be as described in formula (4) when interacting between the o-th and o + 1-th individuals:

X o t and X o + 1 t are a pair of individuals before crossing; X o t + 1 and X o + 1 t + 1 are the individuals after crossing; v o is a uniformly distributed random number in the interval [0,1].

It is assumed that the o-th individual is mutated, and its mutation operator would be as follows:

X o t refers to the individual before mutation, X o t + 1 refers to the individual after mutation, and v o refers to the random number uniformly distributed above interval [ i max − ε 1 − X o t , i max + ε 2 + X o t ] . This would ensure that the mutated individuals remain within the search interval.

Using training sample ϕ 1 to solve the problem, the greatest advantage of the network is not necessarily the extreme point, but the edge point. On this basis, the maximum value under the constraint conditions is taken as the objective function, so that the weight and structure of the network are not necessarily optimal. Therefore, although the accuracy of the function can be improved by increasing the evolution algebra, it is inevitable that “over-fitting” would occur. Moreover, this method is based on the data model of year M and does not take into account the impact of the closest sample on the prediction of the following M − M 1 . M-type samples are divided into three categories. The total sample M is divided into training sample ϕ 1 , test sample ϕ 3 , and training sample ϕ 2 :

After inputting another test sample ϕ 3 , it can be obtained as

This results in a set of weights and implied nodes of the network. Then, a test sample ϕ 3 is input to verify that the generalization of the network can be tested:

3.1 Computer network security test

In the network system security project, the network topology and network security risks are analyzed, and a comprehensive security prevention strategy is proposed:

Virus protection: It prevents virus invasion and killing.

Identity authentication: It identifies legal users.

Information encryption: It is a kind of message transmission password to prevent eavesdropping, leakage, tampering, and destruction of communication lines. It is information storage password-encrypted storage of confidential information.

Security audit: It can monitor the security status of the network in real time and discover the dynamics of the whole network in time; it can find network attacks and violations and record all information truthfully to provide evidence.

Access control: It restricts the use of users’ files and data, mainly to prevent users from unauthorized access.

Security and confidentiality management: Security and confidentiality work is a very important link. In terms of security and confidentiality, effective management should be strengthened to prevent its occurrence.

Vulnerability scanning: Network vulnerability scanners can be used to detect existing security risks and assist system administrators in finding security defects, thus providing better solutions for system administrators.

Intrusion detection: Various methods can be used to collect network or computer data to discover whether there is any violation of security policies and signs of attack in the network or computer system. When an attack is found, it would automatically send an alarm and take corresponding actions. At the same time, the whole process of the attack is recorded in detail, thus providing basic data for the recovery of the network or system and tracing the source of the attack.

The preparation process of the experiment is as follows:

1. Ensure that the protective measures are correctly assembled and the system patch is correct;

2. For unauthorized attacks, check the correctness of the protection policy;

3. Use network vulnerability tools to check system-related vulnerabilities (NBSI and IPhackerIP are two commonly used tools);

4. Collect Trojan tools and check the Trojan;

5. Use a variety of anti-plug-in tools to check the program plug-in vulnerabilities.

Specific experimental operations are as follows:

1. According to the system installation guidelines, install different security monitoring systems on the same two hosts to ensure that the operation parameters of the host are consistent, ensure that the operation of the system is within the monitoring range, and regulate the system status to make the system work effectively with the target host.

2. Simulate network attacks; carry out the same number, type, time, and place of network attacks on two hosts; and number the simulated network attacks to ensure the consistency of experimental parameters.

3. STATA14.0 and EXCEL2007 were used to count number of attacks effectively identified by the two systems, record the operation time of the system, input the result data into the corresponding database, and conduct multiple experiments to ensure the reliability of the experimental results.

In order to further improve the overall comparison effect of the two safety monitoring systems, this article needs to design corresponding experimental parameters for comparison. The experimental parameters are shown in Table 1.

The following would test the network security in the system and set the difficulty of simple, general, and difficult situations, respectively.

Figure 5 shows some contents related to network system security test.

Figure 5

Network system security test: (a) virus protection, identity authentication, and information encryption, and (b) security audit, access control, and vulnerability scanning.

It can be seen from Figure 5(a) that under simple and general difficulties, the accuracy of virus protection, identity authentication, and information encryption of the network system can reach 100.00%; in the degree of difficulty, the accuracy rate of virus protection and information encryption decreased to 98.34 and 98.08%, respectively, except that the accuracy rate of identity authentication still reached 100.00%.

It can be seen from Figure 5(b) that the accuracy rate of security audit, access control, and vulnerability scanning of the network system was still 100.00% under simple and general difficulties; in the degree of difficulty, the accuracy rate of the aforementioned parameters decreased to 99.32, 98.86, and 98.15%, respectively.

Figure 6 shows the relevant data of the system’s intrusion detection test and the security evaluation accuracy of different algorithms.

Figure 6

System intrusion detection test and security evaluation accuracy of different algorithms: (a) system intrusion detection test and (b) security evaluation accuracy of different algorithms.

It can be seen from Figure 6(a) that in the intrusion detection of the system, under the simple difficulty, the alarm accuracy of the network system, the filtering ability of the network system, and the integrity of the intrusion report were all 100.00%; in the case of general difficulty, the aforementioned parameters were 99.77, 99.48, and 99.89%, respectively; in the degree of difficulty, the aforementioned parameters decreased to 98.63, 97.75, and 97.55%, respectively.

It can be seen from Figure 6(b) that in the accuracy of network security evaluation using different algorithms, the accuracy rates of GA, BPNN, and GA–BPNN were 99.58, 99.15, and 99.92%, respectively, under simple difficulty; under general difficulty, the accuracy rates of GA, BPNN, and GA–BPNN were 98.21, 98.34, and 99.52%, respectively; in the degree of difficulty, the accuracy rates of GA, BPNN, and GA–BPNN were 96.72, 96.47, and 98.88%, respectively. It is not difficult to see that in any case, the accuracy rate of GA–BPNN was higher than that of GA and BPNN. It can be concluded that the GA–BPNN algorithm proposed in this article has certain accuracy in the evaluation of computer network security.

3.2 Remote control technology test

The following focused on testing the performance of the system and collecting a large number of test data. The collected data were tested, and the feasibility and correctness of the system were obtained.

In this experiment, the debugging of control performance is very critical. Its control effect directly affects the operation effect of the whole system, the removal of obstacles, and the transmission of safe and lasting information. Therefore, performance debugging should be carried out during the design process of the whole system. The test of control performance mainly consists of PLC hardware control, near-end human–machine interface control (touch screen), and remote monitoring center. First of all, each control module was simulated and debugged separately without information interaction. Second, the control modules were tested, and the corresponding communication debugging was carried out. Through the linkage debugging of each module, each module can cooperate with each other to ensure the stable, safe, and reliable operation of the system.

Independent debugging of control module: In the hardware control performance debugging of PLC, manual keys are required to debug and verify various equipment pieces on site. The digital input and output equipment shall be debugged without power on and verified point by point by software Man–machine interface control performance commissioning. Relevant output and equipment operation can be checked through the touch screen. When the corresponding command is input, it is necessary to confirm whether the device can be controlled as required, and display the relevant data of the sensor on the touch screen, so as to realize the confirmation of each link. The control ability of the remote monitoring center is to confirm the working condition of the relevant equipment and whether the system environment and equipment operation meet the requirements by sending commands. A comprehensive test has been conducted in the remote monitoring center. This is a very targeted method, which can help users quickly find problems and shorten the system debugging cycle.

Communication debugging: Communication debugging is mainly aimed at the communication between PLC and personal computer. Siemens programmable controller (STEP7) sets relevant parameters and corresponding addresses and uses Ethernet for communication. The color change of the indicator light can be observed through the hardware device of PLC, and green indicates that the communication is normal. PLC communicates with other devices using process fieldbus protocol and sets corresponding stations in STEP7. According to the upper communication rate, the switches and other relays can be confirmed one by one and the faults can be eliminated step by step. Then, the hardware circuit is tested until the communication between PLC and each equipment is restored to normal.

Linkage debugging: Based on the completion of each control performance module debugging and communication debugging, the overall linkage debugging is carried out. The control performance, communication performance, and overall program operation performance are linked. Through intuitive and efficient feedback on the operation of the equipment and adjustment of the corresponding programs, the overall requirements for agricultural production are realized.

The main goal of the system function test is to verify the debugging results of the system and to prove the feasibility of the system by verifying the system function.

Function test of PLC: After the hardware debugging on site, the data of the sensor is passed through ZigBee. According to the instructions of the operator, PLC has completed the start and stop of the fan, the closing of the skylight, the opening of the shutter, the light compensation ratio, the spray irrigation system, the heating and humidification system, etc.

Function test of human–computer interaction interface: It is the control center of the near end. It is used to adjust and control the execution device in the monitoring system through communication between PLC and Ethernet, and monitor the working condition of the device, as well as displaying the environmental parameters, reports, alarm information, etc. in real time, which can correct and effectively control the operation of the device.

Overall function test of intelligent agriculture system: The overall performance index of the system is mainly to test the performance of each module and effectively control the equipment under certain conditions. Smart agriculture refers to the collection of environmental parameters of crops in different periods and the real-time and accurate control of the operation of devices, so as to ensure that the growth environment of crops is in the most suitable state in the shortest time. At the same time, the response of the system and equipment in case of abnormal changes in data is tested to ensure that all parameters of the system are stable.

Figure 7 is the relevant data of the overall system test.

Figure 7

Overall test: (a) control performance test, (b) communication performance commissioning, and (c) linkage commissioning.

It can be seen from Figure 7(a) that the success rate of scene to scene jump of PCL control template, human–machine interface, and control center was 99.86, 99.74, and 99.05%, respectively, and the consumption time was 53, 48, and 56 ms, respectively.

It can be seen from Figure 7(b) that the jump success rates of PCL and upper machine, and PCL and equipment were 98.95 and 98.49%, respectively, and the elapsed time was 65 and 58 ms, respectively.

It can be seen from Figure 7(c) that the smoothness of cooperation between modules, devices, and programs was 98.55, 99.05, and 98.40%, respectively, and the consumption time was 66, 72, and 60 ms, respectively.

The aforementioned test results showed that the overall operation of the system was normal; there was no abnormal phenomenon during the detection process, and the functional operation of each module was in good condition; it fully met the environmental control requirements and design requirements of intelligent agricultural IoT system, and it achieved the expected results.

3.3 Application evaluation of smart agriculture

The experimental design mainly took CO 2 concentration and light as the experimental verification points. The experimental results were analyzed, and the control effect of the system and the effectiveness of the hardware architecture were obtained, which were combined with the software to ensure the stability of the whole system. CO 2 concentration and light are essential substances for crop growth. The comparison between them can better reflect the surrounding environmental conditions and system regulation.

CO 2 concentration is a necessary factor for photosynthesis of crops, and maintaining normal CO 2 concentration can ensure the normal growth of crops. According to the reference data, it is found that the comfort level of the real CO 2 concentration in the greenhouse is 530 mL/L. In the morning, the concentration of CO 2 can meet the needs of crops, but the concentration of CO 2 would change over time. Under the same environment, CO 2 was supplemented by building the supply structure of air supply type CO 2 , and corresponding experiments were carried out. In addition, without adding CO 2 , the experimental data were recorded and compared. By comparing the two groups, the CO 2 concentration supply data in Table 2 was obtained.

From the data in Table 2, it can be seen that from 6:00 to 8:30, CO 2 can meet the needs of crop growth with or without supplementation, but the concentration of CO 2 would change with time. After 8:30, if there is not enough CO 2 supplement, the concentration required for the photosynthesis of crops cannot be reached, and the crop treatment is in a stagnant state. When CO 2 concentration is supplied by air, CO 2 concentration can meet the needs of crops and reach the maximum CO 2 concentration at 10:30. As the crops consume CO 2 , the concentration of CO 2 would also decrease with time. By 12:00, the data of whether there is supply is close. The experimental results and the data of concentration change showed that the requirement of crop CO 2 concentration can be effectively adjusted using the air supply-type CO 2 concentration supply.

Light intensity and CO 2 concentration are necessary conditions for plant growth. The experiment proved that the intensity of light can ensure the normal operation of crops, thus providing the most suitable growth conditions for crops. At the same time, it can also test whether the light intensity would change in case of abnormality. The system would adjust it to the predetermined range according to different situations.

Through the collection and analysis of experimental data, the stability of the system and the real-time effectiveness of control can be effectively tested. The data was recorded using the data changes in the database, and the implementation of the actual control indicators was analyzed. The scheme in this article used the light-emitting diode (LED) light-filling system to fill the light and used the control actuator to adjust the start or stop, open, or close control of the lighting equipment (the ratio of red light to blue light is 6:3). The height of LED light-filling device would have a certain impact on the effect of light filling. According to the reference materials, the effect of 3 m light filling is the best. The theoretical light intensity irradiated to the surface is about 220 lux, and the specific changes of the light intensity are shown in Table 3.

It can be concluded from Table 3 that the light intensity at 6:00 am is not enough to meet the needs of crop growth. The light intensity in the morning is weak, but it would increase with time. At 8:00, after suddenly closing the sunshade and closing the curtain, the light intensity decreased rapidly, and the light intensity value reached 85 lux. Between 8:00 and 8:30, the PLC controls the LED fill system to fill the light. The light intensity value immediately increased and stabilized at 255 lux at 8:30. The brightness met the needs of crops. The experimental results showed that the system has a perfect light compensation system, and the light compensation effect is significant; the crops are filled with light in time, and the PLC control structure has good performance, which verified the correctness of the system scheme.