Professional Documents
Culture Documents
IN INTELLIGENT TRANSPORTATION
Received August 11, 2019, accepted September 4, 2019, date of publication September 16, 2019, date of current version October 7, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.2941507
ABSTRACT This paper attempts to solve the security problems in communication, consensus-making and
authentication of nodes in the Internet of vehicles (IoV) for intelligent transport. Considering the defects
of the central node and service complexity in the IoV, the blockchain was integrated with the IoV to
create a decentralized mechanism for communication and consensus-making. In the architecture of the
blockchain-based IoV, the Byzantine consensus algorithm based on time sequence and gossip protocol
is used to complete information communication and consensus authentication, which not only ensures
communication security, improves the consensus efficiency of nodes, but also improves the fault tolerance
of the algorithm. The experimental results show that our algorithm outshined the traditional authentication
method in information security and consensus efficiency of the IoV. The research findings provide a reference
solution to the authentication problems in the IoV for intelligent transport.
INDEX TERMS Blockchain, consensus algorithm, intelligent transport, Internet of vehicles (IoV).
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
VOLUME 7, 2019 139703
W. Hu et al.: Blockchain-Based Byzantine Consensus Algorithm for Information Authentication
based on the highly secure bilinear mapping theory [7]. Refer- without mutual trust; the nodes cannot spoof each other under
ence [8] designs a safe and efficient batch authentication plan the rule of the blockchain. The traceable feature means no
for the IoV, in the light of the short key, fast speed and high information in the blockchain is generated without basis; any
security of the elliptic curve cryptosystem. Reference [9] puts block can access the information about the previous block,
forward a batch anonymous authentication method, drawing for all blocks are linked up chronologically in the blockchain.
on the properties of bilinear pairs on elliptic curves. Ref- The smart contracted feature refers to the fact that, the agree-
erence [10] develops a dynamic trust model based on the ment between the transaction parties in the blockchain will be
messages and behaviors of vehicles, which realizes accurate executed automatically by the prewritten code; the execution
trust evaluation in real time and senses malicious nodes in an is immune to any intervention, making the agreements easier
active manner. to conclude and more efficient and less costly to execute.
The existing studies have the following defects: (1) The The blockchain system consists of six layers, namely,
research has been focused on the security of the central the data layer, the network layer, the consensus layer,
node, without changing the structure of the IoV; (2) The the incentive layer, the contract layer and the application
authentication technologies and structures are far from layer [13]. The architecture of the blockchain system is
diverse; (3) The trust computation has poor real-time perfor- described in Figure 1 below.
mance and accuracy.
In the light of the above, this paper integrates the
blockchain with the IoV, shedding new light on the secu-
rity and authentication of the IoV for intelligent transport.
Firstly, the IoV nodes were divided into the vehicle-mounted
communication nodes (VCNs) and roadside communication
nodes (RCNs). Then, a Byzantine consensus algorithm was
established based on time sequence and gossip sequence
(BCA-TG) to guarantee the security and consensus authen-
tication of the information in the IoV. The experimental anal-
ysis shows that the blockchain-based BCA-TG algorithm can
authenticate consensus in the IoV efficiently and rapidly with
high fault tolerance.
The remainder of this paper is organized as follows:
Section 2 introduces the blockchain and the IoV and discusses
the fitness between the two factors; Section 3 illustrates the
overall architecture of blockchain-based IoV, explains the
communication and update of VCNs and RCNs under
the gossip protocol, and describes the authentication of IoV
information by the BCA-TG algorithm; Section 4 demon-
strates the fault tolerance, convergence speed and scala-
bility of the BCA-TG algorithm; Section 5 experimentally
verifies the authentication process under specific scenarios;
Section 6 puts forward the main conclusions.
II. BLOCKCHAIN AND INTERNET OF VEHICLES FIGURE 1. The architecture of the blockchain system.
A. BLOCKCHAIN
Blockchain, a revolutionary technology in the Internet era, As shown in Figure 1, the data layer encapsulates the
adopts an underlying decentralized collaboration mechanism underlying data blocks and the related techniques like data
to create a distributed ledger. The data are linked up by encryption and timestamping; the network layer consists
chronologically generated blocks, forming a data structure of the mechanisms for distributed networking, data prop-
suitable for any decentralized trust network [11], [12]. The agation and data verification; the consensus mechanism
unique formation mechanism has blessed the blockchain with mainly includes the various consensus algorithms for net-
such features as decentralized, trust-free, traceable and smart work nodes; the incentive layer integrates economic factors
contracted. into the blockchain system, including the issuance and dis-
The decentralized feature is demonstrated by the absence tribution mechanisms of economic incentives; the contract
of central node in the network; all nodes have equal status, layer contains various scripts, algorithms and smart con-
and the failure of any node will not affect the system oper- tracts, making the blockchain programmable; the application
ation. Trust-free means all data operations in the system are layer covers various application scenarios and cases of the
open and transparent; internodal communication is possible blockchain.
all network nodes will have the same information. The pro- of the node ranked in chronological order. Note that =
tocol is implemented in the following steps: Node A, after its {(t1 , i1 ), (t2 , i2 ), . . . , (tn , in )} and t satisfies tj < tj+1 (j =
information is updated, sends the updated information to its 1, 2, . . . , n − 1).
neighbors (e.g. node B); if its information is not updated for The node time sequence is an important basis for the
a period of time, node A will request a neighbor for passive consensus algorithm to judge if a node can enter or exit the
information update, and the neighbor will send its latest network. The entry and exit of nodes must be controlled well.
information to node A upon receiving the request. The above Otherwise, many nodes may enter or exit the network simul-
process is repeated until all network nodes have updated their taneously, exceeding the channel tolerance and undermining
information. the consensus-making. To prevent this situation, thresholds
Under the gossip protocol, any node saves the information need to be set for the number of entering/exiting nodes in the
on its neighbors, and three ways are available for the commu- consensus algorithm. If this number surpasses the thresholds,
nication between any two nodes [19]: then the time window protection mechanism will initiate
(1) Push: Node A randomly selects its neighbor B as the automatically. Under this mechanism, only the first n nodes
communicatee, and sends its own information to the within the thresholds in the node time sequence can enter or
latter. Then, node B will update the local information exit the network, and the other nodes must wait for the next
according to the received information. In this case, time window. The specific flow of this mechanism is given in
node A acquires new information. Figure 3 below.
(2) Pull: Node A randomly selects its neighbor B as
the communicatee, and requests passive information
update. Then, node B will transmit the information that
node A lacks to node A. Finally, node A will update
the local information according to the received infor-
mation. In this case, node B acquires new information.
(3) Push-Pull: In the first phase, this communication mode
is the same as the Pull mode. In the second phase, node
A will transmit the information it lacks to node B. Then,
node B will update the local information according to
the received information.
The time for two nodes to have the same information
through the above modes is called a cycle. In each cycle,
the Push mode only needs to communicate once, the Pull
mode needs to communicate twice, and the Push-Pull mode
needs to communicate three times. Comparatively, the Push-
Pull has the best communication effect, because the two nodes
can have the same information in one cycle.
2) TIME SEQUENCE FIGURE 3. The flow of node entry/exit based on time sequence.
A time sequence refers to a series of information or data
of an object arranged in chronological order [20]. In our
daily life, the common time sequences include the daily 3) PUSH-PULL COMMUNICATION AND UPDATE
price of goods, the stock index and the number of tourists BASED ON THE GOSSIP PROTOCOL
in scenic spots. Depending on the number of information The Push-Pull communication and update of RCNs and
types, time sequences can be divided into unit time sequence VCNs will generate three types of information: local infor-
and n-variate time sequence (n≥2). The former only records mation (LI), update information (UI) and consensus infor-
one type of information of the object at a time, while the mation (CI). The LI is the information to be agreed upon in
latter records n types of information of the object each time, the IoV. The UI is a 1D continuously updated vector formed
creating a multi-dimensional sequence. through the communication between n RCNs; each element
The unit time sequence is more common than the n-variate in the vector corresponds to the LI of different nodes. The CI
time sequence. Each information unit can be abstracted into is the vector element taking up more than half of the UI.
a 2D vector (t, i), where t is the time to record information Under the gossip protocol, the RCNs of the IoV participate
and i is the information content. On this basis, the node time in the consensus-making. The initial UI of each RCN consists
sequence for consensus algorithm can be defined as: of the corresponding LI and n-1 null values. Each pair of
Definition 1 Node Time Sequence: Let t be the time for a nodes have the same UI after update, for the UI of each
node to enter and exit the network and i be the information RCN is updated by the Push-Pull mode. After the first pair
of the node upon the entry and exit. Then, the node time of RCNs have updated their information, the updated UI will
sequence can be defined as the set of the 2D vectors be used to update the information of another pair of RCNs in
TABLE 1. Specific information of the five nodes. other under the Push-Pull mode based on gossip protocol,
and sent their own UIs to the other party. The UIs did not
change after the transmission, because the information had
been encrypted using the Hash function, and applied with dig-
ital signature technology. Based on the received information,
Nodes A and F updated their own information, and obtained
the same UI vector [i, null, null, null, h, null], where i is the
arbitrary value generated by A, null is a null value, and h
is the pre-set correct value. The other serial numbers have
TABLE 2. The information on the consensus-making nodes after the similar meanings with serial number 1. Note that, for serial
entry/exit. number 2, the UI of node A is the updated UI through the
communication between nodes A and F in serial number 1.
After the first round of communication, no null value was
found in the UIs of nodes G and B. In the UIs of the two nodes,
four of the elements were h. However, the corresponding
elements in the UIs of nodes A and D were not h, indicating
that A and D are Byzantine nodes. The second round of
consensus-making was performed, because some nodes still
had null elements in their UIs. The results of the second round
are displayed in Table 4 below.
network normally. The information on the consensus-making TABLE 4. The results of the second round of consensus-making.
nodes after the entry/exit is given in Table 2 below.
Finally, there are 6 nodes involved in consensus-making,
including 2 Byzantine nodes. The value of c is zero because
it is the first consensus-making. Then, the following can be
derived from formula (1).
a = (b + c)%N = (5 + 0)%6 = 5 (3)
According to formula (3), the VCN node which generates
the information for consensus-making will send the informa-
tion to node F. Upon receiving the information, node F will
release the information and initiate the consensus-making
process. Then, each node will determine the UI based on its
own LI. The results of the first round of consensus-making
are recorded in Table 3.
For Table 4, the consensus-making was implemented sim-
TABLE 3. The results of the first round of consensus-making. ilarly to Table 3. Since nodes A and C are Byzantine nodes,
the UI elements of the two nodes were reset arbitrarily, and
the two nodes both searched for the other node in its view for
communication. The results of the second round show that,
after the consensus-making in serial numbers 2 and 3, nodes E
and B could identify nodes A and D as Byzantine nodes; after
the consensus-making in serial numbers 4 and 5, nodes E and
F could identify nodes A and D as Byzantine nodes; after the
consensus-making in serial numbers 6, nodes G and D could
identify nodes A and D as Byzantine nodes. The third round
of consensus-making was performed, because node A after
consensus-making in serial number 1 still had null elements
in its UI. The results of the third round are presented in Table 5
below.
For Table 5, the consensus-making was implemented sim-
Serial number 1 in Table 3 means node A finds its neighbor ilarly to Tables 3 and 4. Since each node has only two
F for communication. However, node A, as a Byzantine node, neighbors in its view, the neighbor for consensus-making
wanted to undermine the consensus-making, and sent a UI of was the same with that in the first round. The results of
arbitrary value to F. The two nodes communicated with each round 3 indicate that, after the consensus-making in serial
TABLE 5. The results of the third round of consensus-making. In summary, the BCA-TG algorithm has the following
advantages. (1) Decentralization: the information in the IoV
can be authenticated without the central node. (2) Byzantine
fault tolerance: the algorithm can reach a consensus when
fewer than half of network nodes are Byzantine nodes, with-
out sacrificing the efficiency of consensus-making. (3) Scal-
ability: the IoV nodes can communicate with each other,
despite the entry/exit of multiple nodes. The research findings
contribute to the security authentication of IoV nodes.
REFERENCES
[1] Y. L. Zhang, ‘‘Design of vehicle network data exchange system based
on blockchain,’’ M.S. thesis, School Electron. Inf. Eng., Tianjin Polytech.
Univ., Tianjin, China, 2018.
[2] H. Teng, Y. Liu, N. N. Xiong, Z. Cai, T. Wang, X. Liu, and A. Liu, ‘‘A novel
number 1, no element in the UI of node A was null; after the code data dissemination scheme for Internet of Things through mobile
vehicle of smart cities,’’ Future Gener. Comput. Syst., vol. 94, pp. 351–367,
consensus-making in serial number 6, none of the nodes had May 2019.
any null element in their UIs, and the correct nodes shared the [3] K. Lin, F. Xia, and G. Fortino, ‘‘Data-driven clustering for multimedia
same UI elements. The CI was finalized as h, which occupied communication in Internet of Vehicles,’’ Future Gener. Comput. Syst.,
vol. 94, pp. 697–711, May 2019.
more than half the elements in the UI. [4] N. Soni, R. Malekian, D. Andriukaitis, and D. Navikas, ‘‘Internet of Vehi-
The above example shows that the proposed BCA-TG cles based approach for road safety applications using sensor technolo-
algorithm obtained the correct CI h in three rounds of commu- gies,’’ Wireless Pers. Commun., vol. 105, no. 4, pp. 1257–1284, Apr. 2019.
[5] F. Kong, J. Li, B. Jiang, and H. Song, ‘‘Short-term traffic flow prediction
nication, allowing the correct nodes to identify the Byzantine in smart multimedia system for Internet of Vehicles based on deep belief
nodes through the consensus-making. This means our algo- network,’’ Future Gener. Comput. Syst., vol. 93, pp. 460–472, Apr. 2019.
rithm can reach a consensus when fewer than half of network [6] S. Liu, Y. Zhang, L. Wang, X. V. Wang, and Y. Liu, ‘‘An ‘Internet of Things’
enabled dynamic optimization method for smart vehicles and logistics
nodes are Byzantine nodes, and there are nodes entering tasks,’’ J. Cleaner Prod., vol. 215, pp. 806–820, Apr. 2019.
and exiting the network, that is, our algorithm enjoys good [7] X. Wang, X. Wei, and L. Wang, ‘‘A deep learning based energy-efficient
computational offloading method in Internet of Vehicles,’’ China Com-
Byzantine fault tolerance. mun., vol. 16, no. 3, pp. 81–91, Mar. 2019.
[8] C. Song, M. Y. Zhang, and W. P. Peng, ‘‘Research on vehicle network
VI. CONCLUSION anonymous authentication scheme,’’ Small Microcomput. System, vol. 39,
no. 5, pp. 899–903, 2018.
Traditionally, the communication and authentication of the [9] J. Kang, Z. Xiong, D. Ye, D. I. Kim, J. Zhao, and D. Niyato, ‘‘Toward
IoV nodes are completed via the central node. The centralized secure blockchain-enabled Internet of Vehicles: Optimizing consensus
mode puts high requirements on the security and communi- management using reputation and contract theory,’’ IEEE Trans. Veh.
Technol., vol. 68, no. 3, pp. 2906–2920, Mar. 2019.
cation ability of the central node. To meet the requirements, [10] L. Park, S. Jeong, D. S. Lakew, J. Kim, and S. Cho, ‘‘New challenges
the common solution is to enhance the defense capability of of wireless power transfer and secured billing for Internet of electric
the central node, or implement asymmetric encryption. But vehicles,’’ IEEE Commun. Mag., vol. 57, no. 3, pp. 118–124, Mar. 2019.
[11] M. Swan, ‘‘Blockchain thinking : The brain as a decentralized autonomous
these solutions cannot change the structure of the IoV, failing corporation [Commentary],’’ IEEE Technol. Soc. Mag., vol. 34, no. 4,
to eliminate security loopholes. pp. 41–52, Dec. 2015.
[12] N. Kshetri, ‘‘Can blockchain strengthen the Internet of Things?’’ IT Prof.,
To solve the problem, this paper modifies the IoV structure vol. 19, no. 4, pp. 68–72, 2017.
in the light of blockchain: the central node was removed, [13] S. Huckle, R. Bhattacharya, M. White, and N. Beloff, ‘‘Internet of Things,
and all network nodes were made equally important. This blockchain and shared economy applications,’’ Procedia Comput. Sci.,
vol. 98, pp. 461–466, Sep. 2016.
approach ensures the security of node communication in [14] L.-W. Chen and Y.-F. Ho, ‘‘Centimeter-grade metropolitan positioning
the Internet of Vehicles. Then, an efficient Byzantine con- for lane-level intelligent transportation systems based on the Internet of
sensus algorithm was constructed based on time sequence Vehicles,’’ IEEE Trans. Ind. Informat., vol. 15, no. 3, pp. 1474–1485,
Mar. 2019.
and gossip protocol (BCA-TG). Firstly, the gossip protocol [15] Q. Kong, R. Lu, H. Bao, and M. Ma, ‘‘A privacy-preserving sensory data
was introduced to Byzantine consensus algorithm, and the sharing scheme in Internet of Vehicles,’’ Future Gener. Comput. Syst.,
Push-Pull mode of the protocol was applied to the com- vol. 92, pp. 644–655, Mar. 2019.
[16] F. Li, H. Zhang, J. Wang, C. Sanin, E. Szczerbicki, and L. Gao, ‘‘A set
munication between IoV nodes. In this way, two network of experience-based smart synergy security mechanism in Internet of
nodes can have the same information through one cycle Vehicles,’’ Cybern. Syst., vol. 50, no. 2, pp. 230–237, 2019.
[17] M. C. Cooper, A. Herzig, F. Maffre, F. Maris, and P. Régnier, ‘‘The epis-
of communication. Next, the time sequence was adopted temic gossip problem,’’ Discrete Math., vol. 342, no. 3, pp. 654–663, 2019.
to improve the efficiency and speed of the communica- [18] L. Altoaimy, A. Alromih, G. Al-Hudhud, H. Kurdi, K. Youcef-Toumi,
tion and consensus-making between the nodes. On this and S. Al-Megren, ‘‘Context-aware gossip-based protocol for Internet of
Things applications,’’ Sensors, vol. 18, no. 7, p. 2233, 2018.
basis, the consensus-making of the BCA-TG algorithm in [19] J. Liu, S. Mou, B. D. O. Anderson, C. B. Yu, and A. S. Morse, ‘‘Request-
blockchain-based IoV was explained in details. The experi- based gossiping without deadlocks,’’ Automatica, vol. 93, pp. 454–461,
mental results show that our algorithm can effectively over- Jul. 2018.
[20] L. Chao, J. Zhipeng, and Z. Yuanjie, ‘‘A novel reconstructed training-
come the defects in consensus and authentication of IoV set SVM with roulette cooperative coevolution for financial time series
nodes. classification,’’ Expert Syst. Appl., vol. 123, no. 1, pp. 283–298, 2019.
WEI HU is currently the Dean of the Depart- WENHUI YAO was born in Zhenjiang, Jiangsu,
ment of Management Science and Engineering, China, in 1996. She received the B.S. degree
Shanghai University of Electric Power, where he in electrical engineering and automation from
is also an Associate Professor. His current research Yangzhou University, in 2018. She is currently
interests include power big data and energy block pursuing the M.S. degree in power engineering
chain. His recent research has been focused on economics and management with the Shanghai
energy and power optimization decision. University of Electric Power, China. Her current
research interests include distributed power dis-
patch model based on blockchain and electricity
market trading. Her awards and honors include
second scholarship and outstanding graduates.
YAWEI HU was born in Shuyang County, Jiangsu, HUANHAO LI was born in Huai’an, Jiangsu,
China, in 1996. He received the bachelor’s degree China, in 1991. He received the degree in
in information and computing science from the electrical engineering and automatic chemistry
Shanghai University of Electric Power, in 2018, from Yangzhou University, in 2016. He is cur-
where he is currently pursuing the master’s degree rently pursuing the master’s degree in electri-
in power engineering economics and management. cal engineering economics and management with
His current research interests include power sup- the Shanghai University of Electric Power. His
ply and demand network trading mechanism based current research interests include power spot
on blockchain, consensus algorithm, and smart trading model research based on blockchain
contract based on blockchain. He was a recipient and blockchain technology in ubiquitous power
of the Second Prize of Challenge Cup in Shanghai, in 2017, and the Shanghai Internet of Things.
Outstanding Graduates, in 2018.