See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/318131748

An Overview of Blockchain Technology: Architecture, Consensus, and Future Trends

Conference Paper · June 2017

DOI: 10.1109/BigDataCongress.2017.85

CITATIONS
READS
2,031
284,969

5 authors, including:

Zibin Zheng Shaoan Xie

Sun Yat-Sen University Sun Yat-Sen University
353 PUBLICATIONS 14,416 CITATIONS 7 PUBLICATIONS 4,211 CITATIONS

Hong-Ning Dai
Lingnan University
198 PUBLICATIONS 6,943 CITATIONS

Some of the authors of this publication are also working on these related projects:

Scalability of Blockchain View project

Blockchain technologies and applications View project

All content following this page was uploaded by Hong-Ning Dai on 06 October 2017.

The user has requested enhancement of the downloaded file.

2017 IEEE 6th International Congress on Big Data

An Overview of Blockchain Technology:

Architecture, Consensus, and Future Trends
Zibin Zheng1, Shaoan Xie1, Hongning Dai2, Xiangping Chen4, and Huaimin Wang3
1
School of Data and Computer Science, Sun Yat-sen University Guangzhou, China
2
Faculty of Information Technology, Macau University of Science and Technology, Macau, SAR
3
National Laboratory for Parallel & Distributed Processing
National University of Defense Technology, Changsha 410073 China
4
Institute of Advanced Technology,National Engineering Research Center of Digital Life
Sun Yat-sen University, Guangzhou, China
Email: [email protected]

Abstract—Blockchain, the foundation of Bitcoin, has received

extensive attentions recently. Blockchain serves as an immutable Things (IoT) [7], reputation systems [8] and security services
ledger which allows transactions take place in a decentralized [9]. Those fields favor blockchain in multiple ways. First of
manner. Blockchain-based applications are springing up, cov- all, blockchain is immutable. Transaction cannot be tampered
ering numerous fields including financial services, reputation once it is packed into the blockchain. Businesses that require
system and Internet of Things (IoT), and so on. However,
high reliability and honesty can use blockchain to attract
there are still many challenges of blockchain technology such
as scalability and security problems waiting to be overcome. customers. Besides, blockchain is distributed and can avoid
This paper presents a comprehensive overview on blockchain the single point of failure situation. As for smart contracts, the
technology. We provide an overview of blockchain architechture contract could be executed by miners automatically once the
firstly and compare some typical consensus algorithms used contract has been deployed on the blockchain.
in different blockchains. Furthermore, technical challenges and
Although the blockchain technology has great potential for
recent advances are briefly listed. We also lay out possible future
trends for blockchain. the construction of the future Internet systems, it is facing a
Index Terms—Blockchain, decentralization, consensus, scala- number of technical challenges. Firstly, scalability is a huge
bility concern. Bitcoin block size is limited to 1 MB now while
a block is mined about every ten minutes. Subsequently, the
I. INTRODUCTION Bitcoin network is restricted to a rate of 7 transactions per
second, which is incapable of dealing with high frequency
Nowadays cryptocurrency has become a buzzword in both
trading. However, larger blocks means larger storage space
industry and academia. As one of the most successful cryp-
and slower propagation in the network. This will lead to
tocurrency, Bitcoin has enjoyed a huge success with its capital
centralization gradually as less users would like to maintain
market reaching 10 billion dollars in 2016 [1]. With a spe-
such a large blockchain. Therefore the tradeoff between block
cially designed data storage structure, transactions in Bitcoin
size and security has been a tough challenge. Secondly, it has
network could happen without any third party and the core
been proved that miners could achieve larger revenue than
technology to build Bitcoin is blockchain, which was first
their fair share through selfish mining strategy [10]. Miners
proposed in 2008 and implemented in 2009 [2]. Blockchain
hide their mined blocks for more revenue in the future. In
could be regarded as a public ledger and all committed
that way, branches could take place frequently, which hinders
transactions are stored in a list of blocks. This chain grows
blockchain development. Hence some solutions need to be put
as new blocks are appended to it continuously. Asymmetric
forward to fix this problem. Moreover, it has been shown that
cryptography and distributed consensus algorithms have been
privacy leakage could also happen in blockchain even users
implemented for user security and ledger consistency. The
only make transactions with their public key and private key
blockchain technology generally has key characteristics of
[11]. Furthermore, current consensus algorithms like proof of
decentralization, persistency, anonymity and auditability.
work or proof of stake are facing some serious problems. For
With these traits, blockchain can greatly save the cost and
example, proof of work wastes too much electricity energy
improve the efficiency.
while the phenomenon that the rich get richer could appear in
Since it allows payment to be finished without any bank or
the proof of stake consensus process.
any intermediary, blockchain can be used in various financial
There is a lot of literature on blockchain from various
services such as digital assets, remittance and online payment
sources, such as blogs, wikis, forum posts, codes, confer-
[3], [4]. Additionally, it can also be applied into other fields
ence proceedings and journal articles. Tschorsch et al. [12]
including smart contracts [5], public services [6], Internet of
made a technical survey about decentralized digital currencies

978-1-5386-1996-4/17 $31.00 © 2017 IEEE

DOI 10.1109/BigDataCongress.2017.85
557
including Bitcoin. Compared to [12], our paper focuses on (ii) Merkle tree root hash: the hash value of all the transac-
blockchain technology instead of digital currencies. Nomura tions in the block.
Research Institut made a technical report about blockchain (iii) Timestamp: current time as seconds in universal time
[13]. Contrast to [13], our paper focuses on state-of-art since January 1, 1970.
blockchain researches including recent advances and future (iv) nBits: target threshold of a valid block hash.
trends. (v) Nonce: an 4-byte field, which usually starts with 0
The rest of this paper is organized as follows. Section II and increases for every hash calculation (will be explained
introduces blockchain architecture. Section III shows typical in details in Section III).
consensus algorithms used in blockchain. Section IV summa- (vi) Parent block hash: a 256-bit hash value that points to
rizes the technical challenges and the recent advances in this the previous block.
area. Section V discusses some possible future directions and The block body is composed of a transaction counter and
section VI concludes the paper. transactions. The maximum number of transactions that a
block can contain depends on the block size and the size of
II. BLOCKCHAIN ARCHITECTURE
each transaction. Blockchain uses an asymmetric
cryptography mechanism to validate the authentication of
transactions [13]. Digital signature based on asymmetric
Block Header
Parent Block Hash
a
Block Header
Parent Block Hash
a
Block Header
Parent Block Hash
a
cryptography is used in an untrustworthy environment. We next
briefly illustrate digital signature.
Transaction Counter Transaction Counter Transaction Counter

TXTXTX TXTXTX TXTXTX

B. Digital Signature
Each user owns a pair of private key and public key.
Block i-1 Block i Block i+1 The private key that shall be kept in confidentiality is used
to sign the transactions. The digital signed transactions are
Fig. 1: An example of blockchain which consists of a contin- broadcasted throughout the whole network. The typical digital
uous sequence of blocks. signature is involved with two phases: signing phase and
verification phase. For instance, an user Alice wants to send
another user Bob a message. (1) In the signing phase, Alice
encrypts her data with her private key and sends Bob the
Block Header encrypted result and original data. (2) In the verification
Merkle Tree Root Hash
TimeParent
phase, Bob validates the value with Alice’s public key. In that
Block Version e
stampnBitsNonceBlock
Hash
way, Bob could easily check if the data has been tampered or
not. The typical digital signature algorithm used in
blockchains is the elliptic curve digital signature algorithm
(ECDSA) [16].
Transaction Counter C. Key Characteristics of Blockchain
TXTXTXTXTXTX
In summary, blockchain has following key characteristics.
• Decentralization. In conventional centralized transaction

Fig. 2: Block structure systems, each transaction needs to be validated through

the central trusted agency (e.g., the central bank), in-
evitably resulting to the cost and the performance bottle-
Blockchain is a sequence of blocks, which holds a complete
necks at the central servers. Contrast to the centralized
list of transaction records like conventional public ledger
mode, third party is no longer needed in blockchain.
[14]. Figure 1 illustrates an example of a blockchain. With
Consensus algorithms in blockchain are used to maintain
a previous block hash contained in the block header, a block
data consistency in distributed network.
has only one parent block. It is worth noting that uncle blocks
• Persistency. Transactions can be validated quickly and
(children of the block’s ancestors) hashes would also be
invalid transactions would not be admitted by honest
stored in ethereum blockchain [15]. The first block of a
miners. It is nearly impossible to delete or rollback
blockchain is called genesis block which has no parent block.
transactions once they are included in the blockchain.
We then explain the internals of blockchain in details.
Blocks that contain invalid transactions could be discov-
A. Block ered immediately.
• Anonymity. Each user can interact with the blockchain
A block consists of the block header and the block body as with a generated address, which does not reveal the
shown in Figure 2. In particular, the block header includes: real identity of the user. Note that blockchain cannot
(i) Block version: indicates which set of block validation guarantee the perfect privacy preservation due to the
rules to follow. intrinsic constraint (details will be discussed in section
IV).

5
 TABLE I: Comparisons among public blockchain, consortium blockchain and private blockchain
Property Public blockchain Consortium blockchain Private blockchain
Consensus determination All miners Selected set of nodes One organization

Read permission Public Could be public or restricted Could be public or restricted

Immutability Nearly impossible to tamper Could be tampered Could be tampered

Efficiency Low High High

Centralized No Partial Yes

Consensus process Permissionless Permissioned Permissioned

• Auditability. Bitcoin blockchain stores data about user

validators, consortium blockchain and private blockchain
balances based on the Unspent Transaction Output
could be more efficient.
(UTX- O) model [2]: Any transaction has to refer to
• Centralized. The main difference among the three types
some previ- ous unspent transactions. Once the current
of blockchains is that public blockchain is decentralized,
transaction is recorded into the blockchain, the state of
consortium blockchain is partially centralized and private
those referred unspent transactions switch from unspent
blockchain is fully centralized as it is controlled by a
to spent. So transactions could be easily verified and
single group.
tracked.
• Consensus process. Everyone in the world could join
the consensus process of the public blockchain. Different
D. Taxonomy of blockchain systems from public blockchain, both consortium blockchain and
Current blockchain systems are categorized roughly into private blockchain are permissioned.
three types: public blockchain, private blockchain and con- Since public blockchain is open to the world, it can at-
sortium blockchain [17]. In public blockchain, all records are tract many users and communities are active. Many public
visible to the public and everyone could take part in the con- blockchains emerge day by day. As for consortium
sensus process. Differently, only a group of pre-selected blockchain, it could be applied into many business
nodes would participate in the consensus process of a applications. Cur- rently Hyperledger [18] is developing
consortium blockchain. As for private blockchain, only those business consortium blockchain frameworks. Ethereum also
nodes that come from one specific organization would be has provided tools for building consortium blockchains [19].
allowed to join the consensus process.
III. CONSENSUS ALGORITHMS
A private blockchain is regarded as a centralized network
since it is fully controlled by one organization. The In blockchain, how to reach consensus among the untrust-
consortium blockchain constructed by several organizations is worthy nodes is a transformation of the Byzantine Generals
partially decentralized since only a small portion of nodes (BG) Problem, which was raised in [20]. In BG problem,
would be selected to determine the consensus. The a group of generals who command a portion of Byzantine
comparison among the three types of blockchains is listed in army circle the city. Some generals prefer to attack while
Table I. other generals prefer to retreat. However, the attack would
fail if only part of the generals attack the city. Thus, they
• Consensus determination. In public blockchain, each n-
have to reach an agreement to attack or retreat. How to reach
ode could take part in the consensus process. And only
a consensus in distributed environment is a challenge. It is
a selected set of nodes are responsible for validating the
also a challenge for blockchain as the blockchain network
block in consortium blockchain. As for private chain, it
is distributed. In blockchain, there is no central node that
is fully controlled by one organization and the
ensures ledgers on distributed nodes are all the same. Some
organization could determine the final consensus.
protocols are needed to ensure ledgers in different nodes are
• Read permission. Transactions in a public blockchain are
consistent. We next present several common approaches to
visible to the public while it depends when it comes to a
reach a consensus in blockchain.
private blockchain or a consortium blockchain.
• Immutability. Since records are stored on a large number A. Approaches to consensus
of participants, it is nearly impossible to tamper trans-
PoW (Proof of work) is a consensus strategy used in the
actions in a public blockchain. Differently, transactions
Bitcoin network [2]. In a decentralized network, someone has
in a private blockchain or a consortium blockchain could
to be selected to record the transactions. The easiest way is
be tampered easily as there are only limited number of
random selection. However, random selection is vulnerable to
participants.
attacks. So if a node wants to publish a block of transactions,
• Efficiency. It takes plenty of time to propagate transac-
a lot of work has to be done to prove that the node is not
tions and blocks as there are a large number of nodes
likely to attack the network. Generally the work means
on public blockchain network. As a result, transaction
computer
throughput is limited and the latency is high. With fewer

5
 TABLE II: Typical Consensus Algorithms Comparison
Property PoW PoS PBFT DPOS Ripple Tendermint
Node identity management open open permissioned open open permissioned

Energy saving no partial yes partial yes yes

Tolerated power < 25% < 51% < 33.3% < 51% < 20% < 33.3%
of adversary computing stake faulty replicas validators faulty nodes in byzantine voting
power UNL power

Example Bitcoin [2] Peercoin [21] Hyperledger Bitshares [22] Ripple [23] Tendermint [24]
Fabric [18]

used for mathematical research.

PoS (Proof of stake) is an energy-saving alternative to
U4 PoW. Miners in PoS have to prove the ownership of the
amount of currency. It is believed that people with more
Shorter Branch currencies would be less likely to attack the network. The
B1 B2
B3
selection based on account balance is quite unfair because the
single richest person is bound to be dominant in the network.
As a result, many solutions are proposed with the combination
B4 B5 of the stake size to decide which one to forge the next
block. In particular, Blackcoin [26] uses randomization to
Longer Branch
predict the next generator. It uses a formula that looks for
the lowest hash value in combination with the size of the
Fig. 3: An scenario of blockchain branches (the longer branch stake. Peercoin
would be admitted as the main chain while the shorter one [21] favors coin age based selection. In Peercoin, older and
would be deserted) larger sets of coins have a greater probability of mining the
next block. Compared to PoW, PoS saves more energy and
is more effective. Unfortunately, as the mining cost is nearly
calculations. In PoW, each node of the network is calculating zero, attacks might come as a consequence. Many blockchains
a hash value of the block header. The block header contains adopt PoW at the beginning and transform to PoS gradually.
a nonce and miners would change the nonce frequently to For instance, ethereum is planing to move from Ethash (a kind
get different hash values. The consensus requires that the of PoW) [27] to Casper (a kind of PoS) [28].
calculated value must be equal to or smaller than a certain PBFT (Practical byzantine fault tolerance) is a replication
given value. When one node reaches the target value, it would algorithm to tolerate byzantine faults [29]. Hyperledger Fabric
broadcast the block to other nodes and all other nodes must [18] utilizes the PBFT as its consensus algorithm since PBFT
mutually confirm the correctness of the hash value. If the could handle up to 1/3 malicious byzantine replicas. A new
block is validated, other miners would append this new block is determined in a round. In each round, a primary
block to their own blockchains. Nodes that calculate the hash would be selected according to some rules. And it is
values are called miners and the PoW procedure is called responsible for ordering the transaction. The whole process
mining in Bitcoin. could be divided into three phase: pre-prepared, prepared
In the decentralized network, valid blocks might be gen- and commit. In each phase, a node would enter next phase if
erated simultaneously when multiple nodes find the suitable it has received votes from over 2/3 of all nodes. So PBFT
nonce nearly at the same time. As a result, branches may be requires that every node is known to the network. Like
generated as shown in Figure 3. However, it is unlikely that PBFT, Stellar Consensus Protocol (SCP) [30] is also a
two competing forks will generate next block simultaneously. Byzantine agreement protocol. In PBFT, each node has to
In PoW protocol, a chain that becomes longer thereafter is query other nodes while SCP gives participants the right to
judged as the authentic one. Consider two forks created by choose which set of other participants to believe. Based on
simultaneously validated blocks U4 and B4. Miners keep PBFT, Antshares [31] has implemented their dBFT
mining their blocks until a longer branch is found. B4,B5 (delegated byzantine fault tolerance). In dBFT, some
forms a longer chain, so the miners on U4 would switch to professional nodes are voted to record the transactions.
the longer branch. DPOS (Delegated proof of stake). The major difference
Miners have to do a lot of computer calculations in PoW, between PoS and DPOS is that PoS is direct democratic while
yet these works waste too much resources. To mitigate the DPOS is representative democratic. Stakeholders elect their
loss, some PoW protocols in which works could have some delegates to generate and validate blocks. With significantly
side-applications have been designed. For example, Primecoin fewer nodes to validate the block, the block could be
[25] searches for special prime number chains which can be confirmed quickly, leading to the quick confirmation of
transactions. Meanwhile, the parameters of the network such
as block size and block intervals could be tuned by delegates.
Additionally,

5
users need not to worry about the dishonest delegates as they • Example. Bitcoin is based on PoW while Peercoin is
could be voted out easily. DPOS is the backbone of Bitshares a new peer-to-peer PoS cryptocurrency. Further, Hyper-
[22]. ledger Fabric utilizes PBFT to reach consensus.
Ripple [23] is a consensus algorithm that utilizes Bitshares, a smart contract platform, adopts DPOS as
collectively-trusted subnetworks within the larger network. In their con- sensus algorithm. Ripple implements the
the network, nodes are divided into two types: server for Ripple protocol while Tendermint devises the
participating consensus process and client for only Tendermint protocol.
transferring funds. Each server has an Unique Node List
PBFT and Tendermint are permissioned protocols. Node
(UNL). UNL is important to the server. When determining
identities are expected to be known to the whole network,
whether to put a transaction into the ledger, the server would
so they might be used in commercial mode rather than public.
query the nodes in UNL and if the received agreements have
PoW and PoS are suitable for public blockchain. Consortium
reached 80%, the transaction would be packed into the ledger.
or private blockchain might has preference for PBFT, Tender-
For a node, the ledger will remain correct as long as the
mint, DPOS and Ripple.
percentage of faulty nodes in UNL is less than 20%.
Tendermint [24] is a byzantine consensus algorithm. A C. Advances on consensus algorithms
new block is determined in a round. A proposer would be
A good consensus algorithm means efficiency, safty and
selected to broadcast an unconfirmed block in this round. It
convenience. Recently, a number of endeavors have been
could be divided into three steps: 1) Prevote step. Validators
made to improve consensus algorithms in blockchain. New
choose whether to broadcast a prevote for the proposed block.
con- sensus algorithms are devised aiming to solve some
2) Precommit step. If the node has received more than 2/3 of
specific problems of blockchain. The main idea of PeerCensus
prevotes on the proposed block, it broadcasts a precommit for
[33] is to decouple block creation and transaction
that block. If the node has received over 2/3 of precommits,
confirmation so that the consensus speed can be significantly
it enters the commit step. 3) Commit step. The node validates
increased. Besides, Kraft [34] proposed a new consensus
the block and broadcasts a commit for that block. if the
method to ensure that a block is generated in a relatively
node has received 2/3 of the commits, it accepts the block.
stable speed. It is known that high blocks generation rate
Contrast to PBFT, nodes have to lock their coins to become
compromise Bitcoin’s security. So the Greedy Heaviest-
validators. Once a validator is found to be dishonest, it would
Observed Sub-Tree (GHOST) chain selection rule [35] is
be punished.
proposed to solve this problem. Instead of the longest branch
B. Consensus algorithms comparison scheme, GHOST weights the branches and miners could
Different consensus algorithms have different advantages choose the better one to follow. Chepurnoy et al. [36]
and disadvantages. Table II gives a comparison between d- presented a new consensus algorithm for peer-to- peer
ifferent consensus algorithms and we use the properties given blockchain systems where anyone who provides non-
by [32]. interactive proofs of retrievability for the past state snapshots
is agreed to generate the block. In such a protocol, miners
• Node identity management. PBFT needs to know the
only have to store old block headers instead of full blocks.
identity of each miner in order to select a primary in
every round while Tendermint needs to know the IV. CHALLENGES & RECENT ADVANCES
validators in order to select a proposer in each round. For
Despite the great potential of blockchain, it faces numerous
PoW, PoS, DPOS and Ripple, nodes could join the
challenges, which limit the wide usage of blockchain. We
network freely.
enumerate some major challenges and recent advances as
• Energy saving. In PoW, miners hash the block header
follows.
continuously to reach the target value. As a result, the
amount of electricity required to process has reach an A. Scalability
immense scale. As for PoS and DPOS, miners still have With the amount of transactions increasing day by day,
to hash the block header to search the target value but the blockchain becomes bulky. Each node has to store all
the work has been largely reduced as the search space transactions to validate them on the blockchain because they
is designed to be limited. As for PBFT, Ripple and have to check if the source of the current transaction is
Tendermint, there is no mining in consensus process. So unspent or not. Besides, due to the original restriction of block
it saves energy greatly. size and the time interval used to generate a new block, the
• Tolerated power of adversary. Generally 51% of hash
Bitcoin blockchain can only process nearly 7 transactions per
power is regarded as the threshold for one to gain control second, which cannot fulfill the requirement of processing
of the network. But selfish mining strategy [10] in PoW millions of transactions in real-time fashion. Meanwhile, as
systems could help miners to gain more revenue by only the capacity of blocks is very small, many small transactions
25% of the hashing power. PBFT and Tendermint is might be delayed since miners prefer those transactions with
designed to handle up to 1/3 faulty nodes. Ripple is high transaction fee.
proved to maintain correctness if the faulty nodes in an There are a number of efforts proposed to address the
UNL is less than 20%. scalability problem of blockchain, which could be categorized
into two types:

5
 • Storage optimization of blockchain. Since it is harder for becomes harder to reveal relationship between Alice and
node to operate full copy of ledger, Bruce proposed a Bob. However, the intermediary could be dishonest and
novel cryptocurrency scheme, in which the old reveal Alice and Bob’s private information on purpose.
transaction records are removed (or forgotten) by the It is also possible that Carol transfers Alice’s funds to
network [37]. A database named account tree is used to her own address instead of Bob’s address. Mixcoin [43]
hold the balance of all non-empty addresses. Besides provides a simple method to avoid dishonest behaviours.
lightweight client could also help fix this problem. A The intermediary encrypts users’ requirements including
novel schem named VerSum [38] was proposed to funds amount and transfer date with its private key. Then
provide another way allowing lightweight clients to exist. if the intermediary did not transfer the money, anybody
VerSum allows lightweight clients to outsource could verify that the intermediary cheated. However,
expensive computations over large inputs. It ensures the theft is detected but still not prevented. Coinjoin [44]
computation result is correct through comparing results depends on a central mixing server to shuffle output
from multiple servers. addresses to prevent theft. And inspired by Coinjoin,
• Redesigning blockchain. In [39], Bitcoin-NG (Next Gen- CoinShuffle [45] uses decryption mixnets for address
eration) was proposed. The main idea of Bitcoin-NG is shuffling.
to decouple conventional block into two parts: key block • Anonymous. In Zerocoin [46], zero-knowledge proof
for leader election and microblock to store transactions. is used. Miners do not have to validate a transaction
The protocol divides time into epoches. In each epoch, with digital signature but to validate coins belong to
miners have to hash to generate a key block. Once the a list of valid coins. Payment’s origin are unlinked
key block is generated, the node becomes the leader who from transactions to prevent transaction graph analyses.
is responsible for generating microblocks. Bitcoin-NG But it still reveals payments’ destination and amounts.
also extended the heaviest (longest) chain strategy in Zerocash [47] was proposed to address this problem.
which microblocks carry no weight. In this way, In Zerocash, zero-knowledge Succinct Non-interactive
blockchain is redesigned and the tradeoff between block Arguments of Knowledge (zk-SNARKs) is leveraged.
size and network security has been addressed. Transaction amounts and the values of coins held by
B. Privacy Leakage users are hidden.
Blockchain can preserve a certain amount of privacy C. Selfish Mining
through the public key and private key. Users transact with
Blockchain is susceptible to attacks of colluding selfish
their private key and public key without any real identity
miners. In particular, Eyal and Sirer [10] showed that the
exposure. However, it is shown in [40], [5] that blockchain
network is vulnerable even if only a small portion of the
cannot guarantee the transactional privacy since the values of
hashing power is used to cheat. In selfish mining strategy,
all transactions and balances for each public key are publicly
selfish miners keep their mined blocks without broadcasting
visible. Besides, the recent study [41] has shown that a user’s
and the private branch would be revealed to the public only
Bitcoin transactions can be linked to reveal user’s
if some requirements are satisfied. As the private branch is
information. Moreover, Biryukov et al. [11] presented an
longer than the current public chain, it would be admitted
method to link user pseudonyms to IP addresses even when
by all miners. Before the private blockchain publishment,
users are behind Network Address Translation (NAT) or
honest miners are wasting their resources on an useless branch
firewalls. In [11], each client can be uniquely identified by a
while selfish miners are mining their private chain without
set of nodes it connects to. However, this set can be
competitors. So selfish miners tend to get more revenue.
learned and used to find the origin of a transaction.
Based on selfish mining, many other attacks have been
Multiple methods have been proposed to improve anonymity
proposed to show that blockchain is not so secure. In stubborn
of blockchain, which could be roughly categorized into two
mining [48], miners could amplify its gain by non-trivially
types:
composing mining attacks with network-level eclipse attacks.
• Mixing [42]. In blockchain, users addresses are
The trail-stubbornness is one of the stubborn strategy that
pseudony- mous. But it is still possible to link addresses miners still mine the blocks even if the private chain is left
to user real identity as many users make transactions behind. Yet in some cases, it can result in 13% gains in
with the same address frequently. Mixing service is a comparison with a non-trail-stubborn counterpart. [49] shows
kind of service which provides anonymity by transferring that there are selfish mining strategies that earn more money
funds from multiple input addresses to multiple output and are profitable for smaller miners compared to simple
addresses. For example, user Alice with address A wants selfish mining. But the gains are relatively small.
to send some funds to Bob with address B. If Alice Furthermore, it shows that attackers with less than 25% of the
directly makes a transaction with input address A and computational resources can still gain from selfish mining. To
output address B, relationship between Alice and Bob help fix the selfish mining problem, Heilman [50] presented
might be revealed. So Alice could send funds to a trusted an novel approach for honest miners to choose which branch
intermediary Carol. Then Carol transfer funds to Bob to follow. With random beacons and timestamps, honest
with multiple inputs c1, c2, c3, etc., and multiple output miners would select more fresh blocks. However, [50] is
d1, d2, B, d3, etc. Bob’s address B is also contained in vulnerable to
the output addresses. So it

5
forgeable timestamps. ZeroBlock [51] builds on the simple be extracted. Users can predict their potential partners’ trading
scheme: Each block must be generated and accepted by the behaviours with the analysis.
network within a maximum time interval. Within ZeroBlock,
selfish miners cannot achieve more than its expected reward. D. Blockchain applications
Currently most blockchains are used in the financial
V. POSSIBLE FUTURE DIRECTIONS
domain, more and more applications for different fields are
Blockchain has shown its potential in industry and appearing. Traditional industries could take blockchain into
academi- considera- tion and apply blockchain into their fields to
a. We discuss possible future directions with respect to four enhance their systems. For example, user reputations could be
areas: blockchain testing, stop the tendency to centralization, stored on blockchain. At the same time, the up-and-
big data analytics and blockchain application. coming industry could make use of blockchain to improve
performance. For example, Arcade City [51], a ridesharing
A. Blockchain testing
startup offers an open marketplace where riders connect
Recently different kinds of blockchains appear and over directly with drivers by leveraging blockchain technology.
700 cryptocurrencies are listed in [52] up to now. However, A smart contract is a computerized transaction protocol that
some developers might falsify their blockchain performance executes the terms of a contract [54]. It has been proposed
to attract investors driven by the huge profit. Besides that, for long time and now this concept can be implemented with
when users want to combine blockchain into business, they blockchain. In blockchain, smart contract is a code fragment
have to know which blockchain fits their requirements. So that could be executed by miners automatically. Smart
blockchain testing mechanism needs to be in place to test contract has transformative potential in various fields like
different blockchains. financial services and IoT.
Blockchain testing could be separated into two phases:
standardization phase and testing phase. In standardization VI. CONCLUSION
phase, all criteria have to be made and agreed. When a Blockchain has shown its potential for transforming tradi-
blockchain is born, it could be tested with the agreed criteria tional industry with its key characteristics: decentralization,
to valid if the blockchain works fine as developers claim. As persistency, anonymity and auditability. In this paper, we
for testing phase, blockchain testing needs to be performed present a comprehensive overview on blockchain. We first
with different criteria. For example, an user who is in charge give an overview of blockchain technologies including
of online retail business cares about the throughput of the blockchain architecture and key characteristics of blockchain.
blockchain, so the examination needs to test the average time We then dis- cuss the typical consensus algorithms used in
from a user send a transaction to the transaction is packed into blockchain. We analyzed and compared these protocols in
the blockchain, capacity for a blockchain block and etc. different respects. Furthermore, we listed some challenges and
problems that would hinder blockchain development and
B. Stop the tendency to centralization summarized some existing approaches for solving these
Blockchain is designed as a decentralized system. problems. Some possible future directions are also proposed.
However, there is a trend that miners are centralized in the Nowadays blockchain- based applications are springing up
mining pool. Up to now, the top 5 mining pools together owns and we plan to conduct in-depth investigations on blockchain-
larger than 51% of the total hash power in the Bitcoin network based applications in the future.
[53]. Apart from that, selfish mining strategy [10] showed that ACKNOWLEDGEMENT
pools with over 25% of total computing power could get more
revenue than fair share. Rational miners would be attracted The work described in this paper was supported by the
into the selfish pool and finally the pool could easily exceed National Key Research and Development Program (2016YF-
51% of the total power. As the blockchain is not intended to B1000101), the National Natural Science Foundation of
serve a few organizations, some methods should be proposed China under (61472338), the Fundamental Research Funds
to solve this problem. for the Central Universities and Macao Science and
Technology De- velopment Fund under Grant No.
C. Big data analytics 096/2013/A3. The authors would like to thank Gordon K.-T.
Blockchain could be well combined with big data. Here Hon for his constructive comments.
we roughly categorized the combination into two types: data REFERENCES
management and data analytics. As for data management, [1] “State of blockchain q1 2016: Blockchain funding overtakes
blockchain could be used to store important data as it is bitcoin,” 2016. [Online]. Available: http://www.coindesk.com/ state-of-
distributed and secure. Blockchain could also ensure the data blockchain-q1-2016/
is original. For example, if blockchain is used to store patients [2] S. Nakamoto, “Bitcoin: A peer-to-peer electronic cash system,” 2008.
[Online]. Available: https://bitcoin.org/bitcoin.pdf
health information, the information could not be tampered and [3] G. W. Peters, E. Panayi, and A. Chapelle, “Trends in crypto-currencies
it is hard to stole those private information. When it comes to and blockchain technologies: A monetary theory and regulation
data analytics, transactions on blockchain could be used for perspective,” 2015. [Online]. Available: http://dx.doi.org/10.2139/ssrn.
2646618
big data analytics. For example, user trading patterns might

5
 [4] G. Foroglou and A.-L. Tsilidou, “Further applications of the
[31] “Antshares digital assets for everyone,” 2016. [Online]. Available:
blockchain,” 2015.
https://www.antshares.org
[5] A. Kosba, A. Miller, E. Shi, Z. Wen, and C. Papamanthou, “Hawk:
[32] M. Vukolic´, “The quest for scalable blockchain fabric: Proof-of-
The blockchain model of cryptography and privacy-preserving smart
work vs. bft replication,” in International Workshop on Open Problems
contracts,” in Proceedings of IEEE Symposium on Security and Privacy
in Network Security, Zurich, Switzerland, 2015, pp. 112–125.
(SP), San Jose, CA, USA, 2016, pp. 839–858. [33] C. Decker, J. Seidel, and R. Wattenhofer, “Bitcoin meets strong con-
[6] B. W. Akins, J. L. Chapman, and J. M. Gordon, “A whole new world: sistency,” in Proceedings of the 17th International Conference on Dis-
Income tax considerations of the bitcoin economy,” 2013. [Online]. tributed Computing and Networking (ICDCN). Singapore, Singapore:
Available: https://ssrn.com/abstract=2394738 ACM, 2016, p. 13.
[7] Y. Zhang and J. Wen, “An iot electric business model based on the [34] D. Kraft, “Difficulty control for blockchain-based consensus systems,”
protocol of bitcoin,” in Proceedings of 18th International Conference on Peer-to-Peer Networking and Applications, vol. 9, no. 2, pp. 397–413,
Intelligence in Next Generation Networks (ICIN), Paris, France, 2015, 2016.
pp. 184–191. [35] Y. Sompolinsky and A. Zohar, “Accelerating bitcoin’s transaction pro-
[8] M. Sharples and J. Domingue, “The blockchain and kudos: A distributed cessing. fast money grows on trees, not chains.” IACR Cryptology ePrint
system for educational record, reputation and reward,” in Proceedings of Archive, vol. 2013, no. 881, 2013.
11th European Conference on Technology Enhanced Learning (EC-TEL [36] A. Chepurnoy, M. Larangeira, and A. Ojiganov, “A prunable blockchain
2015), Lyon, France, 2015, pp. 490–496. consensus protocol based on non-interactive proofs of past states retriev-
[9] C. Noyes, “Bitav: Fast anti-malware by distributed blockchain ability,” arXiv preprint arXiv:1603.07926, 2016.
consensus and feedforward scanning,” arXiv preprint [37] J. Bruce, “The mini-blockchain scheme,” July 2014. [Online]. Available:
arXiv:1601.01405, 2016. http://cryptonite.info/files/mbc-scheme-rev3.pdf
[10] I. Eyal and E. G. Sirer, “Majority is not enough: Bitcoin mining is [38] J. van den Hooff, M. F. Kaashoek, and N. Zeldovich, “Versum:
vulnerable,” in Proceedings of International Conference on Financial Verifiable computations over large public logs,” in Proceedings of the
Cryptography and Data Security, Berlin, Heidelberg, 2014, pp. 436– 2014 ACM SIGSAC Conference on Computer and Communications
454. Security, New York, NY, USA, 2014, pp. 1304–1316.
[11] A. Biryukov, D. Khovratovich, and I. Pustogarov, “Deanonymisation [39] I. Eyal, A. E. Gencer, E. G. Sirer, and R. Van Renesse, “Bitcoin-
of clients in bitcoin p2p network,” in Proceedings of the 2014 ACM ng: A scalable blockchain protocol,” in Proceedings of 13th USENIX
SIGSAC Conference on Computer and Communications Security, New Symposium on Networked Systems Design and Implementation (NSDI
York, NY, USA, 2014, pp. 15–29. 16), Santa Clara, CA, USA, 2016, pp. 45–59.
[12] F. Tschorsch and B. Scheuermann, “Bitcoin and beyond: A technical [40] S. Meiklejohn, M. Pomarole, G. Jordan, K. Levchenko, D. McCoy, G.
survey on decentralized digital currencies,” IEEE Communications Sur- M. Voelker, and S. Savage, “A fistful of bitcoins: Characterizing
veys Tutorials, vol. 18, no. 3, pp. 2084–2123, 2016. payments among men with no names,” in Proceedings of the 2013
[13] NRI, “Survey on blockchain technologies and related services,” Tech. Conference on Internet Measurement Conference (IMC’13), New York,
Rep., 2015. [Online]. Available: http://www.meti.go.jp/english/press/ NY, USA, 2013.
2016/pdf/0531 01f.pdf [41] J. Barcelo, “User privacy in the public bitcoin blockchain,” 2014.
[14] D. Lee Kuo Chuen, Ed., Handbook of Digital Currency, 1st ed. [42] M. Mo¨ser, “Anonymity of bitcoin transactions: An analysis of
Elsevier, 2015. [Online]. Available: http://EconPapers.repec.org/RePEc: mixing services,” in Proceedings of Mu¨nster Bitcoin Conference,
eee:monogr:9780128021170 Mu¨nster, Germany, 2013, pp. 17–18.
[15] V. Buterin, “A next-generation smart contract and decentralized appli- [43] J. Bonneau, A. Narayanan, A. Miller, J. Clark, J. A. Kroll, and E. W.
cation platform,” white paper, 2014. Felten, “Mixcoin: Anonymity for bitcoin with accountable mixes,” in
[16] D. Johnson, A. Menezes, and S. Vanstone, “The elliptic curve digital Proceedings of International Conference on Financial Cryptography
signature algorithm (ecdsa),” International Journal of Information Se- and Data Security, Berlin, Heidelberg, 2014, pp. 486–504.
curity, vol. 1, no. 1, pp. 36–63, 2001. [44] G. Maxwell, “Coinjoin: Bitcoin privacy for the real world,” in Post on
[17] V. Buterin, “On public and private blockchains,” Bitcoin Forum, 2013.
2015. [Online]. Available: https://blog.ethereum.org/2015/08/07/ on- [45] T. Ruffing, P. Moreno-Sanchez, and A. Kate, “Coinshuffle: Practical
public-and-private-blockchains/ decentralized coin mixing for bitcoin,” in Proceedings of European
[18] “Hyperledger project,” 2015. [Online]. Available: https://www. Symposium on Research in Computer Security, Cham, 2014, pp. 345–
hyperledger.org/ 364.
[19] “Consortium chain development.” [Online]. Available: https://github. [46] I. Miers, C. Garman, M. Green, and A. D. Rubin, “Zerocoin:
com/ethereum/wiki/wiki/Consortium-Chain-Development Anonymous distributed e-cash from bitcoin,” in Proceedings of IEEE
[20] L. Lamport, R. Shostak, and M. Pease, “The byzantine generals prob- Symposium Security and Privacy (SP), Berkeley, CA, USA, 2013, pp.
lem,” ACM Transactions on Programming Languages and Systems 397–411.
(TOPLAS), vol. 4, no. 3, pp. 382–401, 1982. [47] E. B. Sasson, A. Chiesa, C. Garman, M. Green, I. Miers, E. Tromer, and
[21] S. King and S. Nadal, “Ppcoin: Peer-to-peer crypto-currency with proof- M. Virza, “Zerocash: Decentralized anonymous payments from bitcoin,”
of-stake,” Self-Published Paper, August, vol. 19, 2012. in Proceedings of 2014 IEEE Symposium on Security and Privacy (SP),
[22] “Bitshares - your share in the decentralized exchange.” [Online]. San Jose, CA, USA, 2014, pp. 459–474.
Available: https://bitshares.org/ [48] K. Nayak, S. Kumar, A. Miller, and E. Shi, “Stubborn mining:
Generaliz- ing selfish mining and combining with an eclipse attack,” in
[23] D. Schwartz, N. Youngs, and A. Britto, “The ripple protocol consensus
Proceedings of 2016 IEEE European Symposium on Security and
algorithm,” Ripple Labs Inc White Paper, vol. 5, 2014.
Privacy (EuroS&P), Saarbrucken, Germany, 2016, pp. 305–320.
[24] J. Kwon, “Tendermint: Consensus without mining,” URL
[49] A. Sapirshtein, Y. Sompolinsky, and A. Zohar, “Optimal selfish mining
http://tendermint. com/docs/tendermint v04. pdf, 2014.
{ strategies in bitcoin,” arXiv preprint arXiv:1507.06183, 2015.
[25] S. King, “Primecoin: Cryptocurrency with prime number proof-of- [50] S. Billah, “One weird trick to stop selfish miners: Fresh bitcoins, a
work,” July 7th, 2013. solution for the honest miner,” 2015.
[26] P. Vasin, “Blackcoins proof-of-stake protocol v2,” 2014. [Online]. [51] S. Solat and M. Potop-Butucaru, “ZeroBlock: Timestamp-Free
Avail- able: https://blackcoin.co/blackcoin-pos-protocol-v2- Prevention of Block-Withholding Attack in Bitcoin,” Sorbonne
whitepaper.pdf Universites, UPMC University of Paris 6, Technical Report, May 2016.
[27] G. Wood, “Ethereum: A secure decentralised generalised transaction [Online]. Available: https://hal.archives-ouvertes.fr/hal-01310088
ledger,” Ethereum Project Yellow Paper, 2014. [52] “Crypto-currency market capitalizations,” 2017. [Online]. Available:
[28] V. Zamfir, “Introducing casper the friendly ghost,” Ethereum https://coinmarketcap.com
Blog URL: https://blog. ethereum. org/2015/08/01/introducing-casper- [53] “The biggest mining pools.” [Online]. Available: https:
friendly-ghost, 2015. //bitcoinworldwide.com/mining/pools/
[29] C. Miguel and L. Barbara, “Practical byzantine fault tolerance,” in [54] N. Szabo, “The idea of smart contracts,” 1997.
Proceedings of the Third Symposium on Operating Systems Design and
Implementation, vol. 99, New Orleans, USA, 1999, pp. 173–186.
[30] D. Mazieres, “The stellar consensus protocol: A federated model for
internet-level consensus,” Stellar Development Foundation, 2015.

564
View publication stats