Abstract
The present industrial Internet platform (IIP) is subjected to numerous complex security issues, and thus technologies that focus on achieving data security of the IIP has become top priority. The data security crisis of IIP gets mainly reflected in the malicious deletion and theft of data, the random access of terminal devices, and the low security of traditional security authentication methods. Blockchain technology adopts distributed network architecture, wherein through asymmetric encryption mechanism, it helps to solve problems associated with single point of downtime and privacy data leakage of central management of industrial Internet. In addition, considering the accept of random access and deletion of industrial interconnection terminal equipment, this paper uses the role access control mechanism based on Hyperledger, thereby regulating user access according to the user’s role in the platform and management of terminal equipment through chain code. The paper includes the following contributions: (1) firstly, it expounds the research background and significance of the IIP and introduces the research status of the industrial Internet and blockchain at home and abroad. (2) Secondly, the IIP architecture is designed based on blockchain technology and implements a sensor node management and monitoring system. (3) Thirdly, the paper uses the gateway to preprocess the collected data, thereby reducing the data transmission delay and improving the real-time and effectiveness of the data. The paper finally implements neural network to evaluate the construction quality of industrial Internet sensor nodes and obtains promising test results.
1. Introduction
With the rapid development and update of information technology, it has become the current development trend to connect sensors and hardware devices to the network for management, and the IoT has also become a research hotspot at home and abroad. The IoT has changed the original concept in the continuous development and is no longer limited to sensing technologies such as radio frequency identification, and the coverage has been expanded. It breaks the limitation that the network is only connected by terminal devices such as computers and realizes the interconnection and information exchange between people, people and things, and things and things [1, 2]. The various sensing devices connected to the IoT become the eyes and ears of humans to perceive the surrounding environment and objects, thereby realizing intelligent management, remote control, real-time positioning, and environmental monitoring of various things connected to the IoT. Further, due to the improvement of people’s requirements for quality of life and the needs of social development, large-scale IoT applications such as smart cities, smart homes, and smart transportation are also popular [3, 4]. In recent years, with the emergence and rise of 5G technology, it has provided new opportunities for the development of the IoT. In this context, all walks of life have joined the ranks of information and industry integration. At the same time, it has also accelerated the revolution of industrial Internet technology and has gradually formed an automated and information-based architecture system. Industrial Internet services take advantage of IoT communications and cover a wide range of applications including heavy industry, medical industry, aerospace, transportation, and other key national industries that have played an important role [5, 6]. In order to promote the development of industrial intelligence and build an enabling platform for efficient data collection and function management, the industrial Internet architecture is proposed. At the same time, with the continuous rise of new technologies such as AI, cloud computing, edge computing, 5G, and blockchain, integration with new technologies has become one of the current development directions. According to the main application scenarios and actual needs of the industrial Internet, the future will develop in the direction of intelligent management, optimization of production processes, flexible production, and industrial upgrading [7]. At the same time, the scale of the industry is expanding, and the upgrading of the industry also needs to be carried out simultaneously. Among them, it is inevitable to use large-scale sensors to collect a large amount of data, and data communication technology to implement effective real-time monitoring and management of production. Due to the alternation of new and old equipment in industrial production, the factory area is generally large, which leads to complex and scattered types of production equipment and sensor nodes connected to the industrial Internet [8]. At the same time, in order to ensure efficient and stable production, the number of monitoring nodes set up has also greatly increased, which has brought great pressure to equipment management and data transmission. In addition, compared with general IoT devices, the industrial production process is rapid and large scale; the production environment is complex, and the network environment is poor, which affects the transmission of data. In this way, the staff cannot obtain effective data for the first time to effectively monitor and manage the production process and industrial equipment. If the problem is not detected in time, it will lead to production delays and economic losses, so there are higher requirements for the real time and stability of data [9]. The traditional supervision platform is centralized in the cloud, and there are many factors affected by data transmission, and there is no good way to deal with the huge data. Therefore, traditional factory management methods have been unable to keep up with the speed and demand of the development of emerging industries at this stage. Therefore, enterprises urgently need a set of efficient and intelligent equipment management and control system to supervise and control their production equipment scientifically and efficiently [10]. The industrial Internet needs to adapt to various protocol devices to aggregate, collect, and transmit data of various types and formats. Gateways serve a crucial role in connecting one generation of technology to the next by bridging the gap between different types of networks and data transfer [11]. On top of standard gateway hardware, the industrial Internet gateway provides data management, filtering, analysis, monitoring, and administration capabilities to more broadly address the issue of data management [12]. Some security features of blockchain technology provide possible solutions to the security problems faced by the industrial Internet. The so-called blockchain is a multiparty participation and maintenance, and the data information on the chain supports anyone through the theory of data structure and cryptography, distributed ledger technology that cannot be modified. The blockchain has the characteristics of decentralized management, the entire network maintenance of the ledger data, the data information cannot be tampered with by anyone, and the data information security encryption. Each data block in the blockchain ledger is subject to multiple valid transaction confirmations by other nodes in the entire network. The consensus mechanism is used to maintain the consistency of all node information in the entire network, and cryptography is used to ensure the data security and unforgeability of the distributed ledger [13]. All of the nodes in a distributed network work together to update the blockchain ledger. The data in the global ledger is backed up locally on each node, where it is verified and managed. In addition, the identity access concept of blockchain technology may guarantee that only the data’s rightful owner can access the data. At the same time, blockchain enables the decentralized trading and storing of digital assets. The usage of smart contracts in a blockchain environment guarantees that agreed-upon conditions will be met [14]. A smart contract is a piece of code written in a computer language, which is a protocol that is run by nodes in a blockchain for automation [15]. This research designs and implements a set of industrial Internet sensor node management and monitoring system based on blockchain. According to the results of actual data collection through multiple tests, there is a big difference between the actual required valid data and the collected complete data. Invalid data will lead to slow system processing progress and increase response time, so this paper uses gateway to preprocess the collected data, thereby reducing the data transmission delay and improving the real-time and validity of the data. At the same time, a new sensor node identification method is designed, which is easy to manage, and each piece of data can be quickly traced back to the source to improve management efficiency. This paper also uses smart contracts to manage configuration files in an Industrial Internet environment. After completing the above work, this paper uses neural network to evaluate the construction quality of industrial Internet sensor nodes.
The unique contributions of the paper include the following: (i)Development of an IIP architecture considering blockchain technology to implement sensor node management and system(ii)Use of gateway for processing the collected data, reducing the data transmission delay thereby improving the real-time and effectiveness of the data
The organization of the paper is as follows. Section 2 discusses the related study. Section 3 presents the methods followed by experimental results and analysis in Section 4. Section 5 presents the conclusion.
2. Related Work
The Internet of Things (IoT) is a chance to construct an intelligent environment made possible by the integration of the physical and network worlds, which emerged as a result of the steady progress of science and technology in modern civilization. The Industrial Internet is a branch of the IoT, focusing more on the interaction and information transfer between devices and people, and has very broad development and application prospects. With the rapid rise of a new round of scientific and technological innovation revolution around the world, it not only promotes the sustainable development of the manufacturing industry, but at the same time, its transformation to digitalization and intelligence has also been further promoted to the national strategic height [16, 17]. The IIP is an important part of the industrial Internet. It is mainly supported by new-generation innovative technologies such as big data, cloud computing, machine learning, and AI as well as smart devices. It is an application of “Internet +” deep integration in the industrial field. The IIP contains several main goals: to effectively allocate industrial production resources and comprehensively improve production and management efficiency [18]. In addition, it also has a positive impact on the digital and intelligent transformation of the manufacturing industry, completing the upgrading of industrial intelligent production and management, and is also conducive to the upgrading of industrial production and flexible customization. From a global analysis, the industrial Internet has gradually formed its own technical system, and the R&D of related companies has begun to take shape, and the applicable scenarios are increasingly enriched [19]. There are some typical examples of IIP, mainly including HiaCloud platform as an equipment and automation enterprise; the research and development focus on scenarios such as reconstructing the production and maintenance operations of enterprises in the digital environment. Siemens’ MindSphere platform uses the cloud platform to realize the fault early warning function. In addition, the Haier-COSMOPlat platform is also a relatively mature IIP at this stage [20]. As a big manufacturing country, although China is still in its infancy in the construction of an IIP, its technological infrastructure and comprehensive capabilities are relatively weak, and the platform covering the entire industry has not yet been completed. However, the platforms for certain industries and fields have begun to take effect and are gradually expanding their influence. The relatively mature IIP in my country at this stage include Yonyou IIP, Aerospace Cloud INDICS Platform, Inspur IIP, etc. [21]. Reference [22] proposes a cloud-based IIP that combines “device management” and “application enablement”. The use of microservice technology reduces the coupling between management platforms, supports flexible access to multiple applications, is easy to expand and develop, and breaks the current status quo of fragmented development and low utilization in the current industrial Internet industry. Integrate partners on the entire industry line to provide a safe and efficient “one-stop” service platform. This solution mainly solves the problem of multiple application multiplexing and access. It does not flexibly use the gateway as the core device in the industrial Internet and does not provide a good solution at the data level. Reference [23] designed and developed a blockchain-based industrial IoT service platform. In order to adapt to large-scale equipment access, the article selected the Logchain blockchain system. Based on M2M IoT standards and blockchain hybrid applications, a blockchain-enabled IoT service layer platform is built to provide IoT users the option of using either centralized or decentralized databases to keep track of their data. As an interconnection device of a complex network, the gateway plays an indispensable role in the network system and is used to realize the interconnection and intercommunication of different networks and perform protocol conversion. Reference [24] proposed a smart home system platform based on Raspberry-PI gateway and cloud service for interoperability between various traditional home appliances and different communication technologies and protocols. It provides the functions of controlling home appliances and analyzing data, but the application scenarios are relatively limited, and there is no better way to deal with the real time and effectiveness of data transmission. It is not suitable for complex scenarios such as industrial environments. Reference [25] designed an IoT gateway monitoring system based on edge computing. In view of the diversity of IoT device communication protocols and the lack of cloud computing power, edge computing technology was used. Based on EdgeX Foundry framework, a set of temperature and humidity alarm system using Modbus protocol is constructed, which has high application value. However, it is mainly aimed at the early warning function, and there are still deficiencies in the functions of monitoring and management. At the same time, the expansion performance and maintainability of the system are poor. Reference [26] proposes a dual-deposit escrow transaction protocol that combines bilateral payment deposits with simple cryptographic primitives, implemented using blockchain-based smart contracts. The security argument of the protocol is carried out by means of a game, which proves that the complete Nash’s equilibrium of the subgame of the game is a game in which buyers and sellers cooperate and are honest and trustworthy. Reference [27] proposed to use the characteristics of blockchain technology to solve the problems of information asymmetry in the current instrument leasing platform. Combining blockchain technology in the leasing platform allows lessor and lessee nodes to build a decentralized blockchain network and install a smart contract in it to complete the leasing process, so that the transaction information generated during leasing and the data generated when the instrument is used can be uploaded to the blockchain, and then a consensus is reached to form a block record in the network. Reference [28] proposes a general framework for authentication and authorization in a restricted environment. Access to terminal devices requires authentication and authorization to ensure the security of the platform. Addresses a major limitation of the datagram transport layer security protocol by protecting application-layer paid payloads. Reference [29] also proposed the distributed ledger management technology through blockchain to realize the sharing of IoT data and solve the problem of trust in traditional centralized institutions. The monitoring and management of terminal equipment is very important to the security of the IIP, which ensures the continuous, efficient and trouble-free operation of industrial equipment. Reference [30] mainly measures and tracks the state of the network. They provide different levels of management through plugins and extensions. One method of network monitoring is to use probes to measure network indicators to actively monitor terminal device status. By configuring pairs of network devices, a certain amount of traffic is introduced into the network to monitor key indicators and ensure the quality of device operation. Reference [31] developed a food anticounterfeiting traceability system using Blockchain and the Internet of Things. The framework used decentralized storage technology and Blockchain to store traceability related data for food during the process of food production; sale and transportation to ensure uniqueness of the food are retained. The IoT technology helps in maintaining authenticity and reliability of the data stored in the Blockchain. Reference [32] a layered architecture using Blockchain and machine learning is proposed. The study includes using industrial internet-of-things for smart manufacturing applications. The architecture consisted of five layers, namely, sensing, network layer, transport layer in association with Blockchain, applications, and advanced services. The Blockchain technology helped in gathering access control information, and machine learning helped to detect various forms of attacks, namely, Denial of Service (DoS), Distributed Denial of Service (DDoS), injection, man in the middle (MitM), brutforce, cross-site scripting, and scanning attacks. The framework was evaluated against the state of the art models considering accuracy, precision, sensitivity, and the Matthews correlation coefficient.
3. Method
3.1. Design of Industrial Internet Security Platform Based on Blockchain
3.1.1. Requirements’ Analysis
In recent years, while the IIP has brought convenience, intelligence, and full-factor connection, the security problems faced by the IIP have become increasingly serious. Infrastructure as a service (IaaS) technology has gained quite some momentum having shorter life cycle. The developers of platform as a service have enough knowledge to understand the various industrial aspects and implement professional technologies. The research and development of PaaS have not been able to bridge the gap between manufacturing and consumption. Thus, the development of IaaS on the existing cloud platforms leads to increased investment in cost and poor usability. There are also issues relevant to transmission of complex unstructured data that are diversified and also has variability. With the continuous development of information technology, the security of sensitive data on the IIP has become an increasingly serious problem, restricting the further development of the IIP. In the new strategic system for the development of the global industrial Internet, how to protect the security of platform data has become the research focus of the outlines of various countries. Since most of the traditional IIP use a centralized network architecture, when the central server encounters an external hacker attack, the platform will not only face the consequences of paralysis but also face the serious situation of data theft or malicious modification. Businesses will face irreparable losses. In addition, in terms of terminal equipment management, the management of underlying terminal equipment by traditional IIP is rather chaotic. The data stored in the database is greatly polluted, resulting in the embarrassing situation of being unable to convert data into value in the face of massive data information. In terms of permissions to the platform, the central server will have a super administrator who controls the permissions of the entire platform. He can manipulate any data information on the platform, making the platform data lack of credibility and credibility, and also causing the problem of self-stealing within the personnel. Therefore, based on the above security issues, it is very necessary to develop a blockchain-based decentralized secure Internet platform, and it is also an imminent practical need. Hyperledger Fabric is an open source Blockchain permission which was initiated in 2015. It is a modular general-purpose framework that provides unique identity management and also renders access control features suitable to be implemented in various industrial applications. The framework includes a unique organization of members that interact with each other on the network. The transaction flow is initiated when a client application sends transaction proposal to the peers in each organization for endorsement. The peers authenticate the identity of the submitting clients and authority for submitting the transaction. The outcome of the proposed transaction is simulated, and if it matched the expected result, an endorsement signature is sent back to the client. The client collects the endorsements from the peers, and when defined numbers of endorsements are received, the transaction is sent to the ordering service. The ordering service checks if the required numbers of endorsements are received; then, the approved transactions are packaged into blocks, and the blocks are sent to peer nodes in each organization. The peer node validates the transactions and then adds a new block to the ledger, and the status of the ledger gets updated making new transaction committed. Based on the Hyperledger blockchain development platform and using a series of development tools for the Hyperledger ecosystem, this paper builds a blockchain-based industrial Internet security platform. Through the role-based access mechanism, it is ensured that the data in the platform can only be operated by specific personnel, and the terminal data is encrypted, packaged, and uploaded to a traceable and nontamperable blockchain network. At the same time, it manages specific terminal equipment based on role-based permissions to achieve the effect that a specific person is responsible for the specific equipment. At the same time, the faulty equipment can be quickly held accountable, repaired and operated quickly. The specific functions are as follows. (1)User Role Management. Different management rights are given based on the roles assigned by users; each role has its own key, and the key can be used as user authentication and rights management rights. The platform has different user rights to define the terminal equipment that the user can operate and the database information content that can be operated. In this way, the effect of data isolation can be achieved according to different roles of users(2)Data Query Function. According to the user’s role level in the platform, the user’s operation authority is determined, such as data reading, data uploading, etc. For some sensitive data, the platform sets that only roles with specific permissions can view it, preventing the possibility of unrelated personnel from leaking data within the enterprise. At the same time, users can also choose the data information on the chain according to their own needs. Users can also view the running status of the entire blockchain network and quickly repair faulty nodes(3)Device Management Function. It can collect and process terminal device data and use the device’s own key to package and upload it to the blockchain network. The configuration file of the device can be viewed, updated, and replaced through the blockchain network(4)Chaincode Management Function. Administrators can deploy new chaincodes according to business needs and can also update and replace old chaincodes(5)Blockchain Network Management Function. Administrators can deploy network nodes, establish secure channels, add or delete platform organizations, and authorize and revoke certificates
3.1.2. Platform Design Principles
Business requirements are the source of platform design. This is the beginning of designing a platform. The first criterion for evaluating the quality of a platform is suitability. The data security of the Industrial Internet is one of the main reasons that restrict its development, and the blockchain technology is unique in its data structure and storage method which can ensure the security of data and information and can perfectly solve the pain points of the current industrial Internet. With the continuous change of industrial Internet requirements and the sharp increase in the amount of data, this platform can quickly add new services on the original basis through a modular mechanism to meet changing business needs. The reliability of this platform includes two aspects; one is to effectively ensure the safety and reliability of platform data. Instead, it can ensure the long-term safe and stable operation of the platform. The spatial data and attribute data of this platform can be organically combined and interacted, and the data information between each module can be transmitted to each other to achieve smooth interaction. This platform adopts visual page design and process-style structured design. After simple business training, users can take up jobs quickly, saving enterprise training costs, saving time, and improving personnel efficiency. This platform adopts a role-based data access and management method. For nonprivate data, it can realize data sharing and improve the closeness of collaboration between various departments. For private data, the data is encrypted on the chain through asymmetric encryption, and only those with the private key can access and view the data. Compared with other traditional platform development solutions, it can greatly reduce development time and platform development costs.
3.1.3. Platform Architecture Design
Through the analysis of the requirements mentioned above, the overall architecture of the blockchain-based feature design platform is shown in Figure 1.
The blockchain network is added to the architecture of the traditional IIP, and the blockchain technology is used to manage terminal equipment, data resources, and access rules. The main change is reflected in the addition of the blockchain platform to the IaaS layer, connecting the edge layer and the platform layer through the blockchain network. IaaS is a cloud service that provides access to completely provisioned on-demand computing infrastructure that is possible to be managed over the internet. The IaaS model enables companies to access all the benefits of computing resource without facing overhead issues in the deployment, maintenance, and operation of in-house infrastructure. IaaS also provides scalability and resource management enabling consumers to work on a “pay and use” model wherein they can pay rent and use additional computational resources when enhanced performance is required. The use of IaaS in blockchain helps organizations unlock inexhaustible pool of computing, network, and storage resources in order to develop the best possible infrastructure for their business. The management of terminal devices at the edge layer is mainly controlled by the chain code. The chain code runs in the blockchain network, and the terminal device can obtain updated configuration information from the chain code in real time. At the platform layer, data management is mainly carried out through asymmetric encryption technology, and the data is packaged, encrypted, and stored in a distributed database to ensure the security of data resources. Users use their own digital certificates to authenticate through the client and then enter the blockchain network. Users can do the following operations according to their operating permissions. (1) Users can access information and data within their own authority. Some of the data on the blockchain are privately encrypted data, and only users with a specific key can view this part of the data. The other part is the public dataset which can be viewed and audited by all users on the blockchain. (2) Users can perform specific operations on devices within their authority, such as viewing or updating device configuration files. Of course, the records of these operations of the user will be written into the blockchain, and if the equipment is abnormal, it can be quickly repaired and the responsibility can be determined. The data generated by the terminal device is encrypted by the device’s own unique Mac address to form a unique pair of public and private keys. In this way, information with sensitive data can be encrypted with its own public key and uploaded to the blockchain network. Then, only operators who have the private key of the device can decrypt the terminal data and view the private data. Once the data is on the chain, the data cannot only be traced back but also cannot be tampered with, which not only increases the credibility of the enterprise but also ensures the integrity and credibility of the data.
3.2. Permission Management
3.2.1. User Rights Division
The model file includes definitions for all users who interact with the platform, including administrators who create access control rules and other authenticated users who need access to restricted resources. In this scenario, each participant is identified by a unique ID, and their characteristics are also tracked. Users with different identities have different attributes and operation permissions. Ordinary users act as requesters of information access and obtain data information through the authority assigned by the organization manager. Compared with ordinary users, organization administrators have the authority to manage ordinary users, user certificate security authentication, and channel management. Organization administrators ensure the safe operation of the entire platform. In this article’s application, there is one transaction for granting access and another for revoking user access. Additionally, there is a transaction that delegates permissions to other users, who will then be able to transfer the access they have been granted to other users.
3.2.2. User Permission Operation
This article defines the access control policy when implementing user rights access control. Below is a list of the rules considered in the application in this article. These rules include the following. (1) A specific role can only access the resources specified by the permission.(2) Users with specific roles can send transactions. (3) Access rights for different modules of the platform. (4) Members of certain groups will be granted access to the archive. The five sets that make up an access policy are the actors, the resources, the circumstances, the behaviors, and the actions. The access control model in this study is a role-based access control model because, in the base model, users have access to resources based on the roles established in the platform. This article provides many classes of actors to represent various organizational responsibilities within the context of the access control architecture presented here. They may also have automatic access to some resources, depending on their jobs and circumstances. While initial ACP module definitions are static, an authorized user may submit a transaction to dynamically modify a user’s access control. Events are an important part of the platform when used with platform queries. The event module is used to query the log of transaction information. The log entry indicates that the result of the event was fired from the transaction function. Also, they can call external applications. This article considers the case of persisting access requests and denying requests after several consecutive efforts. As a result, triggers for external applications will be triggered in response to necessary security considerations such as triggering intrusion alarms to prevent unauthorized access.
3.2.3. Role-Based Permission Control Process
Access control and authorization is extremely important in hyperledger composing wherein the security architecture of business networks shared by organization members in the Blockchain. Hyperledger composer enables admin control on the resources or data; a participant is entitled to access in a business network. Hyperledger fabric uses access control lists to manage access to the resources by associating policies with the resources. The role access control mechanism based on ACP policy is implemented through chain code, which can be regarded as a smart contract in Hyperledger. The main task of chaincode is to define the logic of each transaction and the conditions that need to be met. Once these logics and rules are written into the chain code, they will be automatically executed and will not change unless the new chain code is used to replace the rules on the old chain code. When the corresponding transaction is committed, the corresponding transaction handler function is automatically called. The above figure shows the access control process based on ACP. First, the user submits the access rights transaction, and the platform first checks the user’s rights according to the ACP access rights policy. If the transaction submitted by the user does not have permission to access, the platform will send an error message and return. If the user passes the ACP mechanism, the platform will call the prewritten authorization rules of the chain code on the blockchain for rejudgment. If the authorization rule is passed, the authorization API will be called for authorization, otherwise an error message will be sent and returned. In the permission access process, it is necessary to pass two permission checks because, if each permission check directly calls the chain code on the blockchain for judgment, the efficiency is very low. If some unauthorized transaction accesses are filtered through the ACP policy first, the efficiency of the platform will be greatly improved, and the smooth operation of the platform will be ensured.
3.3. Terminal Device Management
3.3.1. Network Management Protocol
The supervision and management of the enterprise network is very important, and it is the basic issue to ensure the network service. Changes in network node configuration may be made automatically based on observations of network behavior, which can be gathered via a variety of methods of monitoring the status of network nodes and measuring their performance. In order to have a full picture of the state of a network, it is necessary to monitor not only the performance and traffic but also the status of each device and interface. Syslog is a widely used standard protocol for logging and transmitting networked notification messages about state transitions. While simple network management protocol (SNMP) is a specific protocol for monitoring and basic administration of network devices, syslog is routinely used to track changes on any PC, server, or device. In the IIP, there are many ways to manage network devices. Most of the IIP are directly managed through the command line interface, but this method also requires the most manual labor. Therefore, it is not economical and cannot scale equipment efficiently, and the management network of increasing scale and complexity requires more platform based and automated solutions. The SNMPv3 protocol adds configuration rules for remote devices. However, the SNMP protocol is still mainly used for monitoring, because the support of this protocol for modifying device properties is very limited, and it needs to rely on the technical support of the manufacturer. NETCONF, on the other hand, focuses on device configuration through an open application programming interface using an extensible markup language-based device behavior model.
3.3.2. Device Management Process
The sequence flow of blockchain-based terminal device management is as follows. To begin, digital certificates are used by legitimate network administrators to prove their identity. Following this, users may make changes to the blockchain-recorded configuration of devices, so long as they have permission to do so for that device or set of devices. Hyperledger describes a blockchain-based architecture that may be used to maintain authentication certificates. To prevent adding unintentional human mistake into the configuration stored in the blockchain, it is important to perform syntax checking on device configuration files that have been updated. Moreover, new configurations may be signed with the administrator’s certificate for operational identification and attribution. Time stamp, administrator ID, device ID, and encrypted device configuration are the components of a transaction. When a transaction is recorded in a newly added block to the blockchain, that block’s peers get a copy of the transaction. When a new block is introduced to the blockchain, an event will be sent to alert all managed devices to incorporate the new block into the blockchain. This will allow the device ID to determine whether the change changes the device’s settings. The device then uses its private key to decrypt the blockchain-stored configuration and apply the updated settings locally. The blockchain keeps a record of every modification that can be checked by audit and security teams. The specific sequence of the process of changing the configuration file of the terminal device is as follows. (1) The administrator obtains the old configuration file of a specific device or device group from the blockchain network and decrypts the new configuration file through his own private key. (2) The administrator modifies the old configuration file and (3) performs semantic verification on the modified configuration file, so that the modified configuration file conforms to the grammar rules. (4) The semantically verified configuration file is encrypted and written into a new block, and the block is added to the blockchain after being sorted by the sorting node. (5) The administrator receives a notification that the configuration file has been successfully distributed into the blockchain network. (6) The device goes to the blockchain network to query the configuration file information if it is selected to download the configuration file and decrypt the configuration file with its own private key for loading. (7) After the device loads the application, the new block information, including whether the new configuration file has been successfully applied, the hash value of the configuration file, and the download and application timestamps are packaged and written into the block for security auditing.
3.3.3. Chain Code Design
In the Fabric architecture system, chain code is equivalent to the implementation of smart contracts in the blockchain network. There are two types of chain codes: user-level chain codes and platform-level chain codes. The platform-level chain codes are responsible for the processing of the Fabric node’s own platform configuration, endorsement, and verification. User-level chaincode is designed by developers according to their own development needs. It provides state processing logic based on blockchain distributed ledger, and a variety of complex applications can be developed based on it. After the chaincode is deployed, it automatically runs in the blockchain network of Fabric and runs in the isolation sandbox. Nodes can interact with the chain code according to the protocol and operate the distributed ledger data. After starting the Fabric network, users can operate the chaincode in various ways to check whether the network is running well. The fundamental rule of creating a chaincode is that it must not include any secret information. All mandatory data is sent in the form of parameters, and authentication is handled by means of a key-pair consisting of a username and a password. When attempting to utilize any of the CRUD functions in chaincode, an administrator must first input the appropriate credentials. CRUD function is an acronym used in computer programming that includes four functions that are implemented for performing storage related applications. These functions are create, read, update, and delete. If the right login and incorrect key are given, the read, update, and delete operations will fail. This article refers to this process as the key verification method since the chaincode will attempt to decrypt the configuration file as a security precaution, and if this decryption fails, the whole request will fail. When performing the create operation, the configuration file is encrypted with the specified key since it does not yet exist, and there is no way to check that the key is legitimate. Forged requests still need an attacker to change the network device’s settings in order to cause it to download the configuration file and retrieve the device ID. If they have, this security issue is beyond the control of this article. Therefore, the worst case is that any configuration file located in the blockchain is set, or the database is never accessed by the network device.
3.4. Data Safe Storage
3.4.1. Data Upload Process
Data security is the core of an enterprise. Ensuring the integrity, privacy, and availability of data is the basic requirement of an IIP and the basis of network security. The collection and storage of data is mainly manifested in the edge layer and the IaaS layer in the blockchain-based platform architecture, and the terminal data is uploaded to the chain through six links. They are data collection, network isolation and data caching, signature packaging, edge data processing, sorting consensus, and distributed database storage.
3.4.2. Data Collection and Signature
The packaged blockchain has its own unique identity rights management function. Through the security authentication of the terminal device, a unique public and private key is generated for each terminal device. The public and private keys correspond to the unique IP of the terminal device itself, so that a list corresponding to the public and private keys of the terminal device is formed. Through the function of identity authentication authority, on the one hand, the random access and malicious destruction of terminal devices are prevented, and the security of data is protected. On the other hand, blockchain adopts a distributed network architecture. Compared with the centralized data collection for high-frequency data collection in the production process of industrial equipment, the load pressure of data collection and storage on the platform can be greatly reduced. Through the industrial gatekeeper technology, network isolation and data caching are performed on the data collected from the terminal equipment. Then, by calling the SDK interface function of the blockchain platform, the public key of the terminal device is used to encrypt and package its own data. The encrypted data will have a unique data identity, so if the encrypted data is specifically tampered and deleted, the data identity will not match the original one. In this way, when the edge data is processed, it can be deleted, or an alarm can be notified to the administrator for subsequent processing. When the PaaS layer needs to process data, the encrypted data can be decrypted through the public-private key correspondence table. Platform as a Service (PaaS) focuses on the developers and the programmers enabling them to create, run, and manage their applications without botheration of developing and maintaining complex hardware infrastructures. The components that are required for developing and maintaining software applications are performed by the cloud provider ensuring that the developers have enough time to focus on code and new feature development. If it cannot be decrypted effectively, the data is an invalid data that has been tampered with or deleted, and it can be deleted and isolated or discarded. In this way, data tampering and leakage can be effectively prevented by packaging the signature of the data.
3.4.3. Network Node Consensus and Distributed Storage
The consensus mechanism is the security barrier for the blockchain to ensure the consistency of distributed ledgers, and the consensus mechanism is also the main factor restricting the efficiency of data on chain. The frequency of data collection by the IIP is generally at the HZ level, so it is very important to choose a consensus mechanism suitable for the IIP. At present, the mainstream consensus is PoW, POS, and other consensus mechanisms that use incentive measures. This type of consensus mechanism ensures the safe and smooth operation of the blockchain network, but it is inefficient and cannot meet industrial-level data requirements. Hyperledger Fabric adopts PBFT and Kafka sorting consensus, which can greatly reduce the time required for consensus and effectively relieve the load pressure of the data uploading process. Different from the centralized data storage method used by traditional IIP, blockchain technology uses distributed data storage. There is no central node in the server nodes; they are all equal, and each node must back up the complete data storage. In this way, even if a node is attacked and cannot work, other nodes can still run, ensuring the normal operation of the platform and not causing the entire platform to stop serving due to a single point of downtime. Blockchain technology also uses cryptography for data storage, which ensures that data cannot be tampered with integrity and authenticity and fundamentally ensures the security of data in the industrial Internet platform.
3.5. BP Neural Network Theory
3.5.1. BP Neural Network Idea and Network Structure
In terms of multilayer feedforward neural networks, the BPNN is a classic model. To ensure that each layer is error-free at all times. Both signal forward propagation and error reverse propagation are included in the BPNN’s learning process. Each hidden layer and connection weight processes and calculates the input signal, which is then output in the complete forward direction by the input layer. This is the end outcome of the propagation process itself. Back propagation of errors is a process that uses an error function to determine the difference between a target anticipated value and the final output. If the error reaches the desired error level, the learning process finishes; otherwise, the mistake will propagate backwards through all hidden levels to the input layer in some form. When the final error value achieves the network’s goal error requirement or the number of iterations specified by the network is reached, the learning process comes to a conclusion, and the network’s error value is assigned to each layer. For nonlinear processing, the three-layer BPNN with one hidden layer is extensively utilized.
3.5.2. BP Neural Network Learning Process
What happens when there are three layers of neurons (input, hidden, and output) with data samples that are all equal in number? The process of learning is as follows. Network target expects the output feature vector to be , which is the input feature vector of the input layer. is an in-layer feature vector. Neuron thresholds in each of these layers are based on these weights and these thresholds, which are used to connect each layer together (, , , and ). The activation function is a sigmoid function. These are the steps in the BPNN learning process as a result of this. (1)Establish a connection to the network and begin working. Weights and thresholds of the network are given random integers between 0.5 and 0.5, and the network’s target accuracy, the maximum number of repetitions M, and the error function are established(2)In the data collection, choose sample , input vector , and the anticipated output value by random selection ()(3)Using the sample data , the connection weight and the hidden layer’s threshold determine the input value and the output value for each neuron in the hidden layer(4)Calculate each neuron’s input and output values and based on the hidden layer’s output, connection weight, and output layer threshold
A forward propagation technique is used in the BPNN’s learning process. There are several layers of transmission of data from the input to the output; this section discusses the error back propagation process, which is used to remedy forward propagation mistakes. (5)Calculate the error between the actual output and the predicted goal output using the error function, and then apply the partial derivative to each output layer neuron(6)Use the output of hidden layer neurons and the partial derivatives of neurons in the output layer to change connection weights and thresholds between hidden and output layers. is the value before correction, while is the value after correction. The correction formula is as follows:(7)In order to get a partial derivative of neurons in the hidden layer, you must take into account their connection weight, partial derivative of output layer error, and their hidden-layer output. As a result, the weight and threshold of each neuron in the input layer may be adjusted to match those of neurons in the hidden layer(8)Calculate the global total error as (9)Judge whether the network error satisfies ; if so, the BPNN learning process ends. Otherwise, randomly select the next sample and turn to step (3) to continue learning and training the sample until the error meets the requirements or the number of iterations reaches the maximum number of iterations and the training ends
3.5.3. Limitations of BP Neural Networks
According to mathematical theory, BPNN can handle issues involving intricate internal systems and can do any sophisticated nonlinear mapping. There are certain restrictions to the capacity of the BPNN to perform nonlinear mappings. (1) The network structure is difficult to decipher. The fundamental reason for this is the absence of adequate theoretical advice for determining the number of hidden layers and the number of neurons in each hidden layer. (2) During the learning phase, the error convergence speed is sluggish. This happens because the gradient descent approach of BPNN convergence means that the number of iterations of network training rises, while the error drops slowly or even stay the same. (3) The learning process is easy to fall into the minimum value. In practical applications, the BPNN may not find the desired solution and fall into a local minimum during the training process, resulting in the failure of the network structure error to converge well. (4) The learning step size affects the network convergence speed. If the learning step size of BPNN is too large, the network will be unstable, and if it is too small, it will affect the convergence speed and cause long training time. (5) The training and learning of the network is unstable. There is no fixed method to find the best weights and thresholds for the initial selection of weights and thresholds between the connection layers. When the samples change, the trained network model has to retrain the network. For nonlinear systems, the initial value has a lot to do with whether the learning can reach the minimum value and whether the result converges. The BPNN has its limitations wherein the actual performance of the model is dependent on the input data. Also the model tends to be extremely sensitive towards noisy data, and hence, matrix-based approach is considered preferable instead of using minibatches.
4. Experiment and Analysis
4.1. Data Source and Preprocessing
In this paper, a dataset is constructed for experiments based on the data collected by the gateway industrial Internet sensor nodes. The dataset includes 1800 sets of data.
The sample data is normalized and preprocessed since the network’s input data has distinct dimensions and physical meanings; the input features have different numerical ranges, and the numerical ranges between different features vary substantially. To make weight adjustment easier given the high fluctuation in the input value, the data is normalized and then transformed to (0, 1) or (-1, 1). However, the BPNN has a sigmoid excitation function with a derivative that varies between (0, 1) and (-1, 1) across a wider range of values. In order to improve the BPNN’s computational efficiency, normalizing sample data aids the network in reaching a quick convergence. In this study, we choose to normalize our data using the following procedure.
4.2. System Function Test
Each test is more than 1500 times; the average value is taken after the test, and the results of 8 tests are obtained in total. (1)System Delay. This indicator is also a very important indicator in the system. The final result of the experiment is shown in Figure 2. It can be seen that the delay of the system is very small, which proves the efficiency of the system to transmit data(2)System Downtime Rate. This indicator is one of the most important indicators of the information system, which is related to whether the system can complete the basic functions efficiently and accurately. The same 8 experiments were performed on the system to verify its downtime times, and the results obtained are shown in Figure 3. As can be seen from the figure, the downtime rate of the system is very low, which can prove that the system is very stable(3)System Accuracy Rate. The system accuracy rate is one of the important indicators to ensure the accurate transmission of system data. If the accuracy rate is too low, it will affect the user experience and reduce trust. The same method of taking the average of 8 experiments is adopted, and the results obtained are shown in Figure 4. It can be seen that the accuracy of the system is very high and can meet the needs of users
4.3. Model Parameter Analysis
We utilize the method to get a ballpark estimate of how many neurons should be in the hidden layer of the model network structure, and then we use experimentation to hone down on the best value. 15 neurons are present in the input layer, but only 1 is present in the output layer. Training data is used to build a reliable model, whereas test data is used to double check the model’s accuracy. Figure 5 depicts the prediction accuracy change curve when the number of neurons in the hidden layer of the model is varied. As can be observed in the picture, the prediction accuracy of the model peaks at 11 neurons in the hidden layer, which is optimal for gauging the quality of IIP creation. Therefore, it is preferable to choose 11 neurons for the hidden layer in the subsequent tests, taking into account the total performance metrics. Since the learning rate might affect the experiment’s precision, we choose several learning rates and plot their effects in Figure 6. It is clear that the optimal performance is reached with a learning rate of 0.7.
4.4. Model Evaluation Accuracy Experiment
In order to verify the validity of the model proposed in this chapter in evaluating the construction quality of industrial Internet sensor nodes. The test data of each dataset is fed into the model, and the obtained results are compared with the expert results, as shown in Table 1. From the experimental results in the table, the output of the model proposed in this paper is very close to the expert evaluation results, and the error is very small. Therefore, it can be proved that the model proposed in this paper has good performance in evaluating the construction quality of industrial Internet sensor nodes.
5. Conclusion
Aiming at the problems such as the difficulty in guaranteeing the data security of the IIP, the confusion of terminal equipment management, and the unclear division of roles and permissions, this paper designs a secure industrial Internet architecture based on the Hyperledger fabric framework in combination with the security features of the blockchain. This paper has completed the following work: (1) this paper firstly expounds the research background and significance of the IIP and introduces the research status of the industrial Internet and blockchain at home and abroad. (2) Design an IIP architecture based on blockchain technology, and implement a sensor node management and monitoring system. 3) According to the results of actual data collection through multiple tests, there is a big difference between the actual required valid data and the collected complete data. Invalid data will lead to slow system processing progress and increase response time, so this paper uses gateway to preprocess the collected data, thereby reducing the data transmission delay and improving the real time and validity of the data. At the same time, a new sensor node identification method is designed, which is easy to manage, and each piece of data can be quickly traced back to the source to improve management efficiency. This paper also uses smart contracts to manage configuration files in an Industrial Internet environment. After completing the above work, this paper uses neural network to evaluate the construction quality of industrial Internet sensor nodes.
Data Availability
The datasets used during the current study are available from the corresponding author on reasonable request.
Conflicts of Interest
The author declares that he has no conflict of interest.