How Can Applications of Blockchain and Artificial Intelligence Improve Performance of Internet of Things? -- A Survey

The IoT is the latest technology evolving to connect heterogeneous devices for communication. The IoT is known as a growing technology through which anything can be communicated and connected to each other anywhere and anytime. The IoT defines an environment where computing and communication systems are embedded to each other to achieve specified goals by interchanging and collecting data. The IoT technology refers to a scenario where sensors, computation ability, and network are integrated with each other to consume, generate, and exchange data with a minor or without human interference. The generation of advanced technological age is possible using the IoT where things in real world are enabled using various applications. The IoT is also able to provide a network-based environment where sharing of information, handling of resources and storage, properly organization of data, etc., can be done quickly and efficiently [49]. This environment can be distributed in any of IoT-based applications areas.

Blockchain elevates the developments of Industrial world. It has designed and processed new business models which were impossible to shape few years ago. Blockchain secured many segments, like finance, healthcare, manufacturing, etc., through the effect of current business models. In organizations, the business, individuals, software agents, and a group of users are supported by many methods of blockchain for sharing secure records and transactions among them. Blockchain strengthens a group of individuals to make a settlement on a definite activity and this settled activity can also be registered, secured, shared, traced, tracked, and measured through blockchain. This agreed activity may be a payment deal, purchase activity, and medical lab-tests for a patient, voting activity, and contract agreement or it may be the supply chain of the process of planning, implementing, controlling, storage, services, and consumption of goods. As a technology, blockchain unites the benefits of both equal-level networks and the ordered steps of mathematical problems of the confidentiality of messages, integrity of messages, sender authentication, and assurances to make sure the reliability of the managed agreements. Agreed, approved, and recorded activity cannot be changed by any particular participant without connecting others. Blockchain can assure the state of error free, arranged functions over time. Connecting strongly two or more entities to register an agreement securely of definite tasks without knowing each other is one of the main characteristics of blockchain. This agreement cannot be recreated or removed by anyone in that group. But, they can only review the previous transactions easily without removing and modifying it. Activities can be easily recorded and preserved securely through blockchain by a group of people or organizations [51]. Moreover, the process of maintenance of blockchain makes an unchanged history of activities of a group’s members. The process, known as ’mining’, is used to authenticate and to grow dependency of the supervised agreements which are attached in the chain. Many industrial applications take support the internet transactions of blockchain. Blockchain is a developing technology which was previously created in Bitcoin area. But nowadays blockchain is applicable in multiple sectors by its secure and updated structure.

Blockchain technologies can be utilized in distinct sets of applications which are shown in Fig. 3. Blockchain technology can find remedies in multiple sectors like voting, healthcare, supply chain, identity management, energy resources, governance, etc. Different applications of blockchain are as follows.

• 1.

  Blockchain can be served in healthcares by its ability to identify and measure of drugs and patients’ data. In the pharmaceutical industry, inauthentication is one of the vital problem. By the Health Research Funding Organizations report, it was disclosed that $10\%$ to $30\%$ of the medicines have sold in the economically advanced countries. World Health Organization (WHO) evaluated that $16\%$ of drugs have improper mixture of substances and $17\%$ of drugs are comprised inexact level of necessary ingredients. So, these kind of medicines can be harmful to patients and will not cure them from diseases, rather they cause fatality. All these major issues can be directed towards workable solutions by blockchain technology because all the data and time of the agreements are recorded to the developed register, which can observe data and make information unchangeable. Blockchain can grow a better system to maintain the previous and present data of patients in a secure manner.

• 2.

  The implementation of blockchain technology is associated with the energy industry through the major uses of microgrids. Blockchain can make easier to record and verify the power selling and buying deal in the small power supply grid containing a distributed generation of electricity. Moreover, energy related applications of blockchain have the capability to lower energy costs and to resume functionality [52].

• 3.

  Through demoralization and automatic technology, blockchain could be most appropriate in the stock exchange. Blockchain can lower costs by demolishing source of intermediaries and accelerating transaction agreements. Furthermore, blockchain technology can offer possible uses in clearing and settlement of a transaction by easing repetitions of paper works in the business. Blockchain technology overcomes the need of third-party regulators in transactions.

• 4.

  Blockchain can solve the distrustfulness in the segment of voting by providing a developed record to assure the countability of voters.

• 5.

  Blockchain technology can reduce administration costs in the insurance sector by the arrangement, purchase, submission, claims, and record of insurance policies between clients, policy holders, and different companies. Furthermore, blockchain empowers the insurance industry to globally increase the productivity.

• 6.

  Blockchain technology can be an effective system for protecting the online identification. This technology can assure to save an individual’s identity information from being theft and dishonest activities, by generating a updated technical platform.

• 7.

  The updated implementation of blockchain technology can able to speed up and secure the procedure of documentation in the trade finance in an assuring manner.

In blockchain, the digital signature which is achieved using cryptography of private key is an important part to initiate a transaction. In blockchain networks, transactions are used to transfer digital assets within peers. These peers are responsible for choosing and validating the transactions. After the proper verification and validation procedure, one transaction is inserted in a block that is created by the miner node. The miner node has the ability to create a new block only when it can solve first a computational puzzle. After creating a new block in the network, it must be verified by all other peers and after that, it will be joined in existing chain. At this moment, transactions are specified as confirmed [53].

Two types of transactions are commonly used in blockchain: (i) Bitcoin transaction and (ii) Ethereum transaction. Bitcoin is a confidential, authentic, and integrated digital currency, which is performed over an accessible, public, and unidentified blockchain networks. In case of Ethereum transactions, it has sufficient fields to transfer ether (transferable amount) and messages which are beneficial to properly trigger the contracts. Blockchain is the basic technology of a number of digital currency based on a cryptographic system. Chain of blocks can save information with digital indication in distributed and allocated networks. Dissolution, constancy, clarity, and the power of examining ability of blockchain characteristics build the transactions more protected and strongly unchangeable. Blockchain technology can be applied in various segments, such as healthcare, energy industry, stock market, voting, insurance, identity management, trade finance, social services, etc., to make feasible solutions in a secure manner [54]. Although blockchain-based remedies become popular for different business sectors, there are many implementations of blockchain, which need more analysis and improvement. Moreover, functions of blockchain need to remove the errors fast to develop more fruitful, effective, protected, and appealing form of blockchain technology in the long run.

In the modern world, artificial intelligence is one of the prominent and rising technology, which can ease the tasks of human beings to complete the tasks efficiently and effortlessly. A person can use his intellect and creative mind through the performances of artificial intelligence which is the property of machines, computer programs, and systems. Artificial intelligence is often useful to find ways in problem solving and decision making independently. Learning, thinking, planning, and ability of taking actions independently are the fruitful outcomes of recent researches in AI. Some researches make conclusions that AI is wittier than human beings. On the contrary, some of the other researches criticize its functional ability and intelligence level in comparison with human brains. But large number of researches are involved to explore the possibilities of multiple capacities of AI in future. In recent times, AI is becoming one of the highest place of interests among the community of researchers [55]. Many successful researches in AI already revealed the various ways to apply it in solving problems easily associated with software engineering processes. By observing the rapid developments of AI in yesteryears, we all are hopeful for the positive revolutionary outcomes of researches to take digitalization on the top of the success history.

Artificial intelligence is introduced in various ways by many authors through multiple subjects and their scientific implementations in different times. ‘Thinking ability’ is the common element among all the fields of AI. The term ‘intelligence’ is actually the combination of the way of thinking, acquiring knowledge, and ability to implement in order to solve a given problem, as per requirements. AI achieves the capabilities and limitations of human beings in many tasks that were tough to be unable to accomplish in earlier times. Advancements in AI were achieved due to the increase of valid information, betterment in hardware, and the progress in the designs and operations of systems, which were planned precisely in ordered steps for the computational procedure. Nowadays, developers and researchers are allowed to code and test models fast by high quality open-source libraries which are the fruitful outcomes of advancements of AI [56]. Although some applications based on AI disappointed for their trust-related problems, developments in speech recognition, object detection, and many more have successfully led to their enlargement and filtration to the real-world applications, mostly in supervised learning by the rising performance of AI. Although the progress in the development of AI is increasing, there are lot of challenges that need to be overcomed to control the improvement in the applications of AI. Lot of questions and doubts have been arised repeatedly about the implementations and fruitful outcomes of AI because of distrustful results and unreliability in multiple segments like vision models, visual recognition, Natural Language Processing (NLP), etc. To protect future society, these hazards should be addressed fast [57]. Although recent researches of deep learning have been performed noticeably, there are lot of darknesses in the area of improvements. There are various methods in AI, which are used by their applications in software engineering associated with software development processes, for solving problems. Although lot of researches already drew the conclusions that AI is notably difficult course to experiment through teaching, learning, and execution, the current improvements of AI can make the incompetent functions of many software products less severe.

Applications of AI are:

• 1.

  GPS navigations

• 2.

  Customer services, finance, sales, and marketing

• 3.

  Military applications

• 4.

  Medicinal applications

• 5.

  Space applications

• 6.

  Industrial applications

• 7.

  Telecommunication applications

An example of working principle of AI is shown in Fig. 4. Lot of researches revealed that AI can achieve remarkable discoveries in future [58]. Continuous improvements of AI will provide multiple options to make human life more easier. Although AI could not be $100\%$ reliable for human beings because of its executional errors in several times, AI is continuing its development consistently to provide accurate outcomes of its performances.

The IoT system along with blockchain is used in [59] to reduce accidents and improve traffic conditions by using decentralized peer-to-peer communication instead of cloud-based centralized system, so that data can be transmitted efficiently. Confidentiality, integrity, and availability of data are ensured by using blockchain which also makes external attacks negligible since users communicate and self-execute automatically. However, there are some vulnerabilities including power consumption and excessive overload of the network throughput, which are not taken into consideration in [59]. In [60], it is observed that decentralized ledger technology is based on various cryptocurrencies such as bitcoin and Ethereum. Cryptocurrency provides a reliable, stable, and robust way to store, transmit, and validate data to ensure democracy, transparency, and safety in various organizations. With the development of the IoT 4.0, the impact of cryptocurrency would be enormous, and it will be applied in various fields like banking, healthcare, financial, etc. Using blockchain as the underlying communication architecture, a security auditing system is constructed in [61], which can predict and detect any anomalies to guarantee confidentiality of data, issue alerts for security incidents, collect evidences for auditing purposes, and probe event trajectory via data analysis in order to track the involved devices. It uses decentralized system to prevent tampering of data and help in encryption and decryption of data records. Using smart contract scheme on a single blockchain node, the effectiveness and efficiency of the security records are examined.

Two Local Energy Market (LEM) mechanisms are proposed in [62] with artificial prosumers and these mechanisms are implemented on private blockchain in a decentralized manner based on economic evaluation. It promotes increase in the production of renewable energy and decrease in consumption to reduce usage spikes at peak-hour. Through experiments, it is found that prosumers energy behavior can affect their economic benefits; and if users’ energy supply is increased and demand is decreased during peak-hour, their economic benefits increase, which helps in boosting the local economy. The decentralized WEB revolutionized the technologies such as blockchain by introducing symbiotic webs as discussed in [38]. It helps in linking web pages with AI, NLP etc., and makes it transparent and efficient. Due to several characteristics of the IoT such as mobility, distributed deployment, and limited performance, centralized access control method does not support access control in a large scale environment. So, based on attributed-based access control (ABAC) and Hyperledger Fabric Blockchain framework, an access control system is proposed in [39] making use of blockchain’s decentralization, traceability as well as tamper-proof property. The proposed system has a device authority model which ensures the security of the device using three smart contracts, and helps in maintaining high throughput along with ensuring data consistency. In a network, to address the mining process which has limited resources, a blockchain-based solution is presented in [40], which verifies a transaction. To overcome the problem of Proof of Work (PoW) which arises due to limited resources as well as its expensiveness, a decentralized blockchain system is used, where the nodes of the IoT devices are grouped together as a cluster in the network, and multiple transactions occurring at the same time are verified making it more efficient. First Fit Decreasing (FFD) method is modified so that node which leaves the cluster can be replaced with proper fit and helps in overcoming the limitations of PoW.

A transaction settlement system, NormaChain, is proposed in [63], for IoT-based e-commerce based on blockchain in order to provide security against collusion and cryptoanalysis, and ensures privacy. NormaChain overcomes various problems such as huge computational overhead and non-supervisability. It increases system scalability and transaction efficiency. A decentralized public key searchable encryption scheme (Decentralized PEKS scheme) for cryptosystems is designed, which helps in achieving crime traceability and illegal transactions. A Decentralized IoT Collectability Data Marketplace (DCDM) market model is presented in [64], where various operational factors are involved such as contextual, operationality, geographical location, and data provider availability. It uses a combination of crowdsensing and blockchain. It employs crowdsensing to build a business model for trading the IoT and collect data on-demand from customers, whereas blockchain is used to ensure the privacy and transparency. Using community-based decentralized architecture, security is maintained and there is a reward mechanism to boost the performance.

A decentralized application (DApp), as presented in [65], is used for sharing IoT data based on blockchain technology using smart contracts without the involvement of a third party. It is accessible to everyone. It applies Sensing-as-a-Service (SaaS) which provides opportunity for monetization of data by helping in selling sensor data easily. By using cryptocurrency, DApp ensures security and flexibility to the users transacting on peer-to-peer level. However, the system faces various issues, such as scaling and high delay in transaction confirmation, which are to be addressed. BlendCAC is used in large scale IoT devices for access control processes to provide fine-grained, scalable, and lightweight control mechanisms. In [66], IoT devices control their own resources. Local private blockchain network is used and smart contracts are implemented for registration, revocation of the access authorization, and propagation. A capability delegation mechanism is proposed, which helps in access permission, and a token management strategy is also implemented. It is found that the system helps in validation and access control authorization in a decentralized and trustless IoT network.

With the increased number of IoT devices, there is a need to create an IoT back-end, and to improve the method of communication for IoT gateways, which can be achieved using trustless and decentralized nature of blockchain providing peer-to-peer network, as discussed in [67]. Based on the computing and storage capabilities, all the IoT devices can be integrated, which will lead to data-centric models where data processing and application development are conducted by smart contracts. In order to create a smart environment, various benefits, drawbacks as well as trending research topics of using the decentralized nature of blockchain along with IoT are explored in [68]. Potential business applications are explored in order to form a smart world. Conventional IoT systems suffered from various issues like lack of transparency, limited scalability, and single point of failure. In [69], a general architecture is presented by combining IoT systems with blockchain technologies to implement the decentralized nature of blockchain for managing IoT data for further utilization like retrieving and auditing. A case study of a resource allocation learning-based method is also proposed for intelligently managing data. Table 1 highlights existing works which consider the integration of the IoT and blockchain concentrating on decentralization.

The application of blockchain in the perspective of safety and transparency is highlighted in [60]. Blockchain helps in digitally transferring money safely to the intended recipient by protecting against various external attacks, losses, and errors. With the development of the IoT 4.0, its impact would be enormous and it will be applied in various fields like banking, healthcare, financial, etc. The WEB revolutionized the technologies such as blockchain by introducing symbiotic webs making it transparent and efficient. The Industry 4.0 encompasses smart contracts, smart supply chain, smart products, and smart factory, as discussed in [38]. Blockchain has various features, such as immutability and transparency, which make it distinguishable. There are various blockchain-based IoT projects but there is no method to compare their performances based on energy consumption. Blockchain technologies provide trust while transacting at the cost of energy instead of manpower. A research to model the energy consumption behavior is done in [70] by collecting real-world data. The model reflects on various algorithms such as PoW, Proof of Stake (PoS), Stellar Consensus Protocol (SCP), and Ripple Protocol Consensus Algorithm (RPCA) based on linear models.

To improve the enactment of the IoT and its application in an uncontrolled, hostile environment, and to automate living conditions ensuring the privacy and security, a blockchain model is proposed in [71]. It provides various advantages to the IoT system, such as security, cost reduction, immutability, anonymity, robustness, creditability, trust building, authentication, verification, etc. However, in the proposed model, there are some challenges such as consensus mechanism and computational cost for verification of transactions. A comprehensive analysis is presented in [72] in the direction of the disruptive innovation occurring in the field of smart contracts and blockchain. A shared distributed ledger ensures secure and tamper-proof transactions between various users in the network without involving third parties. With the application of Directed Acyclic Graph (DAG) and its multi-forked structure, high scalability, efficient provenance, and optimized validation are observed. Together the IoT and blockchain help in real-time monitoring of resources and ensure non-repudiation, privacy, accountability, predictive analysis, transparency, and stakeholder visibility.

Blockchain technologies implemented on the IoT system in [73] increase the efficiency by improving the transparency of data. It is found that cryptocurrency IOTA is not the best solution with respect to quality improvements. It ensures high scalability by storing a large amount of data but due to the usage of central server, it is prune to attacks. A study is done to recommend which platform is to used based on the demand. Converging cloud computing, IoT, blockchain, and data analytics, a prototype of transportation scheme is proposed in [74], where both private and public blockchains are used. Hyperledger fabric saves the massive trip data and Ethereum framework is implemented as an insurance and payment model. Based on a driver’s behavior, insurance premium is accessed and deducted from IoT data, which is collected from mobile sensors. The proposed model helps in encouraging drivers to drive safely. It is evaluated after experiments that resource utilization is correlated with request load. The system is capable of handling high request load and helps in achieving better throughput.

To ensure scalability to constrained IoT devices, blockchain is used as it offers various strengths, such as transparency, auditability, decentralized consensus, security, etc., which make it an ideal component of IoT solutions. The work [75] provides a scalable, generic, and manageable access control system. It also implements proof-of-concept (PoC) which enables various devices to connect to the same network. Due to the versatility of management hub nodes, high flexibility of the network is ensured but the network suffers from overhead of waiting due to the time required to issue control information by the blockchain network. Smart logistics solution in [70] uses logistics planner and smart contracts to secure trust, provide visibility, and help in condition-monitoring of assets using distributed ML framework in supply chain management areas. It demonstrates liability, transparency, accountability, optimal planning, and traceability, for handling assets by different parties involved in the logistics of supply chain management with the help of various technologies such as IoT, blockchain, ML, and Big Data. Table 2 highlights existing works which consider the integration of the IoT and blockchain concentrating on safety and transparency.

Blockchain based communication architecture, NGBN framework, is designed in [76] based on Named Data Networking (NDN) by integrating the IoT and wireless communication network. It provides a scalable, robust, and secure network environment for other applications. Two application scenarios illustrated under NGBN -- cyberspace scenario which focuses on the digital life and society, where the blockchain offers virtual coins so that the data can be transmitted efficiently and economically, which are useful for smart transportation-related application. Using blockchain as the underlying communication architecture, a security auditing system is constructed in [61], which can predict and detect any anomalies to guarantee confidentiality of data. The proposed system issues alerts for security incidents, collects evidences for auditing purposes, and probes event trajectory via data analysis in order to track the involved devices. It uses decentralized system to prevent tampering of data and help in encryption and decryption of data records. By using smart contract scheme on a single blockchain node, the effectiveness and efficiency of the security records are examined. The IoT system presented in [37] is based on blockchain to secure data and support efficient outsourcing services. To ensure confidentiality of data as well as to secure the keys against key-leakage attack, a blockchain-based key aggregation encryption system, known as BAI-KASE, is proposed, so that the encrypted data can be shared and searched securely. After performance analysis, the results indicate that the proposed model is efficient in computational costs as well as communication overheads.

A framework in mobile IoT is designed in [77], which is based on Wireless Sensor Network (WSN) for secure routing, to improve data reliability, network lifetime, and network security against various malicious attacks. It uses shortest routing chains to efficiently use energy with an optimum decision. Two subcomponents are proposed to generate and maintain non-overlapping clusters and to secure end-to-end routing paths based on the blockchain architecture which leads to the decrease in communication costs and overheads under large network and increase security. Fabric-IoT, an access control system based on ABAC and Hyperledger Fabric Blockchain framework, is proposed in [39], which has a device authority model ensuring the security of the device accessed. It helps in maintaining high throughput ensuring data consistency. A blockchain based IoT system is proposed in [78] to provide advance security and privacy to the remote patient monitoring (RPM) system. The network provides reliable data communication and stores the data over cloud securely using advanced cryptographic techniques such as add–rotate–xor (ARX) encryption. Using digital signatures and ring signature, the privacy can be improved. To increase security, double encryption scheme is used, which protects our data from third party intruders.

A blockchain-based IoT system is proposed in [79] using the distributed and immutable nature of blockchain which helps in building trust and making the system robust, secure, and immune to failure. Crypto-token which is pre-generated using prediction models, like Long Short-Term Memory (LSTM), is used to provide continuous security with seamless user authentication. A Global IoT Device Discovery and Integration (GIDDI) solution in [92] consists of a GIDDI Marketplace that provides functionality required for query, integration, payment, IoT device registration, security, and a GIDDI blockchain which manages metadata. The blockchain stores metadata related to IoT devices, which provides scalability, authentication and security. Wireless battery management systems (WBMSs) of [80] provide secure communication with external devices and data. It is used in cyber-physical environment to solve wiring-harness issues faced by battery management systems (BMSs), and utilizes cloud computing and wireless IoT networks efficiently. It incorporates smart contracts, peer-to-peer transaction, demand side management, supply chain, and IoT’s privacy and security. Due to the application of the private blockchain system, less energy is required. It provides a key management and access control, thereby reducing latency significantly compared to other platforms. The latency of the experiment can be further reduced if instead of using cloud-based blockchain server, local blockchain server is used.

FastPay, a payment method for blockchain-based edge-IoT platforms, is proposed in [42]. This method is both secure and fast, and is based on smart contracts mechanism. On front-end IoT devices, digital payments are made while in the back-end, blockchain’s distributed nature ensures the validity of the system. It protects the system from double-spending attacks. A blockchain-based B-IoT system is presented in [81], which utilizes credit-based PoW mechanism. It decreases power consumption while increasing the complexity of computing for malicious nodes. It improves security as well as transaction efficiency. To protect the confidentiality of data, a data authority management system is designed in [87], which regulates the access to data and protects it without affecting the performance of the IoT system. It faces various challenges such as quality control of sensor data as well as storing large amounts of data. Blockchain technology is used to revolutionize IoT system for Machine-to-Machine (M2M) communications [87]. One M2M is a global project which achieves interoperability with respect to the communication system, which can be achieved using blockchain system. Its decentralized and peer-to-peer nature help secure the system from external attacks and it is based on blind voting. The proposed system in [87] is applied in the IoT service layer; and it helps in preserving privacy and it uses smart contracts in various IoT related applications. NormaChain, as proposed in [63], provides security against collusion and cryptoanalysis, and ensures privacy.

A proxy re-encryption scheme is proposed in [93] in combination with the blockchain to tackle trust issues and scalability of the system, and to ensure the confidentiality of data. The atomized system stores the encrypted IoT data in a distributed cloud and helps in securely sharing the data by using dynamic smart contracts without the involvement of a third party. A multi-robot path planning is deployed in [82]. The plan combines Internet of Robotic Things (IoRT) with Hyperledger Fabric. IoRT acts as a bridge between the IoT domain and robotics domain. Hyperledger Fabric is an enterprise-grade and permissioned blockchain platform. Hyperledger Fabric provides an extensible and elastic architecture solution for control system by adapting power-efficient mechanism known as ordering service. It shows potential to enable scalability and security for IoT use cases. The results obtained reveals that while making transactions, a minimum latency is observed compared to other public blockchain platforms. To improve the enactment of the IoT, and to ensure privacy and security, the IoT is integrated with blockchain, as discussed in [71]. A Decentralized IoT Collectability Data Marketplace (DCDM) market model as presented in [64], where security is maintained, and to boost performance, a reward mechanism is used. Interledger is referred to as the class of operations, that is stretched across two or more distributed ledger technologies (DLTs), as discussed in [94]. Group of researchers demonstrated a way in which a resource-constrained IoT device interacts with an ‘‘IoT-friendly’’ ledger by using interledger as a gateway service that mediates the payment and access controls between different parties using private and public permission-less ledgers such as blockchain and Ethereum. Using hashed timelock agreements (HTLAs), it is seen that payment can be made in an auditable and secure manner.

A decentralized application, DApp, is proposed in [65], which ensures security and flexibility to the users transacting on peer-to-peer level using cryptocurrencies. An analysis done by a group of researchers is shown in [72], which ensures the security between various users in the network without involving third parties. An IoT system is proposed in [95], which collects, publishes, sends, and stores all the relevant data for smart sensors. It is free, instantaneous and trustworthy as reliable, and immutable information can be securely stored and can be accessed anytime by various stakeholders. Cloud-based IoT systems introduce constrained interaction in [96] with actuators and sensors as well as network latency. An IoT management construct, which is software based, is presented, and it enables multi-tenancy and load distribution on constrained edge-devices. It evaluates that, by reducing the number of virtual resources used by the host, the delay is increased, and it shows the use of permission-based blockchain. It is believed that the security and safety of the system can be increased by using blockchain for storing data.

Various blockchain platforms such as Ethereum and Hyperledger fabric are compared in [83] based on their consumption of energy for real workloads and by analyzing the performance trade-offs so that it could be securely implemented in IoT devices. It is found that Hyperledger fabric consumes less energy compared to Ethereum. The mechanism, proposed in [97], is used for trading IoT sensors to provide security against malicious attacks and preserve privacy of the parties. A blockchain-based IoT system is proposed in [84], which is applied in healthcare sectors and helps in increasing the security and robustness by utilizing cryptographic algorithm. It helps in providing trustworthy and peer-to-peer platform to manage data efficiently without involving a third party. Blockchain provides a common platform so that different IoT devices can communicate in a distributed manner securely and protects against malicious attacks.

The 5G technology promises reliability, scalability, real-time applications, ubiquitousness, and cost efficiency. It helps in improving IoT devices and makes it more efficient along with the help of blockchain. Communication constraints faced by the blockchain are resolved by the 5G communication in [85], where the security and privacy issues are resolved by using blockchain. It faces some issues such as limited storage capacity of IoT systems, poor scalability, time consumption while mining, etc. It also increases the traffic overhead. Blockchain-assisted privacy-preserving authentication system (BPAS) is proposed in [86], which helps in transmitting messages securely without involving a centralized third-party. Similarly, data sourcing, crowdsourcing, edge computing, and industrial IoT (IIoT) architectures are achieved in overcoming some of the issues faced by the blockchain when it is implemented on IoT devices.

A blockchain technology addresses the problems of IoT security as discussed in [88]. A self-clustering methodology is proposed, which helps in clustering the IoT network into $K$ number of unknown clusters by using Particle Swarm Optimization (PSO) and Genetic Algorithms, that help in optimizing the lifetime of the network. Hyperledger Fabric that helps in authorization, authentication and verification. The local blockchain helps in providing framework to improve safety and ability; and the global blockchain is implemented to explore upper layer communications. Due to the multi-layer approach, the network load, delay, and computational load are reduced. Due to the peer-to-peer nature of the model, communication efficiency is increased along with the integrity and security. A survey is conducted in [98], where the security aspects of blockchain-based IoT system is observed. A blockchain-based IoT solution is presented in [89], which helps in dealing with preservation and collection of digital forensic evidences. A private evidence database is utilized along with a type of permissioned blockchain, which helps to provide security services like non-repudiation, integrity, and authentication. To identify malicious attacks, an intrusion detection tool is used, which helps in enhancing the security and determining the source of attacks. Cyber-Trust Blockchain (CTB) is built on top of Hyperledger Fabric, and it dematerializes the chain-of-custody (CoC) process by preserving and recording the history of handling the evidences.

To ensure scalability for constrained IoT devices, blockchain is used in [75], as it offers various strengths such as transparency, auditability, decentralized consensus, and security. The work provides a scalable, generic, and manageable access control system. It also implements PoC which enables various devices to connect to the same network. High flexibility of the network is ensured but the network suffers from overhead of waiting due to the time required to issue the control information by the blockchain network. On the Software-Defined Networking (SDN)-based IoT network, a lightweight forensic architecture is developed in [90] by using the peer-to-peer decentralized nature of blockchain, in order to increase the security, integrity, and assist Chain of Custody (CoC) for evidence collection effectively. The blockchain-based controller uses Linear Homomorphic Signature (LHS) algorithm to validate users. The results indicate an increase in performance with a minimal delay, overhead, processing time, and response time. There is also an increase in the throughput, security, and accuracy. A blockchain-based security framework is proposed in [91] for the vehicular IoT environment in 5G-VANETs enabling SDN architecture. Accountability of the message is also validated by using the immutability feature of blockchain. Using blockchains, trust management is established such that the messages cannot be tampered. Though the model ensures feasibility, accuracy, and efficiency, there is an overhead introduced due to the encryption and transmission of messages and videos.

In [99], the uses of blockchain in financial security, financial accounting, and cyber security are discussed. The different ways of impacts of blockchain in auditing are also highlighted. Authors in [99] find that blockchain needs to be effectively applied into different directions of cyber security and accounting. The authors in [100] focus on the development of a marketplace framework that can address design challenges in IoT models, and accordingly propose a three-tier framework. The mechanism proposed in [101] provides a suitable smart contract scheme for guaranteeing the integrity of IoT devices, and immutable and dynamic management of malicious applications. In [102], a blockchain ledger is established to make more blockchain nodes from IoT devices. Here, the authors design a mechanism that selects multiple IoT devices to enable the real-time policy decisions in a distributed way.

In [103], a proxy re-encryption mechanism is proposed for secure sharing of data in cloud environments, where an edge node behaves like a proxy server to handle extensive computations. The work [104] designs a generic cloud-based IoT framework to overcome the security issues of provenance data by leveraging blockchain. The work [107] discusses blockchain-related security enhancement solutions by highlighting key points required to develop different blockchain systems. The authors in [108] review the security and privacy related issues in IoT systems, and present some security solutions using blockchain technology. For blockchain platforms, the incentive mechanism is investigated in [105] to purchase more power for computation from server devices in order to take part in the mining. For the access control, BorderChain [106] is based on blockchain in IoT endpoints, where the security protocol is ensured by using verified IoT devices in the communication with IoT gateways. BorderChain also applies access tokens which are used by IoT services and users for any query to IoT resources. Tables 8 and 9 highlight existing works which consider the integration of the IoT and blockchain concentrating on security.

A remote patient monitoring (RPM) system is presented in [78]. Using digital signatures and Ring Signatures, the privacy is achieved to protect data from third party intruders. WBMSs proposed in [80] ensure secure communication with external devices and data. The credit-based PoW mechanism in [81] builds DAG structured blockchain which overcomes various problems such as power intensiveness, privacy and low-throughput, and makes the system more efficient. A scalable and trustworthy mechanism is proposed in [109] which enforces traffic management by integrating blockchain, pseudo IoT applications and SDN at the edge networks, for verification and authentication purposes. It reduces the risk of IoT deployment by interacting various IoT devices, stakeholders, and networks to manage the traffic efficiently and ensure security, reliability, and information tracking by the use of peer-to-peer nature of blockchain. Various IoT services and devices are listed, which are circulated among the stakeholders to prevent traffic and reduce attacks on edge networks. Various lightweight IoT devices face the problem of storing the blockchain due to low storage capacity. A storage compression consensus (SCC) algorithm is designed in [110] which overcomes the problem such that the devices compress the blockchain using this algorithm to increase storage capacity. It is observed that, by using SCC, the storage capacity is reduced by $63\%$ compared to the existing algorithm like Byzantine fault tolerance (BFT) which uses private blockchain. However, the proposed scheme increases the maintenance of the system by acquiring free storage spaces.

M2M communications presented in [87] secure the system from external attacks. It helps in preserving the privacy. Interledger discussed in [94] acts as a gateway service that mediates the payment and access controls between different parties, where the payment can be made in an auditable and secure manner. In [112], various challenges as well as current prospects of blockchain are surveyed. Communication constraints faced in [85] are resolved by the 5G communication using blockchain ensuring the privacy. However, it faces some issues such as limited storage capacity, poor scalability, time consumption, and it increases overhead in traffic. BPAS proposed in [86] helps in transmitting messages securely. Two protocols, BRICS-BB and BRICS-SSP, are introduced in [111] for resource-constrained IoT devices using Multi-Party Computation (MPC) and blockchain. These protocols are applied in variety of cases. For example, BRICS-BB is applied on software updates and BRICS-SSP can be used for data privacy management, supply chain, and identity management. A survey of blockchain of things (BCoT) is conducted in [113] which provides an overview of both the blockchain technology and Internet of Things as well as the challenges addressed by the blockchain technology in IoT systems. Industrial applications of BCoT and issues faced by using blockchain technology for fifth generation in IoT are also highlighted. Table 10 highlights existing works which consider the integration of the IoT and blockchain concentrating on maintaining the privacy.

An IoT device residing on blockchain uses smart contracts to provide authentication, integrity, and non-repudiation. Due to the problem faced by IoT devices using centralized platform, a decentralized peer-to-peer based blockchain network is implemented in [114], which helps in preventing data-tampering by intruders. The GIDDI solution proposed in [92] is IoT-scaled and IoT-owned. In this solution, the blockchain stores metadata related to IoT devices, which provides scalability, authentication, and security. An update framework is proposed in [115], which prevents attacks and ensures authentication and repudiation to the messages sent between the firmware update service and IoT devices. A blockchain-based architecture is proposed in [41] for edge IoT devices so that they can be uniquely identified through registering physically unclonable functions (PUF) attributes at the time of manufacturing. The dual-use of blockchain enhances the usability and reliability of the system, and counteracts the problems of cloning and counterfeiting. A low-cost communication protocol is also introduced to securely authenticate the local devices.

A trust list is proposed in [109], which enforces traffic management for verification and authentication purposes. A blockchain-based IoT system is proposed in [43], which is used to validate blockchain data helping IoT devices with random nodes to confirm block headers without only depending on the central server. But the large size of the blockchain causes problem for IoT devices with low data storage capacity, memory, and bandwidth. The proposed system works on the top of blockchain protocols supporting IoT devices with low-power. It consists of a gateway which transfers messages from blockchain to IoT devices. It uses a probabilistic approach which is based on Bloom filters. Due to the high cost of deployment, IoT-based solutions are restricted but it can be overcomed by horizontal integration of IoT devices and the use of blockchain system. In this direction, a ticket-based verification protocol is implemented in [116], which does not require the participation of IoT devices in complex transactions, and therefore it can be used in low-end IoT devices. It reduces the networking and processing overhead of IoT devices in the system.

A blockchain framework, ‘‘Sensor-Chain’’, is designed for IoT of mobile devices in [117]. The proposed framework is lightweight and scalable so that the IoT’s limited storage capabilities as well as computational power can be addressed. The designed framework reduces resource consumption and is capable in retaining information about IoT systems of mobile devices. Authentication is concentrated in this work. BPAS is proposed in [86], which helps in transmitting messages securely without involving a centralized third-party. The user authentication scheme proposed in [118] enables the integration of blockchain and fog nodes. The proposed scheme uses smart contracts to authenticate users such that IoT devices can be accessed without involving a third party in a decentralised network. Fog nodes provide scalability by carrying out computation related tasks to communicate with blockchain and authenticate devices. The security analysis is also conducted, which shows that the proposed solution achieves security goals of integrity, confidentiality, and availability. The proposed mechanism provides authentication, data encryption, and security against malicious attacks.

A trusted and distributed system is proposed in [119], which is based on edge computing and blockchain in order to improve the authentication and efficiency. The Practical Byzantine fault tolerance (PBFT) consensus algorithm is implemented in an optimized manner to construct a consortium blockchain which is used for storing authenticated data and logs. It achieves traceability of terminals and guarantees trusted authentication. Edge computing is applied to provide edge authentication and name resolution using smart contracts. Asymmetric cryptography is used to prevent the nodes from being attacked. Caching strategy is proposed which outperforms edge computing in terms of hit ratio and average delay. Using blockchain technology, we can digitally identify the IoT devices, and then authenticate these devices before it join an IoT network. Authenticated Devices Configuration Protocol (ADCP) helps in achieving the authentication process as presented in [120]. The proposed software system consists of ADCP handler, software firewall, WebSocket server, and blockchain database. WebSocket server helps in synchronizing blockchain database. ADCP responds to the requests for connection from other authenticated devices. The software firewall helps in encryption of data automatically. A consensus algorithm is proposed in [121] that reduces the latency in block validation and energy consumption. The proposed mechanism is evaluated over resource-constrained IoT devices. Table 11 highlights existing works which consider the integration of the IoT and blockchain concentrating on maintaining the authentication.

A security auditing system is constructed in [61] which prevents tampering of data and helps in encryption and decryption of data records. Fabric-IoT proposed in [39] makes uses of blockchain’s decentralization, traceability as well as tamper-proof property. The proposed system has a device authority model which ensures the security of the device. Due to the problem faced by IoT devices in using centralized platform, a decentralized peer-to-peer based blockchain network is implemented in [114], which helps in preventing data-tampering by intruders. To update firmware and software, instead of delta updates, a private blockchain mechanism is proposed in [115], which enhances the update performance and helps in integrity verification, by utilizing a tamper-proof server. The blockchain server maintains a record of metadata, checksum, etc.

A shared distributed ledger ensures secure and tamper-proof transactions between various users in the network without involving third parties. With the application of DAG and its multi-forked structure, the high scalability, efficient provenance, and optimized validation are observed in [72]. It affects Key Performance Indicators (KPIs) of the logistics domain. Together the IoT and blockchain help in real-time monitoring of resources, and ensure tamper-proof and efficiency. To address the constrained nature of the IoT, both blockchain and edge computing are used. It decreases device requirements for memory, and increases the performance. The purpose of information circularity is fulfilled in a smart city as data can be extracted to reduce the waste production and resource consumption, and increase the services, as discussed in [122]. In [91], by the use of blockchain, a trust management is established such that the messages cannot be tampered. Although the model ensures feasibility, accuracy, and efficiency, there is an overhead introduced due to the encryption and transmission of messages and videos. A blockchain-based IoT system, known as Beekeeper, is proposed in [123], that allows servers to process data without learning from homomorphic computations on the encrypted data. The data can be collectively verified even by public. Since all the data are recorded in the decentralized blockchain, computational resources and huge memory are not required, and therefore the proposed mechanism ensures privacy, security, and the resistance to tamper. Table 12 highlights existing works which consider the integration of the IoT and blockchain concentrating on tamper-proof and efficiency.

The survey [124] presents the classification of resource-constrained IoT devices and associated wireless communication protocols. The authors in [125] discuss the cloud computing integrated with the blockchain platform for offloading computations from IIoT networks. Moreover, the resource management issues are studied between cloud service providers and miners in blockchain. The work [126] discusses the importance of the blockchain in 6G communications and resource management for multiple applications in the IoT. A distributed matching mechanism is proposed in [127] to maximize the utilization of the resource-restricted fog computing architecture, that guarantees different mining requirements of fog nodes. Table 13 highlights existing works which consider the integration of the IoT and blockchain concentrating on resource management.

Using HTLAs in [94], it is observed that payment can be made in an auditable and secure manner. A decentralised blockchain technology is used along with constrained IoT devices as it offers various strengths, such as transparency, auditability, etc., which make blockchain an ideal component of IoT solutions [75]. In this work, a scalable, generic, and manageable access control system is proposed, which also implements PoC. NGBN framework is designed in [76], which provides a scalable, robust, and secure network environment for other applications, and it ensures trustworthy and robustness. It focuses on the digital life and society, where the blockchain offers virtual coins such that the data can be transmitted efficiently and economically. A blockchain-based IoT system is proposed in [79], which helps in building trust and makes the system robust, secure, and immune to failure. In [41], by accessing the global dataset, the registered device can be verified anywhere without actually locating the manufacturer. The local dataset helps in authenticating local edge IoT devices robustly.

A blockchain-based IoT system is proposed in [128], which helps civil-military cooperation to store metadata collected from IoT devices and sensors, using Hyperledger Fabric. It ensures feasibility, trustworthiness, and reliability of the collected data from IoT devices. A high-level architecture is presented in [61], which helps in metadata binding using blockchain. A decision matrix or an analysis framework is formulated in [129], which helps in identifying key points while integrating the IoT and blockchain. It guarantees trustworthiness, data immutability, and traceability and provides an audit trail to different stakeholders. A blockchain-based IoT system helps in providing trustworthy and peer-to-peer platform to manage data efficiently without the involvement of a third party, as discussed in [84]. It provides a common platform such that different IoT devices can communicate in a distributed manner securely and protect against malicious attacks. It ensures feasibility, fault tolerance, scalability, and operation time.

A model is proposed in [130] to provide IoT devices the trust while making the payment by using blockchain and OAuth 2.0 authorization framework. It ensures the confidentiality and integrity of IoT data. Authorization requests are granted using the smart contract which works on permissioned blockchain. Hash-lock and time-lock mechanism are used to cryptographically link various trusted IoT resources for authorization grants and blockchain payments. In [131], a systematic survey is conducted on the integration of the IoT and blockchain, where the current trends are studied on the usage of blockchain in IoT systems. Challenges faced by IoT and blockchain are also highlighted and analyzed, directing to future works in the integration of the IoT and blockchain.

A supervised learning algorithm, Random Forest, is applied in [146] to correctly identify types of IoT devices from the white list. To train the multi-class classifier, the data was manually labelled and collected from the traffic data. The results show that $99\%$ of the devices were correctly identified from white list. The proposed mechanism also demonstrates good transportability and it is highly resilient in respect to external attacks. In [148], a systematic review of the attack vectors, security requirements, and the security solutions currently implemented for the constrained IoT networks is presented. Here, ML and DL are used in the IoT devices to increase the security and privacy of the system. There are various limitations of using ML and DL, such as the model does not fit in all problems, the convergence takes long time, a minute change in the input creates havoc in the system, and the models are also vulnerable to security breaches. Due to the increase in the poisoning attack as described in [149], the overall performance of the system decreases and presents various security challenges. Using the tamper-free framework and contextual information, the poisonous data is detected and filtered to train the supervised learning model. Two variations for fully untrusted and partially trusted data sets are compared based on the performance. It is found that, by using the proposed method, the network performance is increased and the security is ensured.

To improve the security of IoT devices, a machine learning algorithm is used in [150] within the IoT gateway. By using artificial neural networks, we can detect anomalies in the data intelligently. The results presented in [150] correctly identify invalid data points in IoT systems. By using ML, Denial of Service (DoS) detection is performed in [151], where it is observed that DoS traffic attack can be identified efficiently from customer IoT devices. To restrict the computational overhead, limited feature set is used, which is necessary for the middlebox deployment and real-time classification. Five different classifiers, i.e., KNN, Support vector machine with linear kernel (LSVM), Decision tree using Gini impurity scores (DT), Random Forest using Gini impurity scores (RF), and Neural Network (NN) are tested on both normal and DoS attack traffic datasets. The results show an accuracy of higher than $0.99$ on all the aforementioned algorithms. An anomaly detection framework is proposed in [45] for detecting malicious behaviours and cyber-attacks in IoT devices having limited communication and computation resources. The proposed mechanism can be implemented in any device despite the size and is more suitable for devices running specific type of applications. ML models and statistical techniques can be used in detecting any type of attacks on IoT devices, and the behavioural models of these IoT devices can be learned easily. It is seen that the proposed approach performs better in comparison to other traditional models.

Various attack models for IoT devices are investigated in [46] along with IoT security solutions based on ML techniques which are learning-based including access control, secure offloading, IoT authentication, and malware detection to protect data privacy. The use of the transfer learning at the beginning of the learning process reduces the risk of choosing bad security defence policies. New ML techniques having low communication and computational overhead enhance the security of IoT systems. A backup security solution should be incorporated and designed in order to provide reliability and security to the system. A three-tier architecture is proposed in [152], which helps in storing and processing a large volume of wearable sensor data. Tier-1 helps in extracting data from IoT devices. Tier-2 stores a large volume of data in cloud. Tier-3 develops a logistic regression-based model for heart diseases.

In order to ensure the flexibility of data access and data security in the IoT, the scheme proposed in [155] addresses the issues with the security and privacy in the edge intelligence (EI) model for the IoT, where ML models are trained in EI. The work [156] presents a survey of various classes of cyber-attacks in different IoT systems. Moreover, this work presents reinforcement and deep learning based solutions to combat different types of security attacks in several IoT models. The authors in [157] design a two-way trustworthy communication model in AI-enabled IoT frameworks. The model incorporates both rating and trust information more thoroughly, which can alleviate the sparse trust problem in the IoT. In [158], the security of IoT devices is achieved by detecting spam with the help of ML, where five ML-based schemes are evaluated using different metrics considering a large set of input features. Each model calculates a spam score considering the input features, and this score reflects the trustworthiness of devices under various parameters. Table 21 highlights existing works which consider the integration of the IoT and AI concentrating on the security.

Various attack models for IoT devices are presented in [46], which address the IoT authentication and malware detection. The use of transfer learning at the beginning of learning process reduces the risk of choosing bad security defense policies. An approach is described in [153], which helps in protecting the privacy of users whenever a resource-constrained IoT device runs ML applications based on NN using cloud computing. To improve the privacy, partial processing is done at the client side such that it would be hard to retrieve the original data again. It is observed that the proposed approach has a shorter prediction time compared to other local predictions. However, there is a chance in increasing the time required for the computation, which affects the performance of the system.

To maintain user privacy in the IoT platform, deep learning as well as edge computing are used in [47]. Due to the multiple layers in deep learning, the accurate information can be extracted from the large amount of raw sensor data generated from the IoT devices. Edge computing helps in reducing the traffic of the network to cloud servers. From the performance analysis, it is observed that the proposed solution can rise the number of tasks in edge servers along with achieving QoS requirements. In [154], using person’s heart beat rate, the authors have tried to predict whether or not a person is in stress, by exploiting both the IoT and ML. The IoT informs patients about their stress conditions and ML algorithms are used to predict the condition of the patient. Table 22 highlights existing works which consider the integration of the IoT and AI concentrating on the privacy.

The work [159] proposes a methodology to dynamically estimate the link quality in cloud communications, considering locations of nodes. For this purpose, several machine learning techniques are implemented, such as decision tree, linear regression, neural networks, random forest, etc. Authors in [160] show that fault detection in data driven predictive analysis in remote locations can be achieved using the support vector machine algorithm and low bandwidth IoT. The scheme proposed in [161] utilizes the self-organization based map neural network to group IoT devices considering their data rates. For each of the group, a slot is allocated for channel access.

A learning-based mechanism is proposed in [162] to analyze sentiment of text in IoT frameworks, and for this purpose, a self-attention prototype is used to predict sentiment features. The work [163] discusses sentiment analysis of big data considering its features, definition, and decision value. Here, the challenges of the impact of the IoT on sentiment analysis of big data are also highlighted.

Different challenges of applying blockchain in IoT are discussed as given in the following.

To develop an efficient intelligent IoT system, we need to identify the research challenges in the direction of integrating AI and the IoT together. Therefore, different challenges in applying AI in the field of the IoT are presented next.