JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 1 Blockchain Systems, Technologies and Applications: A Methodology Perspective Bin Cao, Member, IEEE, Zixin Wang, Long Zhang, Daquan Feng, Mugen Peng, Fellow, IEEE, and Lei Zhang, Senior Member, IEEE

Abstract—In the past decade, blockchain has shown a promis- Gartner forecasts that by 2030, blockchain will generate an ing vision greatly to build the trust without any powerful third annual business value of more than US $3 trillion, and party in a secure, decentralized and salable manner. However, envisions that 10% to 20% of global economic infrastructure due to the wide application and future development from cryptocurrency to Internet of Things, blockchain is an extremely will be running on blockchain-based systems [12]. complex system enabling integration with mathematics, finance, Fundamentally, blockchain is a decentralized ledge man- computer science, communication and network engineering, etc. agement system for recording and validating transaction. It As a result, it is a challenge for engineer, expert and researcher to allows two parties to complete a transaction in a peer-to-peer fully understand the blockchain process in a systematic view from (P2P) network [13]. Without involvement of an authority or top to down. First, this article introduces how blockchain works, the research activity and challenge, and illustrates the roadmap third party, all peer nodes work together to maintain public involving the classic methodology with typical blockchain use ledge with aim of realizing trust, security, transparency and cases and topics. Second, in blockchain system, how to adopt immutability. The recorded transaction in blockcahin can be stochastic process, game theory, optimization, machine learning any form of data which involves the ownership transfer or and cryptography to study blockchain running process and sharing of resource, where it can be tangible such as money, design blockchain protocol/algorithm are discussed in details. Moreover, the advantage and limitation using these methods houses, cars, land, or intangible copyright, digital documents, are also summarized as the guide of future work to further and intellectual property, etc. considered. Finally, some remaining problems from technical, Essentially, blockchain is built on a physical network that commercial and political views are discussed as the open issues. relies on the communications, computing and caching, which The main findings of this article will provide an overview serves the basis of blockchain functions such as incentive in a methodology perspective to study theoretical model for blockchain fundamentals understanding, design network service mechanism or consensus. As such, blockchain systems can be for blockchain-based mechanisms and algorithms, as well as depicted as a two-tier architecture: an infrastructure layer and apply blockchain for Internet of Things, etc. a blockchain layer. The infrastructure layer is the underlying Index Terms—Blockchain, stochastic process, game theory, entity and responsible for maintaining P2P network, building optimization, machine learning, cryptography, network connection through wired/wireless communication, computing and storing data. The top is the blockchain layer that can real- ize trust and security functions based on underlying informa- I.INTRODUCTION tion exchanging. More specifically, blockchain features several RIGINALLY proposed as the backbone technology of key components which are summarized as: transaction, block O Bitcoin [1], Ethereum [2], and many other digital curren- and chain of blocks. Transaction contains the information cies [3], blockchain has become a revolutionary decentralized requested by the client and need be recorded by public ledge; data management framework that establishes consensuses and block securely records an amount of transactions or other arXiv:2105.03572v1 [cs.DC] 8 May 2021 agreements in a trust-less and distributed environment [4]. useful information; using consensus mechanism, blocks are In addition to the soaring in the finance sector, blockchain linked orderly to constitute a chain of blocks, which indicates has been attracted much attention from many other major logical relation among the blocks to construct blockchain. As industrial sectors ranging from supply chain [5], transportation a core function of the blockchain, the consensus mechanism [6], entertainment [7], retail [8], healthcare [9], information works in the blockchain layer ensures a clear sequence of management [10] to financial services [11], etc. As such, transactions and ensures the integrity and consistency of the blockchain across geographically distributed nodes [14]. State Bin Cao, Zixin Wang, and Mugen Peng are with State Key Laboratory of a blockchain is updated when a valid transaction is recorded of Networking and Switching Technology, Beijing University of Posts and on chain, and smart contracts1 can be used to automatically Telecommunications, Beijing, 100876, China. E-mail: [email protected], [email protected], [email protected]. trigger transactions under certain conditions [16]. Therefore, Long Zhang is with the National Key Laboratory of Science and Technology due to its autonomy and efficiency, smart contracts are being on Communications, University of Electronic Science and Technology of used for a wide range of purposes, from self-managed iden- China, Chengdu 611731, China. E-mail: [email protected] Daquan Feng is with the Guangdong Province Engineering Laboratory for tities on public blockchains to allowing automated business Digital Creative Technology and Guangdong Key Laboratory of Intelligent collaboration on blockchains. Information Processing, Shenzhen University, Shenzhen 518060, China. E- mail: [email protected] 1The smart contract is a computer program designed to digitally facilitate Lei Zhang is with the James Watt School of Engineering, University of direct negotiations or contract terms between users when certain conditions Glasgow, Glasgow G12 8QQ, U.K. E-mail: [email protected] are met [15]. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 2

Driven by the continuous development of 5G technology, on the blockchain, with the advantages of tamper resistance more and more services have been launched to improve net- and no single point of failure. It can be seen from the Fig. 1 work performance and user experience. Importantly, features that each blockchain node in the blockchain system undertakes such as data immutability and transparency are the key factors all or part of functions, such as communication between to ensure the successful launch of new services such as IoT nodes, maintenance of P2P network, and computation of data collection, driverless cars, drones, and federal learning. consensus mechanism. Thus, the underlying communication, Blockchain is regarded as the most promising to meet these network and computation is crucial to establish effective and new requirements with its decentralization, openness, tamper secure blockchain system. This encourages us to study how resistance, anonymity and traceability. Therefore, in order to communication, networking and computing affect blockchain more thoroughly explore the potential of blockchain and make systems. Fortunately, some classic methodologies can provide it better serve the requirements of modern networks, it is good ideas, such as stochastic process for block generation necessary to comprehensively and systematically understand and nodes communication, machine learning for P2P network blockchain from top to bottom. Methodology advocates going performance improving, and optimization theory for resource to the bottom of the problem, digging into the essence behind allocation. Therefore, it is feasible and valuable to explore the the phenomenon, and forming a theoretical system with a interaction process of communication, network and computing certain depth. The inherent contradictions of the problem can and their impact on the blockchain system from the perspective be revealed and the fundamental solution can be found from of methodology, and even can provides help for revealing the methodology perspective. Therefore, methodology can be well essential problems in the operation process of the blockchain suited to the research of blockchain system performance to re- system. veal the principles and problems of blockchain running process and blockchain protocol/algorithm design in the blockchain B. Existing Surveys system, and provide theoretical support for solving specific problems. Consequently, this paper outlines the theoretical Recognising the wide applications of blockchain technol- model research of blockchain basic knowledge, the design ogy, a novel survey paper can help researchers in various of network services based on blockchain mechanisms and fields to build good foundations on the subject to guide algorithms, and the deployment of blockchain-based applica- actual developments. Recently, several work have reviewed the tions in practical systems from a methodological perspective, advanced development of blockchain from various views. as well as further outlines the application of methodology For security and privacy, T. Salman et al. in [13] present in blockchain from multiple dimensions such as advantages, blockchain-based security services in authentication, confiden- limitations, case studies and challenges. It aims to provide tiality, privacy and access control, etc. N. Waheed et al. in a comprehensive and clear overview for researchers in the [17] summarize the research efforts of using machine learning blockchain field. algorithm and blockchain technology to address security and privacy problems in the field of Internet of Things in the past few years. M. Conti et al. in [18] focus on the security A. How Does Blockchain Work and privacy threats of Bitcoin, and discusses the feasibility The infrastructure layer and blockchain layer are interrelated and limitations of potential solutions. M. Saad et al. in [19] and interact on each other cooperatively, while the detailed focus on how attacks effect public blockchain and discusses procedures can vary among different blockchain systems. the relationships between a sequence of possible attacks. However, for all blockchain systems, they have to follow the Besides, as the core of blockchain, consensus determines the following basic steps, as shown in Fig. 1. Firstly, blockchain performance and security of the blockchain in many ways, M. clients generate transactions and broadcast them to the P2P S. Ferdous et al. in [20] utilize comprehensive taxonomy of network in Step I. More than one nodes may bundle different properties to analyze a wide range of consensus algorithms, subset of unverified transactions into their candidate blocks in and examines in detail the meaning of the different problems Step II. Afterwards, all nodes perform incentive and consensus that are still prevalent in the consensus algorithm. W. Wang et mechanisms to find the winner whose candidate block would al. in [21] review the state of the art consensus protocols and be announced as a new block in Step III. Then, the new game theory in mining strategy management. block will be inserted into all node’s local ledgers in Step IV. In the context of artificial intelligence (AI), Y. Liu et al. This mechanism connects multiple blocks together and builds in [22] discuss feasible solutions integrating blockchain and a chronological chain. In particular, the process of hashing machine learning for communications and networking system. a new block always contains metadata about the hash value Y. Xiao et al. in [23] introduce the classic theory of fault of the previous block, which makes the linked data highly tolerance and analyzes blockchain consensus protocols using non-modifiable. At last, after the transaction is stored in the a five-component framework. blockchain, the client can request the consensus network to For the scalability of the blockchain, J. Xie et al. in [24] confirm whether a transaction is in the blockchain. study the scalability of the blockchain system, analyzes the Through this workflow, the transaction is finally agreed scalability from the perspective of throughput, storage and by the majority of nodes and recorded in blockchain, where network, and introduces the existing enabling technology of malicious nodes cannot subvert the consensus results. This the scalable blockchain system. H. T. M. Gamage et al. in decentralized architecture ensures robust and safe operations [25] discuss issues of the existing blockchains such as 51% JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 3

Candi-Block i Candi-Block m Block Header Block Header Pre Timetamp Pre Hash Timetamp n Step Ⅰ Step Ⅱ Hash n ... Step Ⅲ Merkle Nonce Merkle Nonce ＄

Transaction Data Candi-Block n Transaction Data Block Header Pre Timetamp Hash n Winer Merkle Nonce

Transaction Data Consensus nodes Blockchain network client P2P Network Block Block verification and reaching Transaction publishing Transaction broadcasting Candidate block building consensus

Step Ⅳ Step Ⅴ Block 0 Block n-2 Block n-1 Block n reply Block Header Block Header Block Header Block Header Pre Hash Timetamp Pre Hash Timetamp Timetamp Pre Hash Timetamp 0 ... n-2 Pre Hash n-1 n Merkle Nonce Merkle Nonce Merkle Nonce Merkle Nonce

Transaction Data Transaction Data Transaction Data Transaction Data

Blockchain Blockchain updating and local replica updating

Fig. 1. An overview of blockchain workflow attack, nothing-at-stake problem together with improvements provides comprehensive reviews to elaborate the impact of for the scalability issues in current blockchains. R. Belchior communication networks and computing from methodology et al. in [26] study cryptocurrency-directed interoperability perspective. However, as a system science, blockchain requires approaches, blockchain Engines and blockchain connectors, researchers with multidisciplinary knowledge background in providing a holistic overview of blockchain interoperability. communication, network, computing and cryptography disci- The integration of blockchain and 5G has become a main- plines to understand the operating environment and principles stream trend, D. C. Nguyen et al. in [27] provide the latest of blockchain from different dimensions, so as to make rational survey on the integration of blockchain with 5G networks use of blockchain technology to improve innovative service and other networks. It gives an extensive discussion about applications in 5G/B5G, IoT and other fields. the potential of blockchain for enabling key technologies of This motivates us to review the state of the art of blockchain, 5G, and further explores and analyzes the opportunities that to reveal the reason why a blockchain system needs com- blockchain may give important 5G services. G. Yu et al. in [28] munication, network and computing and how them interact study the sharding problem in the blockchain, mainly including to each other. Furthermore, to facilitate the deployment of providing detailed comparison and quantitative evaluation of blockchain-based applications, how to carry out mathematical the main sharding mechanisms, as well as analysis of the methods to define a specific blockchain problem is a concern characteristics and limitations of existing solutions. Moreover, that many researchers deserve to pay attention to, but there is by enabling the integration of blockchain and other advanced no survey specifically discussing the use of blockchain from technologies, some of work explore potential applications and the perspective of methodology such as stochastic process, research challenges in IoT [29], [30], smart city [31], cloud cryptography and so on. Motivated by this limitation, different computing [32], edge computing [33], and fog computing [34]. from the previous surveys that focus on blockchain application From the perspective of mathematical tools, Z. Liu et al. such as IoT, smart city and edge computing or blockchain in [35] provide reviews and analyses using game theory in technology such as security and privacy, consensus algorithm detail to deal with a variety of problems regarding security, and resource management, we overview the theoretical model mining management and blockchain applications. However, for blockchain fundamentals understanding, and the design of some classic mathematical methods used in blockchain are network services for blockchain-based mechanisms and algo- not limited to game theory, but also include machine learning, rithms from a methodology perspective involving of stochastic cryptography, and so on. Moreover, a survey which provides a process, game theory, optimization, machine learning, and comprehensive overview on theoretical model, network service cryptography, as well as summarizing the advantages and and management, and application for blockchain system on a limitations of these methods. Moreover, this paper reviews methodology perspective is even more needed. some case studies and challenges applying blockchain in IoT, providing feasible guidance and potential research problems. C. Motivation The organization of this article is as follows. Section II The existing surveys mainly focus on blockchain archi- discusses the research activity, challenges and roadmap of tecture, blockchain protocol algorithm, the integration of blockchain system. Section III-VII analysis the application blockchain with other network technologies, etc. Besides, of stochastic processes, game theory, optimization theory, to the best of our knowledge, there is very few survey machine learning, and cryptography for mining pool manage- JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 4

ment, security strategy, resource management and so on in DAG (Tangle) 3%)7 ),6&2%&26 PoS (Nxt) PoW (Bitcoin) blockchain, respectively. Section VIII discusses main issues Byzantine fault tolerance of blockchain and its improvements. Finally, Section IX con- cludes this paper. 33%

II.RESEARCH ACTIVITY,CHALLENGEAND ROADMAP Resource requirements 7KURXJKSXW

736 100TPS  Recently, lots of work have been done on the application 7TPS of blockchain and the improvement of network system per- formance, involving of the blockchain-based functions design and network optimization. The blockchain-based functions 10mins such as consensus protocol, incentive mechanism and smart 60mins contract have significant impacts on the reliability, efficiency Scalability Confirmation delay and scalability of the blockchain system [36]. However, the design of blockchain-based functions also faces challenges due to storage constraints, computing overhead and delay Fig. 2. Multi-dimensional graph for performance comparison of typical consensus mechanisms in blockchain constraints. Moreover, blockchain is seen as a potential tech- nology to improve network system performance, furthermore, blockchain and the network system especially for wireless net- issues of blockchain technology, etc., with the advantage of work system interact and support each other [14]. Specifically, providing theoretical support for blockchain applications. But, on the one hand, with the advent of the 5G era, we are about it may also have certain limitations, which are affected by to enter a highly dynamic wireless connected digital society complex and uncertain factors in the practical system, and thus mainly composed of massive wireless devices. Sequentially, fail to further support the optimization of blockchain function the explosive information will likely be exchanged through design and network performance improvement. Therefore, it wireless networks. However, the scarcity of wireless spectrum is necessary to systematically and comprehensively study the resources poses challenges for public in the safety and effi- blockchain systems, technologies and applications from the ciency of resource/data management and sharing. On the other perspective of methodology, and objectively evaluate their hand, blockchain relies on frequent communication among advantages and limitations. consensus nodes to reach consensus. While the highly dynamic In this Section, we first discuss the research activity clas- wireless network environment will bring performance and sified into theoretical modeling for blockchain performance security degradation to the communication within blockchain analysis, blockchain-based function design for network ser- consensus nodes. vices, and blockchain-based solution for vertical applications. Therefore, the above-mentioned challenges greatly hinder Next, we discuss the remaining challenge of blockchain sys- the safe and efficient application of blockchain in practical tem, and present the technical roadmap which illustrates the systems and weaken the contribution of blockchain to better relationship in methodology, uses case and topic. improve the practical systems performances. To cope with those challenges, it is first necessary to understand the fun- damentals of blockchain, the operating process of practical A. Research Activity communication systems, as well as the impact of the resources 1) Theoretical modeling for blockchain performance anal- (e.g., communication, network computing resources, etc.) and ysis: The establishment of accurate and effective theoretical other uncertain factors on the performance of blockchain. model is very important to study the system performance such Thus, an accurate and efficient theoretical model need to be as the necessary conditions for achieving consistency, delay established to analyze blockchain system performances and and cost of achieving consistency, and processing capacity of its influencing factors in essence. Then the design of network blockchain, which can provide reliable theoretical guidance services for blockchain-based mechanisms and algorithms can for analyzing the performance of blockchain system. Existing be implemented, but which often require the help of classical projects and studies on system performance focus on he the- methodologies. Finally, in order to facilitate the deployment oretically analyzing in stochastic variable analysis, design of of blockchain-based applications in IoT, Internet of Vehicles consensus protocols for specific applications, and decoupling (IoV) and other scenarios, it is also necessary to explore how of the traditional centralized network architectures for security to use mathematical methods to define a specific blockchain and scalability. Specifically, in the practical blockchain system, problem, and examine various constraints and application there are random behaviors such as block generation time, requirements in the practical systems from the perspective confirmation delay and chain growth rate. Thus, the random of methodology. Therefore, methodology is the basis for variable analysis is often used to observe its impact on the studying the fundamentals, performances and applications of throughput, delay and cost of the blockchain system, so blockchain, and has been recognized. However, most of the as to help researchers effectively and accurately study the existing review research directions focus on the combination running process of the blockchain [44]–[46]. In addition, the of blockchain and other advanced technologies, the integration innovation of different types of consistent algorithms which of blockchain and novel networks, and the security and privacy have different characteristics and serve different purposes has JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 5

TABLE I MAJORBLOCKCHAINPLATFORMS

Hyperledger Platforms Bitcoin [37] Nxt [38] ETH [39] Hashgraph [40] IOTA [41] EOS [43] Fabric [42] Enterprise Application Bitcoin DAPP/Nxt DAPP/ETH Public IOTA Operating block layer transaction transaction transaction network transaction system application Programming Solidity/ JavaScript/ JavaScript Java - Go/Java C++ language Serpent Java/C#/Go Transaction Account Data model Account model Account model - Account model Account model model model Block LevelDB - LevelDB - - FileSystem - storage Communication P2P P2P P2P - HTTP P2P P2P protocol Consortium Public Private Public blockchain/ Public Consortium Consortium Category blockchain blockchain blockchain Private blockchain blockchain blockchain blockchain Consensus PoW PoS PoW BFT Tangle BFT DPoS algorithm Continuous Continuous Continuous Based on Based on Continuous Continuous Architecture single chain single chain single chain DAG DAG chain chain architecture architecture architecture architecture architecture architecture architecture Virtual Self-weight Gossip Computational Computational voting, and protocols, Virtual Solution power Coin age power gossip cumulative endorsement voting competition competition protocols weight and ordering Proof-of- High Low Stake scalability, consensus Low consensus, Low low resource cost, high High transaction universal transaction requirements, security, scalability, Flexible, fee, high blockchain fee, high zero-fee high permission scalable, Characteristic security, low framework, security, low transactions, transaction control and user- network decentralized network secure data throughput, modular friendly resource asset resource transfer, low architecture consumption exchange, consumption offline confirmation proven transactions, latency stability quantum immune https://bitbuc https:// https://bit https://geth. https://github. https://github. Open source ket.org/Jeluri github.com coincore.org/ ethereum.org/ - com/iotaledger/ com/hyperledger address da/nxt/src/ /EOSI en/download/ downloads/ iota.js /fabric master/ O/eos great significance to the security and efficiency of blockchain 2) Blockchain-based function for network services: Relying systems, and thus attracted the attention of many scholars [20]. on mechanism and algorithm design, blockchain can achieve Furthermore, in order to cope with the shortcomings of single various functions and provide network services in a distributed point of failure caused by traditional centralized network, way, including management, incentive, security & privacy, the blockchain technology is used to reshape the network and resource allocation. In addition to the initial financial architecture to improve the network scalability, reliability and service, more research related to blockchain services is being positivity, and the interaction between blockchain and network concentrated on specific areas relevanted to network services, can be observed through the fusion of blockchain and networks such as public and social services [48], cloud services [49], [47]. and other Internet services. For distributed systems with blockchain participation, the consistency of state among nodes However, in practical blockchain systems, the factors affect- is a key to ensure the integrity and security. Importantly, ing its performance are rich and complex. Therefore, in order blockchain system exactly uses the consensus algorithm to to better perform blockchain performance analysis, in addition achieve the consistency of state among nodes, which plays a to the aspects mentioned above, resource constraints and key and irreplaceable role. This fact has fuelled the consensus uncertain aspects of communication, network and computing algorithm is the core component that directly dictates how need to be considered. Thus, it is also necessary to further such a system behaves and the performance it can achieve. design a theoretical model that can capture the characteris- So that, the consensus mechanism is the basis for blockchain tics of complex dynamic scenarios for the optimal design to establish trust and agreements without the participation of the blockchain system, and the network service design of third parties. Generally speaking, consensus mechanism based on the blockchain, to provide the necessary theoretical is a system that can constrain each decentralized node in guidance and design ideas for technological innovation and the decentralized network, maintain the operation order and breakthroughs. fairness of the system, and enable each unrelated node to verify JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 6

Methodology Use case Topic Discuss in

Stochastic Theoretical model: Analysis Section III process understand blockchain running process Resource Section IV Game theory Management

Network service: Optimization mechanism & Incentive Section V algorithm design

Machine Security & Section VI learning Scenario: smart city, privacy smart healthcare, edge computing, Cryptography IoT, etc. Application Section VII

Fig. 3. The technical roadmap on the methodology perspective and confirm the data in the network, so as to generate trust and with satisfactory communication, network and computation reach consensus. consequently, consensus plays a key role in capabilities. the blockchain, which largely determines blockchain system security bound and performance. B. Challenges To establish consistency, various consensus algorithms for different projects are proposed. Most consensus algorithms 1) Theoretical model: Targeted at the problem of scalabil- are originated from typical Proof-of-Work (PoW), Proof of ity, transaction throughput, latency and security, etc, various Stake (PoS), Practical Byzantine Fault Tolerance (PBFT) and algorithms and protocols are designed in existing researches on Raft, and the above algorithms have their own advantages blockchain [54] [55]. In these researches, performance analysis and drawbacks in throughput, confirmation latency, security, are conducted through experimental emulations or application transaction fee and tolerate fault. For example, PoW and PoS implements. However, without a complete theoretical model, it are completely decentralized consensus algorithms, but they is difficult to accurately find out the impact of user behaviors require much meaningful computation to win the competition and system parameters on the performance and security, thus [13]. PBFT and Raft [14] [50] have good performance, but leading to the absence of the precise theoretical indicators, they are suitable for private networks. Besides, the perfor- such as performance boundaries, tolerance to malicious attacks mance of Raft depends on honest nodes and cannot solve BFT. and so on. Importantly, the lack of theoretical foundation For some typical consensus algorithms, a multi-dimensional cannot provide theoretical guidance for the rational application performance comparison is shown in Fig. 2. or technological breakthrough of the blockchain, which will 3) Blockchain-based solution for vertical applications: further limit the development of the blockchain. Therefore, Owning to the benefits of building trust, reducing cost and facing the rapid popularization of blockchain application, to accelerating transactions, blockchain technology is expanding make up for the relative backwardness of the theoretical to other areas including IoV, the industry 4.0, smart homes, basis, an accurate and easily extensible set of mathematical artificial intelligence integrated services, and such [14], [51]– theoretical models is necessary to be established. [53]. In addition, as an important carrier of industrial appli- 2) Technique fusion: Known as a secure, reliable and cations, the construction of blockchain platform has attracted transparent distributed ledger, blockchain has been heralded more and more attention from major blockchain enterprises. as a technology that can energize many scenarios for practical The analysis of existing mainstream platforms is shown in systems, such as spectrum management, Industrial Internet of Table I. However, as we discussed before, the performance Things (IIoT) and so on [56] [57]. However, how to integrate of the mainstream blockchain technologies is limited by blockchain into practical systems especially into wireless communication, network and computation, resulting in some networks, while ensuring its security and performance, is still shortcoming such as high confirmation delay, low throughput, challenging. Designed for different purposes, the essence of intensive computation, and redundant storage. Moreover, these blockchain is data storage and consensus while that of wireless limitations restrict the development of blockchain platforms networks is for data perception, transmission and aggregation. and the landing of industrial applications. Naturally, it is nec- It is necessary to propose a new set of architectures and essary to propose feasible architecture and workflow to support processes that can effectively combine them. Furthermore, as blockchain platform construction and application development the mainstream blockchain protocols, PoW is of high resource JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 7 consumption, direct acyclic graph’s performance is highly performance. Meanwhile, stochastic geometry [60] is another influenced by information load and byzantine fault tolerance mathematical tool widely used in the modeling and analysis is of high communication complexity. These characteristics of communications and networks [61]. In communication form the obstacle for them to be applied to practical system, networks, the actual network nodes and spatial locations can be which is with complex communication environment, limited formulated as random point processes, such as Homogeneous resources and diversified service requirements. Therefore, Poisson Point Process (HPPP), which is to eliminate the faced with the mismatches from communication, network and randomness by traversal to analyze the system performance computation, a dedicated blockchain protocol for practical and provide design ideas theoretically. system is necessary to be designed. 3) Joint optimization: Performance and security of wire- A. Model Briefs less network based blockchain are not only affected by the In practical blockchain system, there exist many random designed blockchain protocol, but also by application environ- behaviors such as block generation time, confirmation delay, ment and business requirements of the intelligent wireless net- chain growth rate, and forking. As mentioned before, it is a na- work. For example, as to Blockchain-Enabled MEC Systems ture design to adopt stochastic process to model these random [58], not only delay/time to finality (DTF) for the blockchain behaviors systematically, which can help us to formulate the system but also energy consumption for the MEC system, are blockchain operation process effectively and accurately. Next, the performance metrics that need to be considered. Then, we discuss how to use stochastic process in blockchain system to avoid the sub-optimal performance resulted in by separate on the typical issues of consensus, security and deployment. optimizations, a joint optimization problem can be formulated In order to formulate the consensus process in blockchain to achieve the optimal trade-off between these two perfor- system as a stochastic process, it is necessary to define a mance metrics. Generally, when it comes to the matching metric (like the cumulative block in PoW, or the cumulative of blockchain and other technologies, from the perspective weight in DAG) to indicate the consensus state at the different of multi-dimensional performance requirements, in order to time. Moreover, if it satisfies Markov properties, the consensus eliminate unilateral bottlenecks and ultimately achieve the process can be formulated as a Markov chain model with one- overall system performance, the joint optimization is necessary step transition probability P = P {X = i |X = i }. to be considered. n+1 n+1 n+1 n

P P P C. Roadmap on the Methodology Perspective 1,1 2,2 m, m This article aims to provide a comprehensive overview on P P P theoretical model, network service and management, and ap- 1,2 2,3 1, S S Ă Ă m- m S plication for blockchain system on a methodology perspective. 1 2 m In order to illustrate the relationship in methodology, use cases and topic, a technical roadmap is shown in Fig. 3. First, Fig. 4. The Stochastic Process in consensus process of the Blockchain. we introduce the calssic methodology to investigate running process of blockchain to formulate and analyze. Then illustrate As shown in Fig. 4, we use Markov chain to model the con- the use cases for blockchain theoretical model, blockchain sensus process. In this process, S (i = 1, 2..., m) represents technology design enabling network service, and how to use i the state in the consensus process, and Pi,j is the one-step blockchain as the fundamental of industry vertical applications transition probability in the Markov chain mode. Accordingly, effectively. Accordingly, we review the state of the art of we can learn the consensus process and gradually understand blockchain system for performance and security analysis, the inner-action. Furthermore, it is also useful to analyze the resource management, incentive mechanism, ect, and some malicious forking attack for security. remaining problems are summarized and discussed. As the most famous consensus process, the PoW-based mining task proposed by bitcoin is a competition among III.STOCHASTIC PROCESSIN BLOCKCHAIN minors, where the winner has the right to generate a new block Stochastic Process [59] is a mathematical method to es- to obtain an amount of reward. For malicious purposes like tablish system model and analyze the performance index in double-spending in the blockchain, forking attack launched uncertain environment, it is a set of time-dependent random by the malicious node which generates blocks to build a variables used to describe the system state at a specific parasite chain privately, it would succeed if the parasite chain time. In communications, especially in wireless scenario, the is longer than the main chain built by the honest node due to complex environment can be abstracted as stochastic process Longest-Chain-Rule (LCR). Indeed, this attack can be treated for analysis in a mathematical manner. As well known, Markov as a competition between honest and malicious nodes. Partic- process has been widely used to describe the common random ularly, without any malicious node, this competition is also process. Generally, if the current system state is determined, the consensus process formulated previously. Therefore, the the trend of system state in the future would be also known, competition for block generation can be modeled as a Poisson no matter what the system state is in the past. Therefore, the process to study the relationship between honest and malicious Markov process is an important and effective mathematical nodes, which is shown in Fig. 5. The malicious nodes compete tool to predict the system state, user action and network against the honest nodes to generate the block. The node which JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 8 has the biggest probability will have the right to generate the Confirmed Tips Unconfirmed block. Accordingly, the factors affecting the vulnerability of Indirect approal blockchain can be known, and the successful probability of Cumulative malicious node can be determined. As a result, the theoretical 6 Weight 9 1 2 insight can be provided to resist forking attack, optimize 1 1 1 consensus mechanisms, and improve network security. 11 7 4 1 1 1 1 Own 1 Consensus 8 1 Weight Target data Honest Attack 1 3 2 1 1 Direct approal Conflict with target data Private Fig. 6. The structure of DAG consensus [62].

PM Malicious chain consideration that the new transaction arrives slowly result- Honest chain ing in a low network load, the system state is defined as PH {W (t),L(t)} that is modeled as a discrete Markov chain Fig. 5. The Stochastic Process in the block generation of the Blockchain. {W (k),L(k)}, k = 0, 1, 2, ..., ∞. When a new transaction x arrives, the change in system Like the classic base station deployment problem in hetero- state can be expressed as geneous networks, deployment of blockchain funtion node (or  W (k) a = 1; called as full node in some literatures) can be solved using W (k + 1) = x (3.1) W (k) +1 a = 0. a stochastic method/stochastic methods in the same manner. x The blockchain system is decentralized which is composed of multiple distributed blockchain nodes, their geographical L(k + 1) = L(k) − 1. (3.2) distribution can be modeled as a Homogeneous Poisson Point where the ax = 1 stands for the situation that the observed Process (HPPP). In this way, we can construct an effective transaction has been approved by an incoming new transac- blockchain system architecture, and analyze the impact of tion, and ax = 0 stands for the situation that the observed blockchain node distribution on the communication through- transaction has not been approved. Since the new transaction put, SNR, security and other network metrics. Therefore, a should select two unselected transactions randomly, thus (3.2) valuable theoretical guidance is developed to optimize the indicates that the new transaction replaces two unselected blockchain node deployment in order to imporve the system transactions as a new one. performance. Therefore, the corresponding one-step transition probabili- ties and Markov chain can be shown as Fig. 7. B. Case Study W : The cumulative weight In this section, we will use our previous work [62] as an i L example to introduce how to model the classic blockchain i : The number of Tips problem as a stochastic process. As shown in Fig. 6, Tangle Pij, : The transition probabilities from state i to j is a DAG based distributed ledger for recording transactions. P P P DAG consensus is a typical voting mechanism to accumu- 1,1 i, i m, m late weight for consensus, which allows the new transaction to randomly select two unselected transactions that are called P P P P 1,2 ...... i-1, i i, i + 1 m-1, m as tips. Therefore, according to the random selection in DAG W1, L 1 Wii , L ...... Wm, L m consensus, selected times of the observed transaction within P P the time period [0, t] is a random variable of t based on the 1, i i, m assumption that the arrival time of the observed transaction is Initial state Intermediate state confirmed state 0. Thus, the cumulative weight of the observed transaction is Fig. 7. The Markov chain for DAG-baed consensus process. its own weight plus the overall number of transactions which select it (the weight of each transaction is assumed to be 1). For example, if the network is from high to low with Therefore, the cumulative weight W (t) is a random variable Unsteady State, the one-step transition probabilities can be with time t ({W (t), t ∈ [0, ∞]}), which is a stochastic process. expressed as: Meanwhile, L(t) is the number of tips at time t, and it is also a stochastic process because it is affected by the random   P {i + 1, j − 1 |i, j } = 2/j, i = 1, 2, ··· ,Ln − 1;j = 2, 3, ··· ,Lh, selection as well as W (t). P {i, j − 1 |i, j } = 1 − 2/j, i = 1, 2, ··· ,Ln − 1;j = 2, 3, ··· ,Lh, In summary, the future states of L(t + 1) and W (t + 1)  P {i + 1, 1 |i, j } = 1, i = 2, ··· , ∞;j = 1. are determined by their current states (L(t) and W (t)) only, (3.3) which satisfies Markov properties. Therefore, this consensus This model is for low network load case, and we can also process can be formulated as a Markov chain. Due to the obtain the modelling for high network load case in the same JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 9 manner. According to this Markov chain model, it is possible deployment of base stations and mobile users as a HPPP for to provide a theoretical guidance for the blockchain imple- mobile edge computing, and derive the theoretical expressions mentation, which can help us to analyze the key performance of relevant performance indicators in various modes using indicators in terms of cumulative weight and confirmation stochastic geometric methods. delay under different network loads, and it is able to evaluate the security performance based on the understanding the D. Summary and Existing Problems impact of network loads. Due to the randomness of the blockchain system, stochastic methods is commonly used for formulation, which can accu- C. Related Work rately describe the distribution of blockchain nodes, the arrival 1) Control of mining difficulty: D. Kraft et al. in [63] of transactions, the behavior of blockchain users and etc. discuss the difficulty-of-work re-adjustment in the blockchain Through these mathematical modeling, researchers can track system. In order to achieve a relatively ideal average block the dynamic evolution of the blockchain system, and further generation rate over a period of time, the mining work is analyze of the consensus process, throughput, and security formulated as a non-homogeneous Poisson process and a new performance. Accordingly, the corresponding theoretical basis method of the difficulty-of-work re-adjustment is proposed. can be provided for performance improving, such as malicious However, the randomness of the hash rate in the blockchain attack preventing, and blockchain application accelerating. system has not been addressed yet. To this end, D. Fullmer Although stochastic process has been widely used in the et al. in [64] consider this situation and introduces a random literature, there are still issues that should be considered and model about block arrival time, in which the marginal dis- improved in the future. tribution of block arrival time and its both expectation and 1) Most existing researches formulate the blockchain sys- variance are derived. Accordingly, we can know that the target tem as a certain stochastic process assuming a simple difficulty value both is a function related to the arrival time model such as the Poisson distribution. However, in of the previous block and affects the block arrival time in the the actual environment, the blockchain running process next retargeting period. usually is more complicated, and how to abstract the 2) Modeling and analysis: Using the stochastic reward common random variables accurately without lossing network (SRN), H. Sukhwani et al. propose a new Hyperledger generality should be well studied. In another word, Fabric v1.0 + system model and study the performance indi- the mathematical formulation must accurately describe cators such as throughput, transaction delay, node utilization, the blockchain system in practice, while considering and queue length in [65]. In addition, this work analyze the the property, complexity and constraint in the view of fullsystem as well as the subsystems corresponding to each theoretical approach. transaction phase in details. The proposed model can provide 2) Currently, some typical issues in a few specific scenarios a quantitative framework that helps a system architect estimate have been widely investigated. In contrast, a general- performance as a function of different system configurations ized stochastic model is in need to describe the whole and makes design trade-offs decisions. N. Papadis et al. in blockchain system. In the future, we should further [66] propose a stochastic network model to describe the joint consider how to use stochastic process to model an end- dynamics of “frontier” processes, track the dynamic evolution to-end blockchain system model, which can systematical of blockchain networks, capture important blockchain features, study the actions of blockchain user and the system and study the impact of delay on security. performance. In addition, understanding the interactions 3) blockchain node deployment: Based on the RAFT con- between various functions of hash operation, cache, sensus mechanism, H. Xu et al. in [50] study the security consensus, communication network and smart contract performance of wireless blockchain networks under malicious is another direction for future research. interference, and provides analysis guidance for the actual deployment of wireless blockchain networks. Y. Zhu et al. in IV. GAME THEORY IN BLOCKCHAIN [67] introduce the blockchain-based heterogeneous network in a air-to-ground IoT heterogeneous network. In order to Game theory [70] [71] is a mathematical theory to study the determine the consensus process and obtain the downlink strategy selection in competitive behaviors. A basic game con- information transmission rate through mathematical analysis, sists of four basic elements: player (decision maker), strategy stochastic geometry method is adopted to model the deploy- (the player’s action), reward (the game result obtained after ment of Ground Sensors (GSs), Air Sensors (ASs), and the the player chooses a strategy) and equilibrium (a balance). place of eavesdroppers with interference attacks. Y. Sun et We can use the game theory to formulate the conflicts and al. in [68] model the location deployment and transaction cooperations between selfish and rational decision-makers. By arrival rate in IoT network as a PPP to study the relationship analyzing both expected and actual behaviors of the players, between communication throughput and transaction through- we can study how how each player generate and optimize put, and propose an optimal communication node deploy- individual strategy under different situation. In a game, if no ment algorithm, which achieves the maximum communication player can obtain more profits by changing his own strategy and transaction throughput with the minimum communication alone, we call that the strategy set of all players at this time node density. M. Liu et al. in [69] formulate the location is at the state of Nash equilibrium [72]. Nash equilibrium JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 10 guarantees that each player’s strategy is optimal no matter consumed, the higher cost would be generated. It means there how the strategy of other players changes. is an effective trade-off between cost and benefit needs to be In recent years, game theory has become an important made. Accordingly, this problem can be also formulated as tool for communication and network research, in which most a game to study the interaction among miners and analyze interactions can be analyzed as game behaviors to find the the best strategy. for instance, game theory is often used to optimal competitive strategy. study equilibrium-based strategy, which guarantees the optimal reward of each miner while avoiding the meaningless mining A. Model Briefs competition caused by greedy resource allocation. Fig. 9 is a Many blockchain problems can be structured as a game game theory model based on auction, the miners will bid for based on game theory. Through the analysis of the optimal the computing resource, and the computing service providers strategy and equilibrium solution, we can investigate the will provide different computing resource. The more bid, the operation process of blockchain mechanisms and the behavior more resource. Therefore, the miner need to balance the cost of blockchain users to optimize the systematic performance and revenue. accordingly. Considering the malicious characteristic, a miner launches Miner 1 Comupting an attack to increase his/her own winning probability while Service unfairly reducing the wining probability of others. In fact, Provider 1 any selfish and rational miner would like to maximize its Miner 2 Comupting own overall reward, which is determined by the environment Resource 2 Service Provider 2

feedbacks, cost and successful attacking probability simultane- Ă Ă ously. Therefore, a miner must consider all possible re-actions Miner n Comupting of others to choose the strategy that is most beneficial to itself Service in this typical non-cooperative game. Therefore, the miner can Provider m be treated as the player, its strategy is whether to launch an Bidder Auctioneer Seller attacking, and the reward function is the expected reward if the attacking succeeds minus the cost for attacking. In this Fig. 9. A auction-based a non-cooperative game model for computing resource allocation. manner, we can formulate this mining process as a game shown in Fig. 8, to study the impact of miners’ strategies Using mobile edge computing in blockchain networks, the (to be honest or malicious) on the blockchain network. Based miner can offload its mining task to edge server, which can on the analysis and equilibrium solution, a game model could solve the limitation of computing resource of blockchain users employed and to provide a theoretical guidance to optimize to extend the blockchain application in wireless scenarios the consensus process in order to improve the security (refrain effectively. To encourage the miner to offload reasonably and from launching the attacking action). motivate edge server to process effectively, the miner should pay an amount of payment to edge server for offloading. As shown in Fig. 10, the miners will offload some mining task Mining Pool to mobile edge. The more computing resource the miner buy, the greater the probability of successfully minings. Stackelberg game can be employed based on the concerning of resource allocation as mentioned above. If addressing the mining task reward reward reward reward assignment (the association between edge server and miner), auction model is a common approach. cost cost cost cost

B. Case Study In this section, we will use our previous work [73] as an example to introduce how to motivate honest actions of honest malicious honest honest participants in blockchain-based scenarios. miner miner miner miner In the typical MEC enabled WBN, controlled by the MEC Non -Attack Attack Non-Attack Non-Attack manager, edge servers are just treated as network resources providers, including computational resources and storage re- sources. However, the MEC server manager may become a central node that is independent of the blockchain system, Fig. 8. A Non-cooperative game framework for the mining process thereby undermining the distributed nature of the blockchain system and further damaging its security. To this end, in During the mining process, a miner should allocate certain this example, the underlying P2P network consists of edge amount of computing resources to increase the wining proba- servers, who are treated as blockchain miners, undertaking bility to get the right that generates a new block with corre- blockchain functionality operations and earning transaction sponding reward. Meanwhile, the more computing resources fees, while IoT devices are treated as blockchain users, who JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 11

Blockchain miner Block newest Block generation Edge server Edge Computing Server Block Block Block Block Block propagation newest newest-2 newest-1 newest

Computing Computing Cost Computing Cost Cost Resource Resource Resource Edge server

The noncooperative Game Blockchain miner ...... Health data uploading

Buyer1 Buyer i Buyer n Blockchain user Monitor

Generate Get block reward Fig. 11. The explaination of the case study in Game Theory [73]

Block Block Block i - 1 i new With blockchain miners acting as the leader while blockchain users acting as followers, a single-leader-multiple- Fig. 10. A MEC-basd non-cooperative game for mining strategy followers Stackelberg game can be used to model interaction between them. Based on the Karush-Kuhn-Tucker (KKT) conditions and backward induction method, a distributed algo- ∗ ∗ only need to upload transactions to blockchain with speci- rithm can be designed to reach the optimal strategy (β , γj ) fied transaction rate requirements. As shown in Fig. 11, the in an iterative manner. blockchain users submit transactions to blockchain miners, and then the blockchain miners execute the mining task and successfully generate a block. Finally, it will be added to the C. Related Work local ledger and broadcast to peers and this operations will 1) Mining pool management: Game theory has been widely also be performed by other blockchain miners once they have used for mining pool management. In [74], X. Liu et al. use an verified it as a valid block. evolutionary game to describe the dynamic evolution of pool At blockchain users’ side, the utility of a blockchain user selection strategies for individual miners, analyze the evolution fj includes the satisfaction degree and incentive cost, i.e., of the mining pool selection strategy considering the hash rate, the transaction fee. Thus, in order to maximize the utility and the broadcast delay of the block. In [75], J. Li et al. define by requiring considerable transaction rate γj, the optimization the mining as a non-priority queuing problem that is deter- problem for fj can be expressed as follows: mined entirely by transaction fee, and propose a transaction queuing game model to study the role of transaction fee in the max Ufj = Sfj (γj) − Cfj (γj) γj consensus process. Based on this model, the authors analyze N the relationship between the mining reward and the time cost, X (4.1) s.t. γj ≤ Γmax and prove the existence of Nash equilibriums. In order to j=1 improve the mining rate, the authors in [76] and [77] formulate the mining process in a PoW-based blockchain as iterative where S (γ ) is the satisfaction degree, C (γ ) is the fj j fj j game, and apply Zero-determinant strategy to optimize the Γ transaction fees and max is the maximum transaction rate mining strategy in order to solve the miners’ dilemma problem. blockchain system can afford. 2) Security strategy: Game theory is also potential to study At the blockchain miners’ side, the utility of them is defined the behaviors of blockchain users and the strategies for security as charged transaction fees minus computational resources concerns. S. Feng et al. in [78] introduce a risk management consumption. Thus, to maximize their revenue, the optimiza- framework for blockchain service, and a Stackelberg game tion problem can be expressed as is adopted to describe the interactions among blockchain N N providers, network insurance companies and blockchain users. X X max Ul = Sl( γj, β) − Cl( γj) (4.2) Based on this game model, the existence and uniqueness of β j=1 j=1 equilibrium are discussed, and the three-party equilibrium- PN based strategy is analyzed to avoid the double-spending at- where N is the number of blockchain users, Sl( j=1 γj, β) tacking. S. Kim et al. in [79] use the evolutionary game to PN is the earning by publishing j=1 γj transactions per hour study the dynamics of mining pool strategy, in which the PN and Cl( j=1 γj) is the corresponding cost of resources con- pool can choose some participating miners to infiltrate into sumption. other pools to launch a block withholding attack. Based on JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 12 the formulated model, the authors qualitatively analyzed the V. OPTIMIZATION IN BLOCKCHAIN influence of malicious infiltrators on mining pool strategy and Optimization theory [84] [85] is a well-known computa- the feasibility of automatic migration among pools. tional tool to solve theoretical analysis and practical engi- 3) Resource management: In order to study the issues of neering issues. Generally, the goal of the optimization theory resource management and pricing between cloud computing is to make the best decisions under some constraints. As a providers and miners, H. Yao et al. in [80] propose a multi- famous subfield of optimization theory, convex optimization agent reinforcement learning algorithm to find the Nash equi- [84] has been widely investigated and applied. Since lots librium of the proposed model, and prove that the Nash equi- of optimization problems can be transformed into a convex librium point of service demand in the system is related to the optimization, this section mainly focuses on the applications expected reward of each miner. Y. Jiao et al. in [81] propose an of convex optimization in blockchain. In addition, because auction-based model to study the interaction between miners optimization problems in large-scale and complicated network and edge service providers, and analyze the allocation and environments are usually nonconvex, convex optimization can pricing of edge computing resource in the blockchain network. also be effectively applied by relaxing or/and approximating Considering the reward of service providers, N. C. Luong et al. some nonconvex conditions. A basic form of convex optimiza- in [82] propose an optimal auction model using deep learning tion can be regarded as constrained optimization problems to solve service providers reward and resource management which can be formulated as issues. D. Xu et al. in [83] study the security issues in min f (x) blockchain edge networks using the game theory as well. In x  gi (x) ≤ 0, i = 1, ··· , m (5.1) this work, a penalty scheme based on behavioral records is s.t. T designed considering the conditions of Nash equilibrium. aj = bj, j = 1, ··· , p where the objective function f(x) and the inequality con- strained function g(x) are convex functions on Rn, and the T D. Summary and Existing Problems equality constraint function hi (x) = ai x − bi must be affine.

As an analysis tool, game theory is widely used to study se- A. Model Briefs curity, mining, and resource allocation problems in blockchain networks. A game model can be built by capturing the Optimization theory is widely used in various fields, which characteristic of the addressed problem in terms of the role of help us to adjust the parameters, schedule the resource and blockchain users, behavior of blockchain decision-maker and determine the decision improving the system performance in a the performance of blockchain system. Thus, the game mod- distributed/centralized manner. In blockchain, it can be used to ling can help researchers understand the impact of different manage the mining, improve security, and jointly optimize the strategy and analyze the optimal strategy based on equilibrium blockchain with other technologies. The mining management solution. optimization is a typical problem in blockchain. As well known, the mining reward for miners is an important factor to Meanwhile, some problems are still remaining to be ad- encourage the contribution of computing resource. Therefore, dressed as follows. maximizing the reward and promoting the accomplish of the 1) Some necessary information should be collected and consensus process is a popular topic in mining management. exchanged in game theory, for example, the selling In order to increase the successful mining probability, miners price of edge server for mining task offloading and usually choose to join the mining pool. Generally, as long as the computing resources used by miners for mining. any miner in the pool succeeds in mining, the mining reward Schedule the communication resource into the game will be distributed to each miner in the pool. Different mining theory need to be well investigated. pools may adopt different reward mechanisms. Therefore, 2) Furthermore, to achieve the equilibrium solution, multi- the miner should consider how to choose the optimal pool round iterations in game theory are needed usually, selection strategy to optimize its reward, and this problem but the corresponding overhead should be considered. can be transformed into a mathematical optimization problem. The overhead which cause commnuication delay, would For this optimization problem, the mining reward obtained generate significant impact on the performance and the by miners under different reward mechanisms can be formu- security of blockchain system. Next, efficient and light- lated as the objective function, the variable is the miner’s weight game theoretic mechanism for blockchain system pool selection strategy, and the constraints are determined should be investigated. by the actual situation. Through this mining pool selection 3) Rationality and selfishness are the basic assumptions optimization problem, we can study the impact of different in game theory. However, these assumptions might be reward mechanisms on the blockchain network in term of the invalid especially for the malicious attacker, whose pur- objective function, and determine the optimal selection under pose is to launch an attacking to ruin the blockchain the constraints. system regardless of the cost. Therefore, how to under- When miners join the mining pool and cooperate with oth- stand and study this extreme malicious action in order ers, the competition among miners becomes the competition to improve security and privacy should be studied in the among the mining pools. In order to win for mining reward, future. some mining pools may choose to take the Block Withholding JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 13

relaying and block verification is shown in Fig. 13. In this Computing Power Computing Power for attacking framework, the authors deploy a series of Access Point (APs) to incentivize the transaction relay and block verification Pool Administer Mining honest Mining Maliciously using DPoS-Based Lightweight Block Verification Scheme. Therefore, an important problem for AP is that how to improve Allocate part computing power to attack the efficiency of transaction relaying and DPoS based block verification, while the payment (cost) is small.

6) Consensus Verifier Set process Block manager Verifier Verifier Candidate Candidate ĊDPoS based Ming Pool A (Attack) Blockahin Ming Pool B transaction phase GenerateG Get 5) voting block reward Blockchain users with stake 4) Download opinions on ĊĊ the n -th AP from the D2D Blockchain New Block Block i Block i -1

Fig. 12. The explaination of BWH attack. 3)Transaction relaying 7) Transaction Local AP confirmation Relay device (BWH) attacking on other mining pools. As shown in Fig. 12, ĉtransaction the pool A will allocate a part of computing power α to another 2)Transaction generation confirmation pool B for attacking. The greater computing power consumed 1)Ask for data by the attacker, the greater the malicious impact on other pools Transmitter 8)Provide data would be happened. The malicious impacts would decline the Receiver computing power for consensus and is costly to the attacker Small Cell itself. As a result, the attacker should choose the optimal computing power for attacking, which can be considered as an Fig. 13. The simplified procedure for transaction relaying and block verifi- optimization problem. In this case, the objective function is to cation [86] maximize the mining reward as well as successful attacking In this work, a two-stage contract theory based on joint probability with the computing power for attacking as the optimization scheme is proposed, where the AP serves as an variable. Through the optimization analysis, we can know the employer who designs all kinds of contracts, pays rewards for optimal attacking strategy, which is the baseline to analyze employees, relays devices, and verifiers serve as employees. and design the consensus process for security based on the Relay devices should consider their battery energy, resource of understanding the malicious action of the attacker. occupied bandwidth, etc., and verifiers should consider their As discussed before, mining task offloading from the miners CPU cycles, energy consumption, etc. In order to maximize to the edge servers can be also formulated as an optimization the expected utility of AP while satisfying the individual ratio- problem with resource allocation. In this case, the miner nality and the incentive compatible constraints for transaction should choose an appropriate offloading strategy to determine relaying and block verification, the objective function in the whether to offload, or to which edge server to offload, and how optimization problem is formulated as many resources (computing, communication and cache) the edge server allocates to the offloading task. Using this optimal max UAP = VR − RR + VV − RV offloading formulation, we can analyze the impact of the (5.2) s.t. {a, b, c, d, e} offloading on the performance and security of the blockchain network, and the optimal solution is to provide a theoretical where the limiting conditions {a, b, c, d, e} are respectively guidance for the blockchain development and application. expressed as (a) For each relay device, the reward paid from AP should B. Case Study be not less than its cost due to the rationality. In this section, we use the previous work [86] as an example (b) Similarly, for each verifier, the reward paid from AP to introduce how to perform optimization in blockchain. In should be not less than its cost due to the rationality. this paper, to improve the security and privacy of D2D (c) AP needs to design the contract for the relay device or (device to device) communication, the authors introduce a verifier flexibly according to their corresponding types. new distributed and secure data sharing framework called (d) Similarly, AP needs design the contract for the verifier D2D blockchain. The simplified procedure for transaction flexibly according to their corresponding types. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 14

(e) For AP, the reward paid to relay devices and verifiers in blockchain-based IoT systems. To maximize the system must be reasonable following a limitation. energy-efficiency, the authors use stochastic programming to where VR is the whole value of transaction relaying created solve the joint optimization problem above. M. Wang et al. by all the relay devices, and RR is the whole payment from in [92] propose a blockchain-based decentralized and truthful AP to relay devices. Similarly, VV is the whole value of block framework for MDC (BC-MDC), which enable the decentral- verification created by all the verifiers, and RV is the whole ization and prevented dishonesty by incorporating a plasma- payment from AP to verifiers. based blockchain into the MDC. There are 4 smart contracts However, to solve the optimization problem in a practical designed for distributed management of worker registration, system, the constraint conditions often are set to non-convex, task posting/allocation, rewards and penalties. Furthermore, thus (5.2) cannot be solved directly. MDC task allocation is formulated as a stochastic optimization By reducing some constraints, this optimization problem problem that jointly minimized the long-term processing cost can be transferred into a convex problem. Finally, the optimal and risk of task failure, and an online task allocation is solution is the best strategy for transaction relaying and block proposed to timely allocate the task to workers to accommo- verification, which can maximize the utility of AP while in- date the network variation. Moreover, a rewarding/penalizing centivizing the relay devices and block verifiers to accomplish scheme is designed to ensure the individually rational to their tasks optimally. truthful of rewarding and prevent the free-riding as well. 3) Optimal algorithm and strategy design: The optimiza- tion problem can be also used to describe various purposes C. Related Work in blockchain system. Y. Zhang et al. in [93] study the 1) Security: As one of the most important issue in routing issue in a blockchain-based payment channel network. blockchain, the security topic has been widely formulated as The authors analyze the payment routing problem using the an optimization problem. O. Onireti et al. in [87] propose a optimization theory method. While considering the constraints practical modeling framework for PBFT, and the viable area of timeliness and feasibility, the authors propose a distributed for wireless PBFT network is defined to ensure the minimum optimization scheme to achieve the lowest total transaction number of replication nodes required for protocol security costs from the sender to the receiver. Applying the consortium and activity. Considering the secured communication and data blockchain technology to electric taxi charging scenarios with sharing between vehicles, J. Kang et al. in [88] propose a two- multiple operators, J. Zhang et al. in [94] propose a new stage security enhancement solution to solve collusion attacks Byzantine fault tolerance algorithm to solve the problem of in IoV. In the first stage, the system selects active miners trust among operators of charging stations. In addition, this and standby miners (candidates) based on reputation voting. work designs a system model based on multi-objective opti- Hence, the active miners selected get the chance to generate mization to maximize operating efficiency and customer satis- block. In the second stage, standby miners will verify the block faction while minimizing the time and distance costs of electric generated by active miners. Therefore, the internal collusion taxis. Z. Jin et al. in [95] propose EdgeChain, a blockchain- among active miners is avoided. However, how to incentivize based architecture to make mobile edge application place- the standby miners to participate is an important problem, ment decisions for multiple service providers. The placement which is solved using contract theory. M. Saad et al. in [89] decision is modeled as a stochastic programming problem model the malicious behavior as a Lyapunov optimization to minimize the placement cost for mobile edge application problem. Then they study how attacker uses the impact of placement scenarios. In this scenario, the blockchain is used memory pool overflow on blockchain users to launch an DDoS to store all placement transactions, which can be traceable by attacking. To prevent DDoS attacks, the authors propose two every mobile edge service providers and application vendors effective countermeasures: fee-based and age-based design. who consume resources at the mobile edge. The core of the fee-based design is to conduct transaction relay with minimum relay fee to reject spam transactions. D. Summary and Existing Problem And similarly, the core of the age-based design is to compare For blockchain system, the mathematical optimization tool transaction mining fee and minimum mining fee. While spam is usually introduced to find the best pool selection strategy transactions are rejected, the DDoS attacks are also resolved. for miners to determine the best task offloading strategy, and 2) Resource allocation: Considering the high-resource con- to allocate resource for consensus process. Although existing sumption, mobile edge computing is a nature design for the researches show that the optimization can achieve the better blockchain-enabled wireless networks. Y. Wu et al. in [90] system performance and security, some problems should be convert the computing resource allocation problem in multi- further studied in the future. access MEC-based blockchain into a mathematical form of 1) The optimization problem and corresponding solution joint optimization problem. To maximize the total revenue of are usually for a specific situation in static or semi- the mobile terminals while ensuring the fairness of the mobile static state. However, the practical blockchain system terminals, they consider two different scenarios, namely a is dynamic and stochastic with some important param- single-edge-server scenario and a multi-edge-servers scenario. eters and constraints that are uncertain and might be For the two scenarios, the authors propose two layered algo- changed over time and space. Therefore, these uncer- rithms to solve the non-convex optimization problem above. tainties should be further considered in optimization S. Fu et al. in [91] study the issue of joint resource allocation formulation. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 15

2) Due to the complexity of the blockchain system, the so- A. Model Briefs lution to the formulated optimization problem might be In recent years, machine learning has been widely used in challenging. For example, the original problem should pattern recognition [110], data mining [111], etc., due to its ca- be transformed into a standard convex problem, de- pabilities in data management, analysis, and decision-making. compose into multi-subproblem, or design an iterative In blockchain, machine learning can provide an efficient and approach to achieve a sub-optimal solution instead of the intelligent approach to discover the malicious action and optimal one. Therefore, the solution method is another recognize the attacks to guarantee the data reliability, system issue that should be addressed considering the solution security and user privacy. On the other hand, machine learning quality, convergence speed and overhead. is also used to make the decision for resource allocation, 3) For modelling, some basic assumptions and simplifica- block size setting and transactions scheduling to optimize the tions with some typical mathematical properties are nec- performance of blockchain system. essary especially for problem formulation and analysis, Malicious attackers can launch double spend attacking but they cannot describe the actual blockchain system [112], denial of service attacking, and eclipse attacking on accurately. Therefore, the acceptable gap between the the blockchain network, which will cause the deteriorated reality (practical system) and ideality (mathematical security risk. In recent years, machine learning is introduced model) should be also well studied. in blockchain to solve security problems. By using unsu- pervised/supervised learning algorithms such as the K-means VI.MACHINE LEARNINGIN BLOCKCHAIN algorithm and supervised SVM, etc., we can monitor the trans- action in consensus process and the behavior of blockchain Machine learning [96] [97] is a method of designing and users [113]. Accordingly, as shown in Fig. 14, we can learn analyzing algorithm to “learn” automatically, which allows the characteristic of both honest and malicious actions, identify computers to analyze from a large amount of data, find out suspicious transaction, and malicious attacker or illegal activity the hidden laws for prediction or classification based on in the network, which can reduce the possibility of successful characteristics of data. Machine learning usually involves the attacks. algorithm of supervised learning [98], unsupervised learning, and reinforcement learning [99], and the model of Surport Extract Vector Machine (SVM) [100], Random Forest (RF) [101], Features Features Machine Learning Layer The nodes in and Deep Learning (DL) [102] [103]. Those models and blockchain Of nodes algorithms mentioned above are widely used in the analysis, Train Train machine Model of prediction, and optimization of communications and networks. dataset learning model recognition

Supervised learning is to train labeled data and analyze the The type of Machine learning Blockchain Layer nodes training data to solve classification and regression problems. Test dataset In contrast, unsupervised learning trains data with no labels Data interface to achieve clustering or dimensionality reduction by finding Receive data infrastructure similarities or internal relationships in the data. Different from Send data layer supervised/unsupervised learning, reinforcement learning is mainly used to solve decision-making problems in a trial-and- Terminals collect data error process based on the interaction and feedback between the agent and environment. SVM discriminates two classes by Fig. 14. Using machine learning to recognize the type of blockchain nodes. fitting an optimal linear separating hyperplane to the training samples of two classes in a multidimensional feature space Nowadays, the resource consumption is an barrier for [104]. Random forest is a more accurate and stable model ob- blockchain applications, especially in the case which is tained by building multiple decision trees and fusing them to- resource-limited such as the IoT device in the wireless network gether, with the advantage of high accuracy and efficient oper- [14]. Therefore, it is necessary to consider the energy-saving ation [105], which is suitable for the case of non-differentiable in mechanism design and resource allocation. To develop model with discrete features and limited values. Deep learning an optimal strategy, we can define the state space s(t), the allows computational models that are composed of multiple action space a(t), and the reward function r(t) to represent processing layers to learn representations of data with multiple the agent, environment, and the feedback between them in a levels of abstraction [106]. It involves AtuoEncode, Variational machine learning manner [114]. The interaction among s(t), Auto-Encoder and Generative Adversarial Network based on a(t) and r(t) is shown in Fig. 15. In a real environmental unsupervised learning, Deep Neural Networks, Convolutional transformation, the probability of going to the next state Neural Networks (CNN), and Recurrent Neural Networks s(t + 1), is related to both the current state s(t) and the (RNN) based on supervised learning [107], [108]. It is based previous state s(t−1), occasionally related to even earlier state on basic features and uses multi-layer activation functions s(t − 2), so that the environmental transformation model may to learn high-dimensional nonlinear features, including CNN be too complex to model. Therefore, the way to simplify the networks suitable for image domains, RNN networks suitable environmental transformation model of reinforcement learning for time series and WiDE & Deep networks suitable for is to assume the markov property of state transformation: the recommendation domains, etc [109]. probability of transformation to the next state s(t + 1) is only JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 16 related to the current state s(t), and has nothing to do with Global the previous state [115]. Accordingly, to maximize the long- model term reward, the optimal strategy determination considering the interaction and feedback over time can be treated as a Markov decision process, and it is able to be achieved using reinforcement learning or deep reinforcement learning algorithms. Local model uploading State space st() Local data Device 1 Device 2 Reward r( t ) agent environment (a) Traditional FL

Action a( t ) Block generation

Mining reward M 1 Cross M 2 Fig. 15. The explaination of the interaction and feedback between agent and Verification environment in a machine learning manner. Blockchain Data In addition, blockchain is a potential solution to the prob- reward lems in machine learning. Most typically, the decentralized feature of blockchain can be used to solve the problem of Local model single point of failure caused by aggregating machine learning uploading models using centralized servers. Local data Device 1 Device 2 B. Case Study Ledger Ledger In this section, the previous work [116] is used as an exam- ple to illustrate the benefits of the combination of blockchain (b) BlockFL and machine learning. Federated learning (FL) [117], as a promising training Fig. 16. (a) The structure of traditional FL; (b) The structure of the proposed paradigm, was proposed by Google to tackle the privacy BlockFL [116] and security problems of centralized machine learning and to alleviate the communication load of the core network. As model is aggregated anywhere as long as the latest block can shown in Fig. 16 (a), many devices collaborate in solving a be obtained. In addition, the data reward for devices will be machine learning problem by updating the local model with issued by the associated miner according to the size of the their own local data, under the coordination of the centralized device data sample; the mining reward for miners will be FL server. However, the traditional FL, which relies on a issued by the blockchain network according to the total volume centralized FL server for model aggregation is vulnerable of data used by the connected devices. With the reasonable to experience service paralysis even when a single point of incentives, the blockchain-based FL can be positively driven failure happens. All local models updated from devices will for efficient training. be distorted by the inaccurate global model aggregated at the Due to the decentralized architecture of BlockFL, the mal- FL server. In addition, for some devices with massive amounts function of each miner only distorts the global model of its of data, if there is no credible incentive mechanism, they are own devices instead of paralyzing the entire system. Moreover, usually unwilling to participate in training, which brings great such distortion can be recovered by interaction with other challenge to the rapid convergence of the FL model. regular miners or federating with other devices associated with To address the problems mentioned above, H. Kim proposed regular miners. Besides, by optimizing the block generation a blockchained FL (BlockFL) architecture, which is shown in rate, the time for BlockFL to complete the model training Fig. 16(b). With the distributed and non-tamperable charac- can be reduced and the performance of the system can be teristics of the blockchain, the use of a blockchain network improved. instead of the centralized FL server can effectively overcome the issue of single point failure. The model parameters up- loaded by the devices are taken as the transactions, and will C. Related Work be recorded in the candidate block after being verified by 1) Security: Due to the capability of learning, analyzing the associated miner. After all miners reached a consensus, and classifying, machine learning is used to monitor behaviors devices can obtain the latest global model by aggregating the and to detect the malicious attack for the blockchain security. local model updates contained in the newly generated block S. Dey et al. in [113] combine machine learning and game downloaded from the associated miner. Generally, the global theory to solve the majority-attack problem. In this work, JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 17 the activities of attackers and participants in the network are top of the trust-enhanced blockchain topology, BlockP2P-EP monitored to judge and classify both participants’ motivation executes the parallel spanning tree broadcasting algorithm to and the service value in transactions, and then detect network achieve fast data broadcasting among nodes in terms of intra- anomalies. Therefore, the probability of majority-attack will and inter-clusters. be reduced. H. Tang et al. in [118] introduce a deep learning- 3) Application in IoT: Machine learning is also useful to based algorithm to identify and classify malicious nodes by apply blockchain in various fields such as IoT, IoV and smart classifying behavior patterns in the network. The proposed grid. In order to implement blockchain technology into IoT algorithm can reduce the probability of the blockchain network fields and achieve the condition-based management on the being attacked by malicious nodes. Experimental results show blockchain, T. Id in [125] applies blockchain and machine that the proposed algorithm is significantly more effective learning into the task of anomaly detections in the IoT. The than the existing conventional methods. In order to detect authors transform the collaborative anomaly detection task of anomalies (such as DDoS, double-spend and denial-of-service the blockchain into multi-task probabilisitc dictionary learning. attacks) in electronic transactions of Bitcoin, S. Sayadi et Then, the statistical machine learning algorithms is used to al. in [119] propose an anomaly detection model based on solve major technical issues such as building and validation of machine learning. This work proposes two modes of machine consensus in the blockchain. To ensure the performance of data learning approaches: 1) the One Class Support Vector Ma- aggregation, data storage, and data processing in IoT services, chines (OCSVM) algorithm to detect outliers, 2) the K-Means N. C. Luong in [126] use blockchain to support IoT services algorithm in order to group outliers based on similiar type of which is based on cognitive radio network. To help secondary anomalies. Experimental results show that the proposed model users (IoT devices) to choose an optimal transaction trans- can accurately identify the types of attacks, and can provide the mission under intricate conditions, the authors get an optimal theoretical guidance to improve the security of the electronic transaction transmission policy for secondary users by adopt- trading system based on Bitcoin. P. Thai et al. in [120] ing a Double Deep-Q Network (DDQN) algorithms that can focus particularly on the anomaly detection to the Bitcoin allow the secondary users to learn the optimal policy above. H. transaction network, with the goal of detecting suspicious Yao et al. in [80] introduce blockchain and cloud computing users and transactions. The data is first represented in two into IoT to offload computational task from the IIoT network focal points: users and transactions. And then, unsupervised itself. In addition, this paper models the resource interaction learning methods including k-means, and SVM are utilized to between cloud providers and miners as a Stackelberg game, detect anomalies. M. Shin et al. in [121] propose a clustering while proposing a multiagent reinforcement learning algorithm method for bitcoin block and transaction data analysis, which to find the Nash equilibrium of the proposed game. defines the data that can be collected from the Bitcoin network, 4) Application in smart grid: To protect the smart grid from and the statistics of the blocks that can be extracted from the suffering cyber attacks, M. A. Ferrag et al. in [127] propose a collected data. In addition, this paper performs a clustering novel deep learning and blockchain-based energy framework. experiment by applying Principal Component Analysis (PCA) This framework consists of two schemes: a blockchain based to the extracted data, and also testes how to apply PCA to the scheme and a deep learning-based scheme. The blockchain- clustering data. based scheme is used to facilitate the exchange of excess 2) Performance improvement: Machine learning is widely energy among neighboring nodes. And the deep learning-based used for performance improvement of the blockchain systems. scheme is used to detect attacks and fraudulent transactions to To meet the great demand of the blockchain, the solution need enhance the system reliability and security. to be found for the scalability problem. Nowadays, sharding 5) Application in VANETs: For the problem that the data method is found to resolve the scalability problem of the collected by different entities in the vehicle social network blockchain. A. Bugday et al. in [122] propose a method which (VSNs) usually contains very different attributes, M. Shen uses adaptive machine learning model and Verifiable Random et al. in [100] propose a privacy-preserving SVM classifier Functions together to assign nodes to achieve shards. As shard- training scheme over vertically-partitioned datasets posessed ing method solves the scalability problem, the performance of by multiple data providers. In addition, consortium blockchain the blockchain is improved. In order to optimize the system and threshold homomorphic cryptosystem are used to establish performance of the blockchain-based IoV, M. Liu et al. in a secure SVM classifier training platform without a trusted [123] use Deep Reinforcement Learning Technology (DLT) to third-party. To improve the security and reduce the attack in select the consensus algorithm and the block generated nodes, the vehicular ad hoc networks (VANETs), C. Dai et al. in as well as adjust the block size and the interval between [128] propose an indirect reciprocity security framework. This block generations. The solution proposed can be applied to framework tries to encourage the On Board Units (OBUs) to the dynamic IoV scenario, and it can maximie the system help each others to reduce attacks. According to a designed throughput without affecting the system’s decentralization, social norm, the framework assigns a scalar reputation to latency and security. W. Hao et al. in [124] establish a each OBUs to evaluate their dangerous level to the VANET, trust-enhanced blockchain P2P topology (BlockP2P-EP) that and apply the blockchain technique to protect the reputation considers the transmission rate and transmission reliability from being tampered. Each OBU under the indirect reciprocity to improve the performance of the blockchain network in principle takes actions to another OBU based on the reputation achieving fast and reliable broadcasting. BlockP2P-EP first which is assigned by the framework. Therefore, a selection uses K-means to cluster neighboring peer nodes. Then on strategy for OBUs based on Q-learning (one kind of the JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 18 reinforcement learning) is proposed to solve the core problem blockchain is chain-of-blocks, the chain can be seemed as of action selection. According to the strategy, the OBU can the relationship of blocks, and the block is an encapsulated achieve the optimal action. Therefore, the proposed framework data structure to cache data/transactions. In order to maintain can efficiently increase both the reputation and the utility of the blockchain system, blockchain users perform consensus the each OBU, and improve the security of the VANETs. mechanism to generate blocks continuously, and hash algo- rithm in cryptography is carried out for transaction integrity, D. Summary and Existing Problems proof of consensus, writing new block in chain, etc. Besides, a binary Merkle tree is used for data structure which can In the literature, a number of works have shown that using confirm the existence and integrity of the transactions quickly. machine learning in both data management and analysis can Fig. 17 shows the application of asymmetric encryption in effectively monitor and classify the behavior of blockchain users, as well as recognize malicious behavior, suspicious Public key Blockchain Private key user and illegal activity in the system, therefore both reduce Inforamation block block information the possibility of attacks and optimize the performance of network system. Comparing with the traditional method of optimization or game theory, machine learning can adjust its strategy Fig. 17. The Cryptography used in Blockchain. according to the changing of environment based on the long- term reward maximizing. However, there are still two existing the blockchain. with asymmetric encryption, blockchain users problems that should be further investigated. can use the public key to encrypt information in transaction 1) Machine learning is valuable to understand the to ensure the security. Meanwhile, the private key is used to blockchain process and behavior in order to optimize sign the transaction digitally by any blockchain user, and the system throughput and security (i.e., the transaction others can use the corresponding public key to verify and to processing speed and malicious attack recognition), how prevent in a distributed manner. For the privacy issue caused to use the ability of machine learning in management, by data stored in the blockchain, some scheme can be adopted analysis and prediction to provide an intelligent guide- to provide anonymity decoupling users’ information from line for decision-making and mechanism design is still published transactions, involving of public key encryption an open issue. scheme with keyword searchable, anonymous digital certificate 2) The impact of convergence and accuracy of selected publishing scheme, and zero-knowledge proof. For data pro- machine learning algorithms on the performance and cessing security in the blockchain, homomorphic encryption security of blockchain system should be considered. technology enables users to encrypt the transaction data using the corresponding encryption algorithm before submitting the transaction data to the block chain network. The data exists VII.CRYPTOGRAPHY IN BLOCKCHAIN in ciphertext, which will not reveal any privacy information Cryptography [129] is a method for confidential commu- of users even if it is obtained by the attackers. Meanwhile, nication based on information transformation in a prescribed the ciphertext operation result is consistent with the plaintext way, which is to ensure the confidentiality, integrity, authenti- operation result. cation and non-repudiation. To improve information security, cryptography has been widely used in security identification, such as access control and privacy protection. Symmetric Encrypted encryption and asymmetric encryption are the mainstream data Sensor Cluster head Use public key Data cryptographic techniques to prevent information from being to encrypt eavesdropped. Symmetric encryption use the same key for decrypt encryption and decryption, such as Data Encryption Standard success (DES) [130], Advanced Encryption Standard (AES) [131], and decrypt fail International Data Encryption Algorithm (IDEA). Asymmet- Private Key ric encryption uses public and private keys to encrypt and decrypt, such as Rivest-Shamir-Adleman (RSA) [132] [133], Attribute Miners with Miners with Elliptic-curve cryptography (ECC) [134], and Elgamal [135]. Authorities Right Attributes Flase Attributes Symmetric encryptionis fast and simple, but as the number of users increases, it will face the risk of key management and key leakage [136]. In contrast, asymmetric encryption has high Fig. 18. Cryptography theory for access control and authentication in security but low efficiency [137]. blockchain systems In the meantime, we can apply cryptography theory for A. Model Briefs access control and authentication in the blockchain systems. Cryptography is the basic theory in blockchain, in which As shown in Fig. 18, by controlling the right of data reading hashing, asymmetric encryption, timestamp, Merkle tree, etc. and user access, only the users who have the right attributes have been widely used to guarantee the security of transactions can decrypted the encrypted data and access the blockchain and the privacy of user information. The basic structure of systems, while those users who do not have the right attribute JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 19

cannot decypt data successfully. Thesefore, the security of stor- • Ad Publication: Advertisers with valid certificates pack- ing data and information in the blockchain can be guaranteed age ad information as a transaction and submit it to smart by using crytography. contracts, where the ad information is stored as the form of Merkle hash trees. Advertisers also generate ECDSA B. Case Study signatures so that participants can verify the validity of In this section,we use the previous work [138] to intro- the transaction. Then, they select the suitable roadside duce the application of cryptography in blockchain. Due to unit to broadcast the ad. the attributes of blockchain, many researchers try to apply • Ad Dissemination: The selected Road Side Unit (RSU) blockchain technology to vehicular networks to improve its broadcasts ad to nearby vehicles, or vehicles forward ad- security. However, possible privacy leakage issues will reduce vertising to other vehicles. In addition, both the forward- the enthusiasm of vehicles to participate in ad dissemina- ing vehicle and the receiving vehicle will use ZKPoK tion. In order to solve the vehicle privacy problem of the to generate anonymous credentials to prove that the blockchain-based vehicular networks, the authors use Zero- forwarding process is actually took place. knowledge proof of knowledge (ZKPoK) to propose a new • Reward Payment:. At this stage, RSU checks the valid- blockchain-based ad dissemination framework which is shown ity of the anonymous credential and proof-of-ad-receiving in Fig. 19. submitted by the forwarding vehicle. If the verification is To ensure the validation of ad dissemination and improve passed, the valid advertisement receiver and sender will the security of the vehicular networks, this work designs a receive corresponding rewards. concrete, fair and anonymous scheme under the proposed framework. In order to ensure fairness, Merkle hash trees and C. Related Work smart contracts are used to implement the proof of ad reception to mitigate “free-riding” attacks. In addition,the proposed 1) Security: M. Noel et al. in [139] present basic anal- scheme can protect vehicles’ privacy in terms of anonymity ysis and the background understanding of Stateful Hash- and conditional unlinkability based on zero-knowledge proof based Signature Schemes, particularly the Lamport One-Time techniques. In addition, we briefly explain the five-stage pro- Signature Scheme, Winternitz One-Time Signature Scheme, cess involved in the privacy protection blockchain architecture and the Merkle Signature Scheme. Moreover, the three hash- using Merkle hash tree and ZKPoK. based digital signatures are compared in terms of key gen- eration, signature generation, verification, and security lev- els. K. Chalkias et al. in [140] propose BPQS, a scalable post-quantum (PQ) digital signature scheme best suited for Resgister Authority blockchain and distributed ledger technologies (DLTs). It can Valid leverage application-specific chain/graph structures to reduce Cerificates ESDSA Valid ZKPoK the cost of key generation, signing and validation, as well Public KeyCerificates as decreasing the size of signature. Besides, an open source Wallet implementation of the scheme is provided in this paper. Client   Merkle hash trees Blockchain-based Ad Furthermore, compared with other benchmark schemes (e.g., Transactions ECDSA signatures Dissemination SHA256 and SHA384), BPQS is superior to existing hashing In Vehicular Nerworks Smart Contract algorithms when reusing keys for a reasonable number of signatures. The reason is that it support a fallback mechanism Advertiser RSU Anonymous to allow for an almost unlimited number of signatures if credential and proof Ads of ad receivng needed. M. Sato et al. in [141] propose a scheme to extend ZKPoK Messages the validity of past blocks when the underlying cryptographic algorithms (hash functions and digital signatures) are de- ZKPoK Response stroyed. This scheme can effectively avoid the hard-fork of the original blockchain when the hash function is compromised, and provide the smooth-fork when the digital signature scheme Fig. 19. The blockchain-based AD Dissemination framework [138] is compromised. 2) Privacy: Cryptography is widely used to protect the • System Setup: Register authority initializes the system privacy of users in the blockchain. G. Micaliey al. in [142] and generates public and private keys based on predefined utilize the partial homomorphic encryption to enhance data security parameters. privacy, and resist both collision attack and prime attack in the • Registration: Advertisers and vehicles run sub-protocols blockchain. In order to protect the privacy of blockchain users, to obtain valid certificates or keys individually. In the sub- P. Zhong et al in [143] design a Privacy-Protected blockchain protocol for advertisers, the registration authority adds system to protect the privacy of users in the blockchain. registration information based on the ECDSA public key This system encrypts all data uploaded by blockchain users sent by the advertiser. In the sub-protocol for vehicle within the agreed time. During this period, the user’s personal operation, register authority verifies the vehicle’s identity data will not leak unless he voluntarily publish these data. information according to the ZKPoK sent by the vehicle. When the agreed time is reached, the system publishes the JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 20 decryption key and verifies the user’s behavior to avoid nology with digital signature, called zk-DASNARK. Based on fraudulent behavior by malicious users. Y. Rahulamathavan this, zk-AuthFeed (a zero-knowledge authenticated data feed et al. in [144] apply attribute-based encryption to control scheme) is designed to achieve data privacy and authenticity the permission of data access and usage in the blockchain of smart contracts, and it is implemented using libsnark and system to achieve privacy protection. M. Zhang et al. in [145] Ethereum to prove the efficiency of the proposed scheme. propose an ID-based encryption scheme to improve the data 5) Supervisability: To solve the nonsupervisability problem privacy particularly for the non-transaction applications in in the blockchain-based IoT e-commerce autonomous transac- the blockchain. This scheme encrypts the plaintext into the tion management system, C. Liu et al. in [153] propose a ciphertext to hide the information for preventing disguise and new transaction settlement system called NormaChain. This passive attacks. One innovation of this sheme is that complex work design a three-layers sharding blockchain network and certificate management and issuance in traditional PKI systems an innovative decentralized public key searchable encryption can be avoid without using advance technologies such as zero- scheme (decentralized public key encryption with keyword knowledge proof. search (PEKS) scheme) to resist chosen ciphertext attacks, 3) Zero-knowledge proof: Zero-knowledge proof (ZKP) cryptanalysis, and collusion. H. Kang et al. in [154] propose provides the ability to prove secrets without directly disclos- a blockchain network called FabZK. By only storing the ing them. It guarantees that the proof will not reveal more encrypted data of each transaction, and anonymizing the information about private input, nor can it infer from the transaction relationship between participants at the same time, calculations. A large number of research scholars have used this blockchain network FabZK hides the transaction details in ZKP to solve blockchain privacy issues. M. Harikrishnan et the shared ledger. In addition, it achieves both privacy and au- al. in [146] study the problem of confidentiality of data in ditability by supporting verifiable Pedersen commitments and the blockchain network and design a new interactive ZKP constructing zero-knowledge proofs. Evaluation result shows encryption technology-ZKSTARK, which is achieved by se- that this FabZK offers strong privacy-preserving capabilities, lecting two indistinguishable hash functions in the blockchain while delivering reasonable performance for the applications system and integrating them into the ZKP protocol. M. H. developed based on its framework. MurtazaIn et al in [147] introduce a simple non-interactive ZKP scheme for blockchain-based electronic voting systems, D. Summary and Existing Problem namely zkSNARKs. This work uses digital signatures for message authentication, cryptographic hash functions for get- Cryptography is the key element for confidentiality, in- ting a message digest, and ZKP for obtaining unlinkability, tegrity, authentication and non-repudiation for security and to achieve the security and confidentiality of the electronic privacy in blockchain using the typical scheme like digital voting system. On this basis, D. Ding et al. in [148] propose signature, asymmetric encryption, hashing and etc. These a blockchain privacy protection scheme using accounts and methods focus on security and privacy protection in an external multi-asset model, where the zkSNARKs algorithm is use to manner, which are to design a powerful wall around the generate and to verify the zero-knowledge proof. This scheme blockchain system to deny the attacks from malicious users. is faster than ZKSTARK, but it requires trusted settings, However, the system would be vulnerable if any bugs found by unable to resist quantum attacks, and is less secure than the attacker to break. Therefore, how to empower the security ZKSTARK. Y. Tsai et al. in [149] introduce an improved and privacy ability of blockchain system in a systematic man- non-interactive zero-knowledge range proof scheme based on ner providing inherent safety is still an open issue. Moreover, the predecessors. This work use the Fujisaki-Okamoto com- the mechanism and protocol should be designed to optimize mitment, non-interactive zero-knowledge and computational the performance as well as security simultaneously. bindingness through proof of knowledge in the cyclic group with secret order technique (CBPKCGSO), to achieve better VIII.OPEN ISSUE DISCUSSIONS security and efficiency. This paper discussed the theoretical model research of 4) Verification: A. S. Sani et al. in [150] introduce a blockchain basic knowledge, the design of network services new high-performance and scalable blockchain network. This based on blockchain mechanisms and algorithms, and the blockchain network uses Time-based Zero-Knowledge Proof deployment of blockchain-based applications in practical sys- of Knowledge (T-ZKPK) to perform identity verification and tems from a methodological perspective, and the reference the establishment of key protection transactions, which can classification is shown in Table II. In the meantime, some enhance the security and privacy of IIoT. S. Zhu et al. remaining problems in terms of technical, commercial and in [151] propose a novel hybrid blockchain crowdsourcing political views are still needed to be discussed as the open platform to achieve decentralization and privacy preservation. issues. On this platform, a hybrid blockchain structure, dual-ledgers and dual-consensus algorithms are integrated to ensure secured communication between requesters and workers. In addition, A. Technology Issue the smart contract and zero-knowledge proof are deployed to From the resource consumption perspective, many achieve permission control and privacy protection. Z. Wan blockchain protocols, which consumes less power than PoW, et al. in [152] design a zero-knowledge SNARK scheme for have been proposed, such as PoS, IOTA and so on. However, authenticated data by effectively combining zk-SNARK tech- they struggle to be widely used in wireless networks because JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 21

TABLE II REFERENCE CLASSIFICATION ON METHODOLOGY PERSPECTIVE

Research category Theoretical model Network service Application Method Markov performance and security - - process analysis in IoT [62] Stochastic Poisson mining difficulty node deployment in IoT [68], - process control [63] [64] node deployment in MEC [69] Stochastic - - node deployment in IoT [67] geometry modeling and performance Others - - analysis [65] [66] incentive mechanismin IoT [73], Stackelberg - security strategy in network [78], - game resource management in MEC [80] Game mining pool Evolutionary management [74], - - game mining pool security strategy [79] Iterative mining pool - - game management [76] [77] resource management Auction - - in MEC [81] [82] mining pool security Others - management [75] in edge networks [83] Convex security in D2D - - optimization communication [86] Geometric resource allocation Optimization - - programming in IoT [91] optimal algorithm and Stochastic - strategy design in mobile - programming edge network [95] Lyapunov resource allocation in DDoS attack avoidance [89] - Optimization mobile device cloud [92] security in IoV [88], optimal algorithm and analytical framework modeling resource allocation in MEC [90], Others strategy design in electric for PBFT [87] optimal algorithm and strategy design in taxi charging scenarios [94] payment channel network [93] Supervised majority-attack avoidance [113] - - learning Unsupervised performance security in Bitcoin [119] [120] [121] - Machine learning optimization [124] Learning privacy and security in Federated - centralized machine - learning learning [116] Deep identify malicious application in IoT [126], - learning nodes [118] application in smart grid [127] Reinforcement resource management - - learning in IoT [80] Deep performance optimization reinforcement - security in IoV [128] in IoV [123] learning sharding management [122], SVM traning platform Others - collaborative anomaly for VSNs [100] detection in IoT [125] Asymmetric security issues [139] [140], - - encryption user privacy [143] Cryptography Partial homomorphic - user privacy [142] - encryption privacy [138] [146] [147] ZKP - [148] [149] [154], - identity verification [150] [151] [152] security [141] [153], Others - - privacy [144] [145] JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 22 of their scalability. As to another type of blockchain protocols reducing the probability of successful transactions. Therefore, which vote instead of calculating, such as PBFT, Hashgraph in the wireless blockchain system, a framework is needed and so on, they are still not widely applicable to wireless to measure the communication overhead and communication networks due to their communication complexity. From quality in the communication process. the communication perspective, caused by blockchain’s transaction release, consensus interaction, ledger updating IX.CONCLUSION and so on, additional communication overhead is bound to Blockchain is an emerging technology that is considered be incurred along with the improvement of data security as one of the key enablers of 5G networks due to its brought by blockchain. There have to be a tradeoff between unique features of decentralization, scalability, security and the communication performance and the security performance the corresponding characteristics. In this article, we presented of the blockchain network. In conclusion, essential questions a comprehensive survey focusing on the current most advanced to be technologically addressed include: 1) how to design achievements in exploring the intrinsic nature of blockchain a blockchain protocol with high scalability and appropriate from a methodological perspective. Based on the state of power consumption, communication complexity while art literatures, we outlined the theoretical model research for ensuring its security, 2) how to compromise between blockchain fundamentals understanding, the network service blockchain’s performance and network’s performance. design for blockchain-based mechanisms and algorithms, as well as the application of blockchain for Internet of Things B. Commercial Issue and etc. We first introduced the working principles, research In addition to the technical matters, the lack of killer activities, and challenges of blockchain, as well as illustrated applications [155] and business models also hinders the large- the roadmap involving the classic methodology with typical scale application of blockchain. For private or consortium blockchain use cases and topics. Subsequently, we discussed blockchains, only the untamperability of the data on chain can the contribution of the methodology to the performance of be guaranteed. While the authenticity of the data off chain blockchain systems, focusing on the role of stochastic process, requires the cooperation of other technologies form Internet game theory, optimization, machine learning and cryptography of Things. In terms of public chain, it is more about the in both the study of blockchain operation process and the transformation of the whole system, which is suitable for the design of blockchain protocol/algorithm. Finally, we pointed system without effective incentive system or reliable allocation out several blockchain issues from technical, commercial, and mechanism. The first areas to to be implemented will be those political perspectives. Although the blockchain is still in its that have formed a consensus but lack incentives and cannot infancy, it is clear that blockchain will significantly improve be implemented on a large scale. the landscape and experience of future network services and applications. C. Policy Issue REFERENCES The lack of policies, regulations, standardization and other [1] S. Nakamoto, “Bitcoin: a peer-to-peer electronic cash system,” 2009. related policy issues are hindering the mass commercialization [Online]. Available: https://bitcoin.org/bitcoin.pdf. of blockchain. In the formulation of them, the development and [2] Ethereum, “What is ethereum?” 2020. [Online]. Available: https: limitations of blockchain should be taken into consideration. //ethereum.org/what-is-ethereum/. [3] N. community, “Nxt: a peer-to-peer digital socioeconomic system,” At the mean time, the selection of solutions in blockchain must White paper, 2014. also comply with the restrictions of policies, regulations and [4] G. Zyskind, O. Nathan, and A. S. Pentland, “Decentralizing privacy: standards. Using blockchain to protect personal data,” in 2015 IEEE Security and Privacy Workshops, 2015, pp. 180–184. [5] S. Mondal, K. P. Wijewardena, S. Karuppuswami, N. Kriti, D. Kumar, and P. Chahal, “Blockchain inspired rfid-based information architecture D. Communication Issue for food supply chain,” IEEE Internet of Things Journal, vol. 6, no. 3, Blockchain is built in a P2P network, which means that pp. 5803–5813, 2019. [6] Y. Yuan and F. Wang, “Towards blockchain-based intelligent trans- a large amount of communication overhead will be added portation systems,” in 2016 IEEE 19th International Conference on in terms of network traffic and system processing capacity. Intelligent Transportation Systems (ITSC), 2016, pp. 2663–2668. Because current applications are highly dynamic, data sources [7] D. Bhowmik and T. Feng, “The multimedia blockchain: A distributed and tamper-proof media transaction framework,” in 2017 22nd Inter- frequently change. This indicates that the node may have to national Conference on Digital Signal Processing (DSP), 2017, pp. send a large number of update transactions, which will further 1–5. increase the communication overhead. On the other hand, [8] D. Liu, A. Alahmadi, J. Ni, X. Lin, and X. Shen, “Anonymous rep- utation system for iiot-enabled retail marketing atop pos blockchain,” blockchain deployment in wireless is foreseeable in the near IEEE Transactions on Industrial Informatics, vol. 15, no. 6, pp. 3527– future [36]. In this case, blockchain services rely on wire- 3537, 2019. less communication networks to reach consensus. During the [9] H. Xu, L. Zhang, O. Onireti, Y. Fang, W. B. Buchanan, and M. A. Im- ran, “Beeptrace: Blockchain-enabled privacy-preserving contact tracing consensus process, blockchain nodes are connected through for covid-19 pandemic and beyond,” 2020. wireless channels. However, due to factors such as wireless [10] N. Al-Zaben, M. M. Hassan Onik, J. Yang, N. Lee, and C. Kim, “Gen- channel fading and unauthorized malicious interference, the eral data protection regulation complied blockchain architecture for personally identifiable information management,” in 2018 International wireless connection among blockchain nodes may be attacked, Conference on Computing, Electronics Communications Engineering and the uplink or downlink transmission may fail, thereby (iCCECE), 2018, pp. 77–82. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 23

[11] R. S. Sangwan, M. Kassab, and C. Capitolo, “Architectural [34] H. Baniata and A. Kertesz, “A survey on blockchain-fog integration considerations for blockchain based systems for financial transactions,” approaches,” IEEE Access, vol. 8, pp. 102 657–102 668, 2020. Procedia Computer Science, vol. 168, pp. 265 – 271, 2020, ’Complex [35] Z. Liu, N. C. Luong, W. Wang, D. Niyato, P. Wang, Y. C. Liang, and Adaptive Systems’ Malvern, PennsylvaniaNovember 13-15, 2019. D. I. Kim, “A Survey on Blockchain: A Game Theoretical Perspective,” [Online]. Available: http://www.sciencedirect.com/science/article/pii/ IEEE Access, vol. 7, pp. 47 615–47 643, 2019. S1877050920303914 [36] L. Zhang, H. Xu, O. Onireti, M. Imran, and B. Cao, “How much com- [12] B. Granetto, R. Kandaswamy, J. Lovelock, and M. Reynolds, “Forecast: munication resource is needed to run a wireless blockchain network?” Blockchain business value, worldwide, 2017-2030,” Gartner, 2017. submitted to IEEE Wireless Communications Magazine, 2020. [13] T. Salman, M. Zolanvari, A. Erbad, R. Jain, and M. Samaka, “Security [37] S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,” [On- services using blockchains: A state of the art survey,” IEEE Commu- line]. Available: http://www.ildgroup.si/uploads/product/20/bitcoin.pdf, nications Surveys Tutorials, vol. 21, no. 1, pp. 858–880, 2019. 2008. [14] H. Xu, P. V. Klaine, O. Onireti, B. Cao, M. Imran, and L. Zhang, [38] “Nxt.” [Online]. Available: https://www.jelurida.com/nxt, 2020. “Blockchain-enabled resource management and sharing for 6g com- [39] “Ethereum,” [Online]. Available: https://www.ethereum.org/, 2016. munications,” arXiv preprint arXiv: 2003.13083, 2020. [40] “Hashgraph,” [Online]. Available: https://hashgraph.com/, 2020. [15] B. Cao, Y. Li, L. Zhang, L. Zhang, S. Mumtaz, Z. Zhou, and M. Peng, [41] S. Popov, “The tangle,” [Online]. Available: http://iotatoken.com/IOTA “When internet of things meets blockchain: Challenges in distributed Whitepaper.pdf, 2017. consensus,” IEEE Network, vol. 33, no. 6, pp. 133–139, 2019. [42] E. Androulaki, A. Barger, V. Bortnikov, C. Cachin, K. Christidis, [16] Y. Hu, M. Liyanage, A. Mansoor, K. Thilakarathna, G. Jourjon, and A. De Caro, D. Enyeart, C. Ferris, G. Laventman, Y. Manevich, A. Seneviratne, “Blockchain-based Smart Contracts - Applications and S. Muralidharan, C. Murthy, B. Nguyen, M. Sethi, G. Singh, K. Smith, Challenges,” arXiv e-prints, Sep. 2018. A. Sorniotti, C. Stathakopoulou, M. Vukolic,´ S. Weed Cocco, and [17] N. Waheed, X. He, M. Usman, and M. Usman, “Security & Privacy in J. Yellick, “Hyperledger Fabric: A Distributed Operating System for IoT Using Machine Learning & Blockchain: Threats & Countermea- Permissioned Blockchains,” arXiv e-prints, p. arXiv:1801.10228, Jan. sures,” arXiv preprint arXiv:2002.03488, 2020. 2018. [18] M. Conti, E. Sandeep Kumar, C. Lal, and S. Ruj, “A survey on [43] “Eosio,” [Online]. Available: https://eos.io, 2020. security and privacy issues of bitcoin,” IEEE Communications Surveys [44] Z. Zhang, M. Zargham, and V. Preciado, “On modeling blockchain- Tutorials, vol. 20, no. 4, pp. 3416–3452, 2018. enabled economic networks as stochastic dynamical systems,” Applied [19] M. Saad, J. Spaulding, L. Njilla, C. Kamhoua, S. Shetty, D. Nyang, Network Science, vol. 5, p. 19, 2020. and D. Mohaisen, “Exploring the Attack Surface of Blockchain: A [45] S. Guo, X. Hu, S. Guo, X. Qiu, and F. Qi, “Blockchain meets edge IEEE Internet of Things Journal Comprehensive Survey,” , vol. 22, computing: A distributed and trusted authentication system,” IEEE no. 3, pp. 1977–2008, 2020. Transactions on Industrial Informatics, vol. 16, no. 3, pp. 1972–1983, [20] M. S. Ferdous, M. Chowdhury, M. Hoque, and A. Colman, 2020. “Blockchain consensus algorithms: A survey,” 2020. [On- [46] A. Gopalan, A. Sankararaman, A. Walid, and S. Vishwanath, “Stability line]. Available: https://www.researchgate.net/publication/338738073 and scalability of blockchain systems,” 2019, pp. 71–72. Blockchain Consensus Algorithms A Survey [47] X. Ling, J. Wang, T. Bouchoucha, B. C. Levy, and Z. Ding, [21] W. Wang, D. T. Hoang, P. Hu, Z. Xiong, D. Niyato, P. Wang, Y. Wen, “Blockchain radio access network (b-ran): Towards decentralized se- and D. I. Kim, “A Survey on Consensus Mechanisms and Mining cure radio access paradigm,” IEEE Access, vol. 7, pp. 9714–9723, 2019. Strategy Management in Blockchain Networks,” IEEE Access, vol. 7, pp. 22 328–22 370, 2019. [48] A. Chakravorty and C. Rong, “Ushare: User controlled social Proceedings of the 11th [22] Y. Liu, F. R. Yu, X. Li, H. Ji, and V. C. M. Leung, “Blockchain media based on blockchain,” in International Conference on Ubiquitous Information Management and Machine Learning for Communications and Networking Systems,” and Communication IEEE Communications Surveys & Tutorials, no. c, pp. 1–1, 2020. , ser. IMCOM ’17. New York, NY, USA: [23] Y. Xiao, N. Zhang, W. Lou, and Y. T. Hou, “A Survey of Distributed Association for Computing Machinery, 2017. [Online]. Available: Consensus Protocols for Blockchain Networks,” IEEE Communications https://doi.org/10.1145/3022227.3022325 Surveys & Tutorials, no. c, pp. 1–1, 2020. [49] Q. Xia, E. B. Sifah, K. O. Asamoah, J. Gao, X. Du, and M. Guizani, [24] J. Xie, F. R. Yu, T. Huang, R. Xie, J. Liu, and Y. Liu, “A survey on “Medshare: Trust-less medical data sharing among cloud service the scalability of blockchain systems,” IEEE Network, vol. 33, no. 5, providers via blockchain,” IEEE Access, vol. 5, pp. 14 757–14 767, pp. 166–173, Sep. 2019. 2017. [25] H. T. M. Gamage, H. Weerasinghe, and N. G. J. Dias, “A survey [50] H. Xu, L. Zhang, Y. Liu, and B. Cao, “Raft based wireless blockchain on blockchain technology concepts, applications, and issues,” 2020. networks in the presence of malicious jamming,” IEEE Wireless Com- [Online]. Available: https://doi.org/10.1007/s42979-020-00123-0 munications Letters, vol. 9, no. 6, pp. 817–821, 2020. [26] R. Belchior, A. Vasconcelos, S. Guerreiro, and M. Correia, “A Survey [51] V. Sharma, “An energy-efficient transaction model for the blockchain- on Blockchain Interoperability: Past, Present, and Future Trends,” arXiv enabled internet of vehicles (iov),” IEEE Communications Letters, preprint arXiv:2005.14282, 2020. vol. 23, no. 2, pp. 246–249, 2019. [27] D. C. Nguyen, P. Pathirana, M. Ding, and A. Seneviratne, “Blockchain [52] S. Huh, S. Cho, and S. Kim, “Managing iot devices using blockchain for 5g and beyond networks: A state of the art survey,” Journal of platform,” in 2017 19th International Conference on Advanced Com- Network and Computer Applications, p. 102693, May. 2020. munication Technology (ICACT), 2017, pp. 464–467. [28] G. Yu, X. Wang, K. Yu, W. Ni, J. A. Zhang, and R. P. Liu, “Survey: [53] “Blockchain for smart homes: Review of current trends and research Sharding in blockchains,” IEEE Access, vol. 8, pp. 14 155–14 181, challenges,” Computers & Electrical Engineering, vol. 83, p. 106585, 2020. 2020. [29] M. A. Ferrag, M. Derdour, M. Mukherjee, A. Derhab, L. Maglaras, [54] I. Eyal, A. E. Gencer, E. G. Sirer, and R. Van Renesse, “Bitcoin-ng: and H. Janicke, “Blockchain technologies for the internet of things: A scalable blockchain protocol,” 2015. Research issues and challenges,” IEEE Internet of Things Journal, [55] P. Robinson, R. Ramesh, J. Brainard, and S. Johnson, “Atomic Cross- vol. 6, no. 2, pp. 2188–2204, 2019. chain Transactions White Paper,” arXiv preprint arXiv:2003.00903v2, [30] M. Chowdhury, M. S. Ferdous, K. Biswas, N. Chowdhury, and 2020. V. Muthukkumarasamy, “A survey on blockchain-based platforms for [56] K. Zhang, Y. Zhu, S. Maharjan, and Y. Zhang, “Edge intelligence and iot use-cases,” The Knowledge Engineering Review, vol. 35, May. 2020. blockchain empowered 5g beyond for the industrial internet of things,” [31] J. Xie, H. Tang, T. Huang, F. R. Yu, R. Xie, J. Liu, and Y. Liu, “A IEEE Network, vol. 33, no. 5, pp. 12–19, 2019. Survey of Blockchain Technology Applied to Smart Cities: Research [57] M. B. H. Weiss, K. Werbach, D. C. Sicker, and C. E. C. Bastidas, Issues and Challenges,” IEEE Communications Surveys and Tutorials, “On the application of blockchains to spectrum management,” IEEE vol. 21, no. 3, pp. 2794–2830, 2019. Transactions on Cognitive Communications and Networking, vol. 5, [32] K. Gai, J. Guo, L. Zhu, and S. Yu, “Blockchain Meets Cloud Comput- no. 2, pp. 193–205, 2019. ing: A Survey,” IEEE Communications Surveys and Tutorials, no. c, [58] J. Feng, F. R. Yu, Q. Pei, J. Du, and L. Zhu, “Joint optimization of pp. 1–22, 2020. radio and computational resources allocation in blockchain-enabled [33] R. Yang, F. R. Yu, P. Si, Z. Yang, and Y. Zhang, “Integrated Blockchain mobile edge computing systems,” IEEE Transactions on Wireless and Edge Computing Systems: A Survey, Some Research Issues and Communications, vol. 19, no. 6, pp. 4321–4334, 2020. Challenges,” IEEE Communications Surveys and Tutorials, vol. 21, [59] A. Papoulis and S. U. Pillai, Probability, random variables, and no. 2, pp. 1508–1532, 2019. stochastic processes. Tata McGraw-Hill Education, 2002. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 24

[60] F. Baccelli and B. Błaszczyszyn, Stochastic geometry and wireless [84] H. Hindi, “A tutorial on convex optimization II: duality and interior networks. Now Publishers Inc, 2010, vol. 1. point methods,” in 2006 American Control Conference. IEEE, 2006, [61] S. N. Chiu, D. Stoyan, W. S. Kendall, and J. Mecke, Stochastic pp. 11—-pp. geometry and its applications. John Wiley & Sons, 2013. [85] P. Berck and K. Sydsæter, “Linear and nonlinear programming,” in [62] Y. Li, B. Cao, M. Peng, L. Zhang, L. Zhang, D. Feng, and J. Yu, “Direct Economists’ Mathematical Manual. Springer, 1993, pp. 71–76. acyclic graph based blockchain for internet of things: Performance and [86] L. Jiang, S. Xie, S. Maharjan, and Y. Zhang, “Joint Transaction Relay- security analysis,” arXiv preprint arXiv:1905.10925, 2019. ing and Block Verification Optimization for Blockchain Empowered [63] D. Kraft, “Difficulty control for blockchain-based consensus systems,” D2D Communication,” IEEE Transactions on Vehicular Technology, Peer-to-Peer Networking and Applications, vol. 9, no. 2, pp. 397–413, 2019. 2016. [87] O. Onireti, L. Zhang, and M. A. Imran, “On the viable area of wireless [64] D. Fullmer and A. S. Morse, “Analysis of difficulty control in bitcoin practical byzantine fault tolerance (pbft) blockchain networks,” in 2019 and proof-of-work blockchains,” in 2018 IEEE Conference on Decision IEEE Global Communications Conference (GLOBECOM), 2019, pp. and Control (CDC). IEEE, 2018, pp. 5988–5992. 1–6. [65] H. Sukhwani, N. Wang, K. S. Trivedi, and A. Rindos, “Performance [88] J. Kang, Z. Xiong, D. Niyato, D. Ye, D. I. Kim, and J. Zhao, “Toward modeling of hyperledger fabric (permissioned blockchain network),” in secure blockchain-enabled internet of vehicles: Optimizing consensus 2018 IEEE 17th International Symposium on Network Computing and management using reputation and contract theory,” IEEE Transactions Applications (NCA). IEEE, 2018, pp. 1–8. on Vehicular Technology, vol. 68, no. 3, pp. 2906–2920, 2019. [66] N. Papadis, S. Borst, A. Walid, M. Grissa, and L. Tassiulas, “Stochastic [89] M. Saad, L. Njilla, C. Kamhoua, J. Kim, D. Nyang, and A. Mohaisen, models and wide-area network measurements for blockchain design “Mempool Optimization for Defending Against DDoS Attacks in PoW- and analysis,” in IEEE INFOCOM 2018-IEEE Conference on Computer based Blockchain Systems,” in 2019 IEEE International Conference on Communications. IEEE, 2018, pp. 2546–2554. Blockchain and Cryptocurrency (ICBC). IEEE, 2019, pp. 285–292. [67] Y. Zhu, G. Zheng, and K.-K. Wong, “Blockchain-empowered decentral- [90] Y. Wu, J. Shi, X. Chen, K. Ni, L. P. Qian, and K. Zhang, “Optimal ized storage in air-to-ground industrial networks,” IEEE Transactions Multi-access Computation Offloading for Mobile Blockchain,” in 2018 on Industrial Informatics, vol. 15, no. 6, pp. 3593–3601, 2019. IEEE International Conference on Communication Systems (ICCS). [68] Y. Sun, L. Zhang, G. Feng, B. Yang, B. Cao, and M. A. Imran, IEEE, 2018, pp. 198–203. “Blockchain-enabled wireless Internet of Things: Performance analysis [91] S. Fu, L. Zhao, X. Ling, and H. Zhang, “Maximizing the system and optimal communication node deployment,” IEEE Internet of Things energy efficiency in the blockchain based Internet of Things,” in ICC Journal, vol. 6, no. 3, pp. 5791–5802, 2019. 2019-2019 IEEE International Conference on Communications (ICC). [69] M. Liu, F. R. Yu, Y. Teng, V. C. M. Leung, and M. Song, “Computation IEEE, 2019, pp. 1–6. offloading and content caching in wireless blockchain networks with [92] M. Wang, C. Xu, X. Chen, L. Zhong, Z. Wu, and D. O. Wu, “Bc- mobile edge computing,” IEEE Transactions on Vehicular Technology, mdc: A blockchain-based decentralized truthful framework for mobile vol. 67, no. 11, pp. 11 008–11 021, 2018. device cloud,” IEEE Transactions on Industrial Informatics, pp. 1–1, [70] R. B. Myerson, Game theory. Harvard university press, 2013. 2020. [71] Z. Han, D. Niyato, W. Saad, T. Ba\csar, and A. Hjørungnes, Game [93] Y. Zhang, D. Yang, and G. Xue, “Cheapay: An optimal algorithm for theory in wireless and communication networks: theory, models, and fee minimization in blockchain-based payment channel networks,” in applications. Cambridge university press, 2012. ICC 2019-2019 IEEE International Conference on Communications [72] J. Nash, “Non-cooperative games,” Annals of mathematics, pp. 286– (ICC). IEEE, 2019, pp. 1–6. 295, 1951. [94] Z. Jin, R. Wu, X. Chen, and G. Li, “Charging Guiding Strategy for [73] W. Liu, B. Cao, L. Zhang, M. Peng, and M. Daneshmand, “A Electric Taxis Based on Consortium Blockchain,” IEEE Access, vol. 7, Distributed Game Theoretic Approach for Blockchain-based Offloading pp. 144 144–144 153, 2019. Strategy,” 2020. [95] H. Zhu, C. Huang, and J. Zhou, “Edgechain: Blockchain-based multi- [74] X. Liu, W. Wang, D. Niyato, N. Zhao, and P. Wang, “Evolutionary vendor mobile edge application placement,” in 2018 4th IEEE Confer- game for mining pool selection in blockchain networks,” IEEE Wireless ence on Network Softwarization and Workshops (NetSoft), 2018, pp. Communications Letters, vol. 7, no. 5, pp. 760–763, 2018. 222–226. [75] J. Li, Y. Yuan, S. Wang, and F.-Y. Wang, “Transaction queuing game [96] T. O. Ayodele, “Introduction to machine learning,” New Advances in in bitcoin blockchain,” in 2018 IEEE Intelligent Vehicles Symposium Machine Learning, pp. 1–9, 2010. (IV). IEEE, 2018, pp. 114–119. [97] P. Langley and H. A. Simon, “Applications of machine learning and [76] Y. Zhen, M. Yue, C. Zhong-yu, T. Chang-bing, and C. Xin, “Zero- rule induction,” Communications of the ACM, vol. 38, no. 11, pp. 54– determinant strategy for the algorithm optimize of blockchain PoW 64, 1995. consensus,” in 2017 36th Chinese Control Conference (CCC). IEEE, [98] R. Saravanan and P. Sujatha, “A state of art techniques on machine 2017, pp. 1441–1446. learning algorithms: A perspective of supervised learning approaches [77] C. Tang, C. Li, X. Yu, Z. Zheng, and Z. Chen, “Cooperative mining in data classification,” in 2018 Second International Conference on in blockchain networks with zero-determinant strategies,” IEEE trans- Intelligent Computing and Control Systems (ICICCS). IEEE, 2018, actions on cybernetics, 2019. pp. 945–949. [78] S. Feng, Z. Xiong, D. Niyato, P. Wang, S. S. Wang, and Y. Zhang, [99] L. P. Kaelbling, M. L. Littman, and A. W. Moore, “Reinforcement “Cyber Risk Management with Risk Aware Cyber-Insurance in learning: A survey,” Journal of artificial intelligence research, vol. 4, Blockchain Networks,” in 2018 IEEE Global Communications Con- pp. 237–285, 1996. ference (GLOBECOM). IEEE, 2018, pp. 1–7. [100] M. Shen, J. Zhang, L. Zhu, K. Xu, and X. Tang, “Secure svm training [79] S. Kim and S.-G. Hahn, “Mining Pool Manipulation in Blockchain over vertically-partitioned datasets using consortium blockchain for ve- Network Over Evolutionary Block Withholding Attack,” IEEE Access, hicular social networks,” IEEE Transactions on Vehicular Technology, vol. 7, pp. 144 230–144 244, 2019. vol. 69, no. 6, pp. 5773–5783, 2020. [80] H. Yao, T. Mai, J. Wang, Z. Ji, C. Jiang, and Y. Qian, “Resource trading [101] H. J. Singh and A. S. Hafid, “Transaction confirmation time prediction in blockchain-based industrial Internet of Things,” IEEE Transactions in ethereum blockchain using machine learning,” 2019. on Industrial Informatics, vol. 15, no. 6, pp. 3602–3609, 2019. [102] I. Arel, D. C. Rose, and T. P. Karnowski, “Deep machine learning-a [81] Y. Jiao, P. Wang, D. Niyato, and Z. Xiong, “Social welfare maxi- new frontier in artificial intelligence research [research frontier],” IEEE mization auction in edge computing resource allocation for mobile computational intelligence magazine, vol. 5, no. 4, pp. 13–18, 2010. blockchain,” in 2018 IEEE international conference on communications [103] L. Deng, “A tutorial survey of architectures, algorithms, and ap- (ICC). IEEE, 2018, pp. 1–6. plications for deep learning,” APSIPA Transactions on Signal and [82] N. C. Luong, Z. Xiong, P. Wang, and D. Niyato, “Optimal auction for Information Processing, vol. 3, 2014. edge computing resource management in mobile blockchain networks: [104] B. Waske and J. A. Benediktsson, “Fusion of support vector machines A deep learning approach,” in 2018 IEEE International Conference on for classification of multisensor data,” IEEE Transactions on Geo- Communications (ICC). IEEE, 2018, pp. 1–6. science and Remote Sensing, vol. 45, no. 12, pp. 3858–3866, 2007. [83] D. Xu, L. Xiao, L. Sun, and M. Lei, “Game theoretic study on [105] L. Breiman, “Random forests,” Machine Learning, vol. 43, no. 1, pp. blockchain based secure edge networks,” in 2017 IEEE/CIC Interna- 5–32, 2001. tional Conference on Communications in China (ICCC). IEEE, 2017, [106] Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. pp. 1–5. 521, pp. 436–44, 05 2015. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 25

[107] L. Deng, “A tutorial survey of architectures, algorithms, and applica- [130] F. PUB, “Data Encryption Standard (DES),” FIPS PUB, pp. 43–46, tions for deep learning – erratum,” APSIPA Transactions on Signal and 1999. Information Processing, vol. 3, 12 2014. [131] N.-F. Standard, “Announcing the advanced encryption standard (aes),” [108] G. Litjens, T. Kooi, B. Ehteshami Bejnordi, A. Setio, F. Ciompi, Federal Information Processing Standards Publication, vol. 197, no. M. Ghafoorian, J. van der Laak, B. Ginneken, and C. Sanchez,´ “A 1-51, p. 3, 2001. survey on deep learning in medical image analysis,” Medical Image [132] R. L. Rivest, A. Shamir, and L. Adleman, “A method for obtaining Analysis, vol. 42, 02 2017. digital signatures and public-key cryptosystems,” Communications of [109] H.-T. Cheng, L. Koc, J. Harmsen, T. Shaked, T. Chandra, H. Aradhye, the ACM, vol. 21, no. 2, pp. 120–126, 1978. G. Anderson, G. Corrado, W. Chai, M. Ispir, R. Anil, Z. Haque, [133] S. R. Singh, A. K. Khan, and S. R. Singh, “Performance evaluation of L. Hong, V. Jain, X. Liu, and H. Shah, “Wide & deep learning for RSA and elliptic curve cryptography,” in 2016 2nd International Con- recommender systems,” 2016. ference on Contemporary Computing and Informatics (IC3I). IEEE, [110] C. Bishop, Pattern Recognition and Machine Learning (Information 2016, pp. 302–306. Science and Statistics), 2006. [134] A. Jurisicˇ and A. Menezes, “Elliptic curves and cryptography,” Dr. [111] G. Nguyen, S. Dlugolinsky, M. Bobak, V. Tran, A. Lopez Garcia, Dobb’s Journal, pp. 26–36, 1997. I. Heredia, P. Mal´ık, and L. Hluchy,´ “Machine learning and deep [135] T. ElGamal, “A public key cryptosystem and a signature scheme based learning frameworks and libraries for large-scale data mining: a survey,” on discrete logarithms,” IEEE transactions on information theory, Artificial Intelligence Review, vol. 52, pp. 77–124, 01 2019. vol. 31, no. 4, pp. 469–472, 1985. [112] G. Karame, E. Androulaki, M. Roeschlin, A. Gervais, and S. Capkun, [136] S. Chandra, S. Paira, S. S. Alam, and G. Sanyal, “A comparative survey “Misbehavior in bitcoin: A study of double-spending and accountabil- of symmetric and asymmetric key cryptography,” in 2014 International ity,” ACM Transactions on Information and System Security, vol. 18, Conference on Electronics, Communication and Computational Engi- pp. 1–32, 06 2015. neering (ICECCE). IEEE, 2014, pp. 83–93. [113] S. Dey, “Securing majority-attack in blockchain using machine learning [137] G. Simmons, “Symmetric and asymmetric encryption,” ACM Comput. and algorithmic game theory: A proof of work,” in 2018 10th computer Surv., vol. 11, pp. 305–330, 12 1979. science and electronic engineering (CEEC). IEEE, 2018, pp. 7–10. [138] M. Li, J. Weng, A. Yang, J.-N. Liu, and X. Lin, “Toward Blockchain- [114] R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction, Based Fair and Anonymous Ad Dissemination in Vehicular Networks,” 2nd Edition, 2018. IEEE Transactions on Vehicular Technology, vol. 68, no. 11, pp. [115] C. Szepesvari,´ “Reinforcement learning algorithms for mdps,” Tech. 11 248–11 259, 2019. Rep., 2009. [139] M. D. Noel, O. V. Waziri, M. S. Abdulhamid, and A. J. Ojeniyi, “State- [116] H. Kim, J. Park, M. Bennis, and S.-L. Kim, “Blockchained on-device ful hash-based digital signature schemes for bitcoin cryptocurrency,” federated learning,” IEEE Communications Letters, 2019. in 2019 15th International Conference on Electronics, Computer and [117] H. B. McMahan, E. Moore, D. Ramage, S. Hampson, and Others, Computation (ICECCO), 2019, pp. 1–6. “Communication-efficient learning of deep networks from decentral- [140] K. Chalkias, J. Brown, M. Hearn, T. Lillehagen, I. Nitto, and ized data,” arXiv preprint arXiv:1602.05629, 2016. T. Schroeter, “Blockchained post-quantum signatures,” in 2018 IEEE [118] H. Tang, Y. Jiao, B. Huang, C. Lin, S. Goyal, and B. Wang, “Learning International Conference on Internet of Things (iThings) and IEEE to Classify Blockchain Peers According to Their Behavior Sequences,” Green Computing and Communications (GreenCom) and IEEE Cyber, IEEE Access, vol. 6, pp. 71 208–71 215, 2018. Physical and Social Computing (CPSCom) and IEEE Smart Data [119] S. Sayadi, S. B. Rejeb, and Z. Choukair, “Anomaly Detection Model (SmartData), 2018, pp. 1196–1203. Over Blockchain Electronic Transactions,” in 2019 15th International [141] M. Sato and S. Matsuo, “Long-term public blockchain: Resilience Wireless Communications & Mobile Computing Conference (IWCMC). against compromise of underlying cryptography,” in 2017 26th In- IEEE, 2019, pp. 895–900. ternational Conference on Computer Communication and Networks [120] T. Pham and S. Lee, “Anomaly detection in bitcoin network using (ICCCN), 2017, pp. 1–8. unsupervised learning methods,” 2016. [142] S. Yaji, K. Bangera, and B. Neelima, “Privacy preserving in blockchain [121] M. Shin, U. Baek, K. Shim, J. Park, S. Yoon, and M. Kim, “Block based on partial homomorphic encryption system for AI applications,” analysis in bitcoin system using clustering with dimension reduction,” in 2018 IEEE 25th International Conference on High Performance in 2019 20th Asia-Pacific Network Operations and Management Sym- Computing Workshops (HiPCW). IEEE, 2018, pp. 81–85. posium (APNOMS), 2019, pp. 1–4. [143] P. Zhong, Q. Zhong, H. Mi, S. Zhang, and Y. Xiang, “Privacy-Protected [122] A. Bugday, A. Ozsoy, and H. Sever, “Securing Blockchain Shards By Blockchain System,” in 2019 20th IEEE International Conference on Using Learning Based Reputation and Verifiable Random Functions,” Mobile Data Management (MDM). IEEE, 2019, pp. 457–461. in 2019 International Symposium on Networks, Computers and Com- [144] Y. Rahulamathavan, R. C.-W. Phan, M. Rajarajan, S. Misra, and munications (ISNCC). IEEE, 2019, pp. 1–4. A. Kondoz, “Privacy-preserving blockchain based IoT ecosystem using [123] M. Liu, Y. Teng, F. R. Yu, V. C. M. Leung, and M. Song, “Deep Re- attribute-based encryption,” in 2017 IEEE International Conference on inforcement Learning Based Performance Optimization in Blockchain- Advanced Networks and Telecommunications Systems (ANTS). IEEE, Enabled Internet of Vehicle,” in ICC 2019-2019 IEEE International 2017, pp. 1–6. Conference on Communications (ICC). IEEE, 2019, pp. 1–6. [145] M. Zhang, S. Wang, P. Zhang, L. He, X. Li, and S. Zhou, “Protect- [124] W. Hao, J. Zeng, X. Dai, J. Xiao, Q. Hua, H. Chen, K. Li, and H. Jin, ing Data Privacy for Permissioned Blockchains using Identity-Based “Towards a trust-enhanced blockchain p2p topology for enabling fast Encryption,” in 2019 IEEE 3rd Information Technology, Networking, and reliable broadcast,” IEEE Transactions on Network and Service Electronic and Automation Control Conference (ITNEC). IEEE, 2019, Management, pp. 1–1, 2020. pp. 602–605. [125] T. Ide,´ “Collaborative anomaly detection on blockchain from noisy [146] M. Harikrishnan and K. V. Lakshmy, “Secure Digital Service Payments sensor data,” in 2018 IEEE International Conference on Data Mining using Zero Knowledge Proof in Distributed Network,” in 2019 5th Workshops (ICDMW). IEEE, 2018, pp. 120–127. International Conference on Advanced Computing & Communication [126] N. C. Luong, T. T. Anh, H. T. T. Binh, D. Niyato, D. I. Kim, and Y.-C. Systems (ICACCS). IEEE, 2019, pp. 307–312. Liang, “Joint transaction transmission and channel selection in cogni- [147] M. H. Murtaza, Z. A. Alizai, and Z. Iqbal, “Blockchain Based tive radio based blockchain networks: A deep reinforcement learning Anonymous Voting System Using zkSNARKs,” in 2019 International approach,” in ICASSP 2019-2019 IEEE International Conference on Conference on Applied and Engineering Mathematics (ICAEM). IEEE, Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019, pp. 2019, pp. 209–214. 8409–8413. [148] D. Ding, K. Li, L. Jia, Z. Li, J. Li, and Y. Sun, “Privacy protection for [127] M. A. Ferrag and L. Maglaras, “DeepCoin: A novel deep learning and blockchains with account and multi-asset model,” China Communica- blockchain-based energy exchange framework for smart grids,” IEEE tions, vol. 16, no. 6, pp. 69–79, 2019. Transactions on Engineering Management, 2019. [149] Y. C. Tsai, R. Tso, Z.-Y. Liu, and K. Chen, “An Improved Non- [128] C. Dai, X. Xiao, Y. Ding, L. Xiao, Y. Tang, and S. Zhou, “Learning Interactive Zero-Knowledge Range Proof for Decentralized Appli- Based Security for VANET with Blockchain,” in 2018 IEEE Interna- cations,” in 2019 IEEE International Conference on Decentralized tional Conference on Communication Systems (ICCS). IEEE, 2018, Applications and Infrastructures (DAPPCON). IEEE, 2019, pp. 129– pp. 210–215. 134. [129] W. Diffie and M. Hellman, “New directions in cryptography,” IEEE [150] A. S. Sani, D. Yuan, W. Bao, P. L. Yeoh, Z. Y. Dong, B. Vucetic, and transactions on Information Theory, vol. 22, no. 6, pp. 644–654, 1976. E. Bertino, “Xyreum: A high-performance and scalable blockchain for JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 26

iiot security and privacy,” in 2019 IEEE 39th International Conference Daquan Feng received the Ph.D. degree in information engineering from on Distributed Computing Systems (ICDCS), 2019, pp. 1920–1930. the University of Electronic Science and Technology of China in 2015. He [151] S. Zhu, H. Hu, Y. Li, and W. Li, “Hybrid blockchain design for pri- had been a visiting student with the School of Electrical and Computer vacy preserving crowdsourcing platform,” in 2019 IEEE International Engineering, Georgia Institute of Technology, USA, from 2011 to 2014. Conference on Blockchain (Blockchain), 2019, pp. 26–33. After graduation, he was a research staff in the State Radio Monitoring [152] Z. Wan, Z. Guan, Y. Zhou, and K. Ren, “zk-authfeed: How to feed Center, Beijing, China, and then a Postdoctoral Research Fellow in Singapore authenticated data into smart contract with zero knowledge,” in 2019 University of Technology and Design. He is now an Assistant Professor with IEEE International Conference on Blockchain (Blockchain), 2019, pp. the Shenzhen Key Laboratory of Digital Creative Technology, the Guangdong 83–90. Province Engineering Laboratory for Digital Creative Technology, College [153] C. Liu, Y. Xiao, V. Javangula, Q. Hu, S. Wang, and X. Cheng, of Electronics and Information Engineering, Shenzhen University, Shenzhen, “NormaChain: A blockchain-based normalized autonomous transaction China. His research interests include blockchain technology, URLLC com- settlement system for IoT-based E-commerce,” IEEE Internet of Things munications, MEC, and massive IoT networks. He is an Associate Editor of Journal, vol. 6, no. 3, pp. 4680–4693, 2018. IEEE COMMUNICATIONS LETTERS. [154] H. Kang, T. Dai, N. Jean-Louis, S. Tao, and X. Gu, “FabZK: Sup- porting Privacy-Preserving, Auditable Smart Contracts in Hyperledger Fabric,” in 2019 49th Annual IEEE/IFIP International Conference on Dependable Systems and Networks (DSN). IEEE, 2019, pp. 543–555. [155] W. Kenton, “Killer application,” website, 2018, https://www. investopedia.com/terms/k/killerapplication.asp.

Mugen Peng received the Ph.D. degree in communication and informa- tion systems from the Beijing University of Posts and Telecommunications Bin Cao is an associate professor in the state key laboratory of network and (BUPT), Beijing, China, in 2005. Afterward, he joined BUPT, where he has switching technology at Beijing University of Posts and Telecommunications been a Full Professor with the School of Information and Communication (BUPT). Before that, he was an associate professor at Chongqing University Engineering since 2012. In 2014, he was an Academic Visiting Fellow of Posts and Telecommunications. He received his Ph.D. degree (Honors) in with Princeton University, Princeton, NJ, USA. He leads a Research Group communication and information systems from the National Key Laboratory of focusing on wireless transmission and networking technologies with the State Science and Technology on Communications, University of Electronic Science Key Laboratory of Networking and Switching Technology, BUPT. He has and Technology of China in 2014. From April to December in 2012, he authored/coauthored over 100 refereed IEEE journal papers and over 300 was an international visitor at the Institute for Infocomm Research (I2R), conference proceeding papers. Dr. Peng was a recipient of the 2018 Heinrich Singapore. He was a research fellow at the National University of Singapore Hertz Prize Paper Award, the 2014 IEEE ComSoc AP Outstanding Young from July 2015 to July 2016. He served as a guest editor for IEEE Sensors Researcher Award, and the Best Paper Award in the JCN 2016 and IEEE Journal and IEEE Transactions on Industrial Informatics, he also served WCNC 2015. He is on the Editorial/Associate Editorial Board of the IEEE as symposium cochair for IEEE ICNC 2018, blockchain workshop cochair Communications Magazine, the IEEE Internet of Things Journal, and IEEE for CyberC 2019, IEEE Blockchain 2020 special session and TPC member Access. for numerous conferences. His research interests include blockchain system, internet of things and mobile edge computing.

Zixin Wang received the M.E degree in information and communication engineering from Chongqing University of Posts and Telecommunications, Chongqing, China, in 2020. She currently is pursuing her Ph.D. degree in Lei Zhang is a Senior Lecturer (Associate Professor) at the University of the State Key Laboratory of Networking and Switching Technology, Beijing Glasgow, U.K. He received his Ph.D. from the University of Sheffield, U.K. University of Posts and Telecommunications, Beijing, China. Her research His research interests include wireless communication systems and networks, interests include blockchain and Internet of Things. blockchain technology, radio access network slicing (RAN slicing), Internet of Things (IoT), multi-antenna signal processing, MIMO systems, etc. He has 19 patents granted/filed in more than 30 countries/regions including US/UK/EU/China/Japan etc. Dr Zhang has published 2 books and 100+ peer- reviewed papers. He received IEEE Communication Society TAOS TC Best Paper Award 2019. Dr. Zhang is a senior member of IEEE. He is a Technical Committee Chair of 5th International conference on UK-China Emerging Technologies (UCET) 2020. He was the Publication and Registration Chair of IEEE Sensor Array and Multichannel (SAM) 2018, Co-chair of Cyber- Long Zhang received the M.E degree in information and communication C Blockchain workshop 2019. He is an associate editor of IEEE Internet engineering from Chongqing University of Posts and Telecommunications, of Things (IoT) Journal, IEEE Wireless Communications Letters and Digital Chongqing, China, in 2019. He currently is pursuing his Ph.D. degree at the Communications and Networks. National Key Laboratory of Science and Technology on Communications, University of Electronic Science and Technology of China, Chengdu, China. His research areas include next generation mobile networks and Internet of Things.