IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO.

15, 1 AUGUST 2022 13263

A Trust-Centric Privacy-Preserving Blockchain for

Dynamic Spectrum Management in IoT Networks
Jingwei Ye , Graduate Student Member, IEEE, Xin Kang , Senior Member, IEEE,
Ying-Chang Liang , Fellow, IEEE, and Sumei Sun , Fellow, IEEE

Abstract—Blockchain is a promising technology for future healthcare, augmented, and virtual reality will continue driving
dynamic spectrum access (DSA) management due to its the demand for radio spectrum [1]. However, the current static
decentralization, immutability, and traceability. However, many spectrum management method has resulted in under-utilization
challenges need to be addressed to integrate the blockchain
to DSA, such as the trustworthiness of participating nodes’ of spectrum resources [2]. To this end, dynamic spectrum
spectrum sensing results, privacy protection of sensing nodes’ access (DSA) has been proposed to allow secondary users
identities, and affordable lightweight consensus algorithms for (SUs) sense and then use the idle spectrum bands. Since single
IoT devices. In this article, we propose a trust-centric privacy- SU’s sensing capability is usually limited, cooperative sens-
preserving blockchain for DSA in IoT networks. To be specific, we ing [3] is proposed to fuse the sensing results of multiple
propose a trust evaluation mechanism to evaluate the trustworthi-
ness of sensing nodes and design a Proof-of-Trust (PoT) consensus SUs to improve the sensing accuracy. However, the traditional
mechanism to build a scalable blockchain with high transaction- usage of a center to collect and fuse sensing results faces the
per-second (TPS). Moreover, a privacy protection scheme is risk of a single point of failure and the increasing management
proposed to protect sensors’ real-time geolocation information and regulatory costs. Moreover, SUs usually need to upload
when they upload sensing data to the blockchain. Two smart their sensing data to the fusion center at the risk of leakage
contracts are designed to make the whole procedure (spectrum
sensing, spectrum auction, and spectrum allocation) run automat- of their privacy, such as identity and location.
ically. Simulation results demonstrate the expected computation As an emerging technology, blockchain has attracted atten-
cost of the PoT consensus algorithm for reliable nodes is low, and tion from both academia and industry. By using blockchain, a
the cooperative sensing performance is improved with the help of decentralized resource management system can be established
the trust evaluation mechanism. In addition, incentivization and without the requirement of trustworthy central management
security are also analyzed, which show that our system can not
only encourage nodes’ participation, but also resist many kinds agencies [4]–[8]. Moreover, smart contracts which are imple-
of attacks which are frequently arise in the trust management mented on a blockchain can be used to replace traditional
mechanism and blockchain-based IoT systems. central management agencies and facilitate the cooperation
Index Terms—Blockchain, consensus algorithm, cooperative of users [9]. Recently, researchers have investigated how to
spectrum sensing, dynamic spectrum access (DSA), trust apply the blockchain technology to DSA. Weiss et al. [10]
model. summarized the types of blockchains applicable in different
spectrum sharing scenarios, and the possible advantages and
disadvantages of applying the blockchain technology to DSA.
I. I NTRODUCTION With the help of cryptocurrency issued by the blockchain
ECENT years have witnessed the exponential growth and smart contracts, flexible, and automatic spectrum trad-
R of mobile data traffic. The rapid development of new
wireless applications, such as autonomous vehicles, remote
ing markets are made possible for spectrum sellers and
buyers [11]–[15].
However, the integration of blockchain and DSA still faces
Manuscript received 25 June 2021; revised 11 October 2021 and many challenges. First, although the data recorded on a
13 December 2021; accepted 3 January 2022. Date of publication
13 January 2022; date of current version 25 July 2022. This work was blockchain is prevented from being tampered, the quality
supported in part by the National Natural Science Foundation of China or value of it cannot be guaranteed. Especially, in a public
under Grant U1801261 and Grant 61631005; in part by the National Key blockchain, a malicious user can easily join the blockchain
Research and Development Program of China under Grant 2018YFB1801105;
in part by the Macau Science and Technology Development Fund (FDCT), and record their data in the blockchain. The data from such
Macau SAR, under Grant 0009/2020/A1; in part by the Key Areas of a malicious user or an unreliable user is valueless or even
Research and Development Program of Guangdong Province, China, under harmful to cooperation tasks based on the data, e.g., cooper-
Grant 2018B010114001; in part by the Programme of Introducing Talents of
Discipline to Universities under Grant B20064; and in part by the Fundamental ative sensing. Therefore, there is compelling need to evaluate
Research Funds for the Central Universities under Grant ZYGX2019Z022. the quality of the data from each user. Second, though the
(Corresponding author: Xin Kang.) communication among Ethereum external accounts can be pro-
Jingwei Ye, Xin Kang, and Ying-Chang Liang are with the Center
for Intelligent Networking and Communications, University of Electronic tected by encryption and decryption with their public and
Science and Technology of China, Chengdu 611731, China (e-mail: private key pairs, a contract account is not equipped with a key
[email protected]; [email protected]; [email protected]). pair hence such protection method is not applicable to smart
Sumei Sun is with the Institute for Infocomm Research, Agency for Science,
Technology and Research, Singapore (e-mail: [email protected]). contracts. As a result, sensing results uploaded to the smart
Digital Object Identifier 10.1109/JIOT.2022.3142989 contract cannot be encrypted, and the sensitive information
2327-4662 
c 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13264 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

in the sensing results (such as real-time geolocation) can II. P RELIMINARIES

thus be accessed by malicious users. Third, traditional con- A. Blockchain
sensus algorithms like Proof of Work (PoW) used in pub-
Blockchain is a growing chain of data blocks with a spe-
lic blockchain introduce much computation overhead and
cific data structure constructed using cryptographic algorithms.
thus make the transaction processing speed very low. As
It uses hash pointers to connect the data blocks to form an
a result, it is inefficient to directly use existing consensus
entangled chain. In this way, the integrity of the data can be
algorithms (such as PoW) for spectrum management, espe-
protected. Blockchain uses a measurable and verifiable mech-
cially considering the limited computation capabilities of IoT
anism to reach a consensus among nodes in the network on
devices.
the generation of new blocks. Such a mechanism is usually
In this article, we consider the DSA for an IoT network
referred to as the consensus algorithm. On the premise that the
where IoT devices opportunistically access the licensed spec-
data block entanglement is guaranteed, the structure and the
trum bands through cooperative sensing [16]. Blockchain
consensus algorithm of the blockchain can be flexibly designed
is used as a platform for dynamic spectrum management.
based on the demand of application scenarios.
Specifically, key component designs for such a blockchain-
Blockchain can be roughly classified into three types,
enabled DSA system, including the trust evaluation mech-
depending on the degree of openness, which are public
anism, privacy protection, consensus algorithm, and smart
blockchain, private blockchain, and consortium blockchain.
contracts are studied. The main contributions of this article
Public blockchain is designed to enable and to record the trans-
are summarized as follows.
actions in a public network, which means that any node can
1) We propose a trust evaluation mechanism to evaluate the
freely join or leave the blockchain network without autho-
trustworthiness of sensing nodes which participate in the
rization. Bitcoin, the most famous digital cryptocurrency, is
cooperative sensing. We show that the proposed trust
generated and managed by a public blockchain. Consortium
evaluation mechanism is effective in incentivizing sens-
blockchain is a permissioned blockchain, which is open to
ing nodes to be honest, and thus improving the spectrum
authorized members of external institutions in limited roles
sensing result.
and functions. The authorization of the node to join is all deter-
2) We propose a strong privacy protection mechanism by
mined by an authorized organization. The nodes in a private
combining ring signature with the commitment scheme
blockchain network trust each other. Thus, private chains can
for providing sensing nodes’ privacy such as their real-
simplify operations in data authentication and the consensus
time geolocation. The ring signature is used to protect
algorithm to improve the efficiency.
sensing nodes’ identities when they upload sensing data.
The commitment scheme is proposed to figure out the
B. Smart Contract
connection between sensing data packet and its origi-
nated sensing node after the fusion of sensing results, In the 1990s, Szabo [18] first proposed the concept of a
so that the trust value of each sensing node is updated smart contract. It was defined as computerized transaction
and the incentive tokens are distributed. protocol that execute terms of a contract. In the field of cryp-
3) We propose a new consensus algorithm named as Proof- tocurrency, a smart contract is defined as an application or
of-Trust (PoT) by connecting the mining difficulty with a a program that runs on a blockchain. Normally, it contains
miner’s trust value. We show that the proposed PoT can a set of digital agreements with specific rules. These rules
greatly reduce the computation cost of honest nodes and are predefined in the form of computer source codes, and all
enhance the scalability of the current blockchain-based network nodes will copy and execute these computer source
DSA system. codes independently. Smart contracts are highly customizable
4) We design the system protocol and two smart contracts and can be flexibly designed to provide different services and
to make the whole DSA procedure, including spectrum solutions.
sensing, spectrum auction, and spectrum allocation, run
automatically. In addition, we implement and verify a C. Ring Signature and the Commitment Scheme
prototype of our proposed protocol and smart contracts Ring signature was first introduced in 2001 [19] which fol-
using Solidity [17] on the Ethereum testnet. low a ring-like structure of the signature algorithm. It is a type
The remainder of the article is organized as follows. In of digital signature that can be performed by any member of
Section II, we introduce the basic concept of blockchain and a group of users without the agreement of others. Therefore,
smart contract. In Section III, we describe the works related the only “open” information is that a message signed with a
to the application of blockchain to DSA. In Section IV, we ring signature is endorsed by someone in a particular group
introduce our system model. In Section V, we describe the of people. Ring signatures are deliberately designed so that it
design of our proposed protocol, including the block struc- is computationally infeasible to determine which of the group
ture, the trust evaluation mechanism, the consensus algorithm, members’ keys was used to produce the signature. Ring signa-
and the privacy protection scheme. In Section VI, the design of tures are similar to group signatures but differ in two key ways:
smart contracts and workflow are introduced. In Section VII, 1) there is no group administrator to revoke the anonymity of
we discuss the performance of the trust evaluation mechanism, an individual signature and 2) any set of users can be used
and analyze the security of the system. Finally, Section VIII as a group by the signer without additional setup. In public
concludes this article. blockchains, there is no leader or manager in the system, who

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13265

can serve as the group manager required in the group signa-

ture scheme. Thus, a ring signature scheme is more suitable
for this scenario than a group signature.
Commitment schemes are basic elements in many crypto-
graphic protocols. Usually, the two-stage commitment scheme
is used to enable a party to commit itself to a value while
keeping the value secret first, and then rebuild the connections
between the data packets and their corresponding owner [20].

III. R ELATED W ORK

Blockchain was first proposed as a secure decentralized
ledger for recording spectrum transaction data and analyz-
ing overall spectrum usage in dynamic spectrum management.
Kotobi and Bilen [21] proposed to use the blockchain as
a distributed database to securely store the transactions in
dynamic spectrum sharing, and to use the digital currency Fig. 1. System model.
issued by the blockchain as the currency for spectrum auctions.
In [15], a smart contract running on Ethereum was proposed earn interest through the long-term holding of tokens, which
as a platform to provide spectrum sensing services. In [13], make fewer people willing to sell tokens and reduce the liquid-
a blockchain-based spectrum trading and sharing scheme was ity of currency. For the DPoS adopted in the EOS network, in
proposed for mobile network operators (MNOs) to lease their order to improve performance, there are only 21 super nodes.
spectrum to the secondary unmanned aerial vehicle (UAV). Due to this reason, this blockchain is no longer really decen-
In [22], spectrum sharing contracts deployed on a permis- tralized, but “partially decentralized” [32]. In this article, we
sioned blockchain platform are constructed for multioperators need a consensus algorithm which is more suitable for the IoT
spectrum trading. Ariyarathna et al. [14] proposed to use smart network environment.
contracts to securely store and then autonomously implement On the other hand, the blockchain system is an open system,
the spectrum leasing agreements, thus to avoid the interference hence the reliability of data source needs to be guaranteed.
from the disordered spectrum access. In [33], the data uploaded by the sensors are compared with
However, these works did not make any improvement to the the data collected by the trustworthy validator to determine
existing blockchain. Thus, the transaction-per-second (TPS) the credibility degree of the sensor, in this way, the system
of the blockchain is low, limiting the performance of the can evaluate the credibility of nodes for better performance.
blockchain-based DSA platform. Thus, Fan and Huo [23] However, this method is a semi-centralized method which
proposed a consensus algorithm named as “blockchain-KM needs to deploy trustworthy validators.
protocol” by modifying the PoW, to speed up the transac- Different from the above works, we not only delve into the
tion processing for the license-free spectrum sharing. In [24], design of the block structure and consensus algorithm, intend-
a novel consensus algorithm named as proof-of-strategy was ing to overcome the scalability problem of the blockchain
proposed, where the best strategy to allocate spectrum bands technology, but also design a decentralized trust evaluation
in the whole network is the proof of authority to publish a mechanism and a privacy protection mechanism for the open-
new block. Such a consensus algorithm can not only regu- ness and transparency of blockchain system while protecting
late the transaction generation, but also optimize the spectrum the data’s confidentiality and the sensors’ privacy. To the best
allocation. However, proof-of-strategy couples the generation of our knowledge, this is the first work of joint trust evaluation
of transactions with spectrum allocation strategies which make and privacy protection mechanism in cooperative sensing in a
consensus algorithm complex, and this reduces the efficiency blockchain-enabled DSA system.
of the consensus mechanism and also the system throughput.
In [25], by modifying the PoW, the authors proposed a con- IV. S YSTEM M ODEL , C HALLENGES , AND
sensus algorithm named as proof-of-device, which selects a D ESIGN G UIDELINES
device, represented by its unique identifier, to publish the new
In this section, we introduce the DSA scenario in future IoT
block in a way like lottery. It thus eliminates the heavy compu-
network, and list the difficulties and challenges in applying the
tation cost in PoW. However, this consensus algorithm relies
blockchain technology to this scenario. Finally, we give the
on the unique ID of each device, and the ID is relatively con-
guidelines on how to design a blockchain-based DSA system.
stant. Thus, malicious users may predict the winning miner
of each round, and launch a DoS attack on this device which
may threat the security of the system. A. Blockchain-Enabled DSA System Model
There are also many other public blockchains that improve In this article, we consider the DSA in a two-tier cognitive
TPS through other types of consensus algorithms, such as radio network consisting of primary users (PUs) and SUs. In
Proof of Stake (PoS) [26]–[28], DPoS [29]–[31], etc. They this network, PUs are inactive users who may not use their
can improve the TPS of the blockchain system, but they also spectrum all the time so SUs can detect these spectrum holes
have some shortcomings. For example, for PoS, all nodes can and use them. As illustrated in Fig. 1, SUs can be a variety
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13266 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

contracts. Compared with the current fixed paper-based

contracts signed between mobile users and operators,
this method has higher flexibility.

B. Difficulties and Challenges

Traditional cooperative spectrum sensing is divided into
two modes: 1) centralized mode and 2) distributed mode.
Centralized cooperative spectrum sensing requires the deploy-
ment of a central fusion node to process the data collected
from sensors which usually contain sensors’ geolocation
information. Therefore, this mode not only requires redun-
dant regulation fees (usually charged to maintain the operation
of the central node), but also faces the risk of data leakage
Fig. 2. Blockchain-based DSA when the central node is attacked. While the distributed mode
often suffers from the fake reporting issue caused by mali-
cious nodes and the iterative-based malicious node detection
algorithms [34] introduce extra computation cost.
of heterogeneous IoT devices, which form a mesh network.
Therefore, how to implement the cooperative sensing in a
SUs access the spectrum bands of the PU in an opportunistic
secure and effective way is a challenge. The decentralized
manner based on cooperative sensing, and a blockchain is used
nature of the blockchain system is promising to realize dis-
as the platform of the cooperative spectrum sensing decision
tributed cooperative sensing and encourage all sensors to share
and the spectrum allocation.
spectrum sensing results. However, there are still some chal-
Basically, as shown in Fig. 2, there are four phases in our
lenges when applying blockchain into spectrum management.
blockchain-enabled DSA system.
First, IoT devices usually lack sufficient computing and stor-
1) Individual Sensing: The sensors perform spectrum sens-
age resources to maintain and store the whole blockchain
ing to detect the state of the PU’s spectrum band.
ledger. Second, a malicious node can easily join a public
2) Sensing Fusion: Each of the sensor broadcasts its sens-
blockchain network and threaten the reliability of cooperative
ing data which mainly consists of the sensing result and
spectrum sensing. Moreover, the sensing-related data recorded
the geolocation where the sensing is performed. These
on blockchain or collected by smart contracts are usually in
data can be raw data or quantized bits, which will be
an unencrypted format which may leak the private information
fed to a smart contract implemented upon a blockchain.
of IoT devices contained in their sensing data packets.
Then, the smart contract provides the final sensing result
according to predefined fusion rules.
3) Spectrum Allocation: When the final sensing result indi- C. Design Guidelines
cates that PU is inactive, it should be decided which Considering the above challenges, our discussion on system
spectrum requester can access the spectrum. Spectrum design mainly includes the following three parts.
auction, as a common and fair way for spectrum allo- 1) Trust Evaluation: Specially, in a public blockchain
cation, is chosen to achieve that. In the spectrum auc- network, anyone can join or exit the system at will,
tion, the digital cryptocurrency, which is issued by the and one user can create multiple accounts in the system.
blockchain through incentive mechanisms for spectrum Therefore, if there is no suitable node evaluation mech-
sensing and mining, is commonly considered. anism, malicious users can launch Sybil attacks easily.
4) Spectrum Access: The SU who obtains the access right This is because, malicious nodes can keep on applying
through spectrum auction accesses the idle spectrum for for accounts and continue launching attacks. Therefore,
data transmission. the system should be able to evaluate the trust values
Compared with traditional cooperative spectrum sensing and for nodes, and identify the malicious nodes based on
spectrum allocation methods, blockchain-enabled DSA has the the their trust values [35].
following benefits. 2) Lightweight Consensus Algorithm: Due to the limita-
1) Decentralization: Blockchain-based DSA does not tion of computing power and storage resources of IoT
require the deployment of trusted central nodes which devices, maintaining a normal public blockchain system
avoids the single point of failure. is a heavy burden. To reduce the computing cost, it
2) Transparency: The blockchain ledger can record the is necessary to invent a new consensus algorithm with
whole process of DSA. low computational and storage complexity. On the other
3) Automatic: Using smart contract instead of traditional hand, as time goes by, the length of the ledger keeps
contracts, we can achieve automatic spectrum manage- on increasing. To reduce the storage cost, methods such
ment and on-chain payment settlement. as edge storage can alleviate storage pressure on the
4) Flexibility: For spectrum trading on the blockchain blockchain nodes but may introduce an additional bud-
platform, diversified spectrum trading rules can be get. Thus, a method that can directly reduce the storage
dynamically enforced by adjusting parameters of smart cost of the blockchain is preferred.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13267

TABLE I
L IST OF N OTATIONS

Fig. 3. Block structure.

of each account. Merkle patricia trie (MPT) [37] which

is a commonly adopted data structure, is used here to
store the account states.
3) Privacy Protection Mechanism: Smart contracts does not
have public-and-private key pair and thus can not com-
municate with external accounts in an encrypted manner. B. Trust Evaluation Mechanism
Therefore, when a smart contract is used as the data Although the blockchain can guarantee that data recorded
fusion center of cooperative spectrum sensing, a privacy in a block will not be tampered, it cannot guarantee that the
protection mechanism is needed to ensure that the pri- data source is trustworthy. Especially, in a public blockchain,
vate information of the sensors will not be leaked during a malicious node can upload misleading data which may
the interaction between sensors and the smart contract. interfere the cooperative sensing based on such data. In this
case, the performance of cooperative sensing will be degraded,
V. B LOCKCHAIN D ESIGN , T RUST E VALUATION , AND and the honest and reliable sensors will be discouraged from
P RIVACY P ROTECTION participating in cooperative sensing. To this end, we design a
In this section, we first introduce the block structure of our trust evaluation mechanism to evaluate the credibility of every
blockchain, then we propose the trust evaluation mechanism to sensing node in the blockchain. Instead of using a central
evaluate the trustworthiness of each SU on spectrum sensing. authority to record and maintain every account’s trust value,
After that, a high-efficiency consensus algorithm is proposed we propose to include the trust value as an attribute of the
for our blockchain. Finally, a privacy protection mechanism is blockchain account, which is recorded in the ledger and main-
proposed for protecting participants’ privacy. tained by all the miners. In the following, we will discuss the
design of the trust value updating method.
A. Block Structure Intuitively, the trust value of a sensing node should be
adjusted based on its performance of the current and histor-
The block structure should be tailor designed for the ical cooperative sensing. In detail, if it plays a positive role
blockchain-based DSA system. Some of the unnecessary com- in the cooperative sensing, its trust value should be increased,
ponents in the block can be removed to reduce the overheads and vice versa. Here, we define that the effect of a sensor is
on the block transmission and storage and other necessary positive when its sensing result is consistent with the coop-
information, such as the trust value of the block generator, erative sensing result, and is negative when its sensing result
needs to be recorded in the block header to prevent possible is inconsistent with the cooperative sensing result, or when
attacks in the DSA system. the node does not participate in cooperative sensing. Denote
1) Block Header: As shown in Fig. 3, The block header the number of times when the sensing result of the ith node is
mainly contains: a) Merkle root of transactions; b) the consistent with the cooperative sensing result as Ni,r ; the num-
Merkle root of account states; c) hash value of the header ber of times when the sensing result of a node is inconsistent
of the previous block, as the hash pointer of the chain with the cooperative sensing result as Ni,w .
structure; d) trust value of the miner who mines this In a sensing-based system, trust is generally determined by
block; e) signature of the miner who mines this block; two factors: 1) the number of correct sensing and 2) incorrect
f) a Nonce, which is the solution of Hash puzzle of the sensing. It can be expressed as
consensus algorithm; and g) timestamp when the block
 
is mined. TVi (n) = g Ni,r (n), Ni,w (n) (1)
2) Block Body: The block body mainly stores transaction
generated during spectrum allocation using the Merkle where g(·; ·) is a 2-D function, and should have the following
tree. Unlike the unspent transaction output (UTXO) [36] properties.
model used in Bitcoin, the system based on the account 1) ∀Ni,r (n), Ni,w (n) ∈ [0, +∞), g(·; ·) ∈ [0, 1].
model usually contains an account tree to record the 2) g(·; ·) is an increasing function of Ni,r (n), and is a
account states, such as the balance and the trust value decrease function of Ni,w (n).

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13268 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

(1) (2)
Inspired by [38], we propose a new model to calculate trust where TVi = TVi (n) − TVi (n − 1) is defined as the
(1)
value of sensori denoted by TVi (n) in sensing round n, where original increment of trust value, which can be either posi-
n is also the number of sensing tasks published since this tive or negative. The design criterion of (5) is explained as
system was created follows. We consider three cases. First of all, when the sen-
  sor senses correctly, the original increment of trust value is
TV(1)
i (n) = e
−ρNi,w (n)
1 − e−ηNi,r (n) (2) positive, i.e., TVi > 0. Such increment is first degraded by
the negative effect of sleeps in the previous sensing activities,
where ρ > 0 and η > 0 are coefficient that determine how i.e., f (Rsleep ), and then added to the trust value. Second, when
fast the trust value changes with respect to Ni,w and Ni,r , the sensor gives the wrong sensing result, the original incre-
respectively. ment of trust value is negative, i.e., TVi < 0. In this case,
Moreover, we let Ni,w (n) decay by (1/P) every time when we choose not to consider the negative effect of the sleep
the node participates in sensing, and we only consider the in previous sensing rounds, in order to encourage the sen-
latest P sensing rounds of each node. In this way, the effect sors which unintentionally derive the wrong sensing result in
of a wrong sensing result on the trust value will gradually be this time. Finally, for sensors who do not participate in sens-
degraded over time. As a result, a node that unintentionally ing, their trust value will not remain unchanged but gradually
submitting an inconsistent result will be gradually forgotten. decrease as their sleep time increases. This design is to ensure
Mathematically, Ni,w (n) can be defined as that sensors can only maintain its high trust value by actively

n   participating in spectrum sensing.
n−m
Ni,w (n) = ri (m) 1 − (3)
P
m=n−P
C. Consensus Algorithm
where ri (m) = 1 if sensori broadcasts the wrong sensing result The traditional consensus algorithm for public blockchain
in round m, and ri (m) = 0 otherwise. like PoW is computation-intensive, which may be inappli-
Besides, we denote the number of rounds when a sensing cable in IoT networks, where IoT nodes are usually with
node is inactive as Rsleep , and we add a function f (Rsleep ) limited computational capabilities. One reason why the PoW
which models the negative impact of being inactive on the is designed to be computation-insensitive is that it assumes
trust value. It is designed to satisfy the following criteria. there is nearly no trust among nodes in a blockchain network.
1) ∀Rsleep ∈ [1, +∞), f (Rsleep ) ∈ (0, 1]. Nevertheless, with the trust value mechanism proposed in
2) (∂fdcv /∂Rsleep ) < 0. this article, we can evaluate the credibility of a node
The first criterion is for normalization; the second crite- on spectrum sensing in the blockchain network. Therefore,
rion is to ensure that the negative effect increases as Rsleep based on the trust value, we can optimize our design of
increases. Moreover, as Rsleep increases, the downward trend the blockchain consensus algorithm, forking solution, and
of the function is first gentle, and then becomes severe. This blockchain compression.
is designed to punish the node that does not participate in the Proof of Trust (PoT): Intuitively, the nodes with higher trust
sensing process. The punishment is light when the peer does value on sensing are more likely to be honest in publishing
not participate only a few times, but becomes severe when a new block. Therefore, we can reduce the difficulty of such
the node always does not participate in the sensing. f (Rsleep ) nodes to mine a new block. In this way, the computation con-
will reach a certain low value when Rsleep is big enough. For sumption for reliable nodes will be decreased, and a block will
example, the trust value of a node will reduce to r1 (r1 < 1) thus be more quickly mined and published. Based on PoW and
of its original value for k1 consecutive nonparticipation, and trust value, we propose a PoT consensus algorithm. Different
to r2 (r2 < r1 < 1) for k2 consecutive nonparticipation. Here, from PoS, PoT assigns high-trust nodes lower mining difficul-
a piecewise function consisting of multiple linear functions is ties, and trust can be gained only by providing correct sensing
utilized as our f (Rsleep ) results and will decrease over time. Thus, high-trust nodes are
⎧ k −R not the same as high-stake nodes. To be specific, for success-
⎪ 1
⎨ k1
sleep
· (1 − r1 ) + r1 , 0 ≤ Rsleep ≤ k1 fully mining, the miner with higher trust value is required to
f (Rsleep ) = R sleep −k2
⎪ · (r1 − r2 ) + r2 , k1 ≤ Rsleep ≤ k2 (4) find a hash value with fewer leading zeros and vice versa.
⎩ k1 −k2
r2 , k2 < Rsleep . Like PoW, PoT can work securely under the assumption that
the majority of hashrate is controlled by honest nodes.
These parameters can be set flexibly. If the actual deployment Mathematically, assuming that the difficulty of node i at
requires stricter penalties, then all these parameters should be block n is denoted as Din . Din is determined mainly by βn and
set smaller. a function g(TVi ), where g(TVi ) should have the following
By adding the attenuation function, the model in (2) can be properties.
modified as 1) g(TVi ) is a decreasing function of TVi .
⎧  
⎪ (2) 2) ∀TVi ∈ [0, 1], g(·) ∈ [0, 1].
⎨ TVi (n − 1) + TVi · f Rsleep , if TVi > 0
In this article, Din is denoted as
TV(2) (1)
i (n) = ⎪ TVi (n), if TVi < 0
⎩ (2)  
TVi (n − 1) · f Rsleep , if inactive  π 
(5) Din = βn · 1 − sin · TVi (6)
2
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13269

Algorithm 1 Block Selection value. If the trust values of multiple blocks are the same, the
Input: the block to be compared: Blocki , Blockj ; block whose timestamp is earlier is selected as the valid block.
Output: The winner block; If it is still tied, it compares the hash value of the blocks, and
if TVi < TVj then then selects the block with the smallest hash value as the valid
Blocki wins. block. Since the probability of hash collision is negligible,
else if TVi > TVj then Algorithm 1 can select only one valid block eventually.
Blockj wins. Blockchain Compression: Since IoT devices are usually
else with limited storage space, each time when the blockchain
if Timestampi < Timestampj then grows by L blocks, compression will be performed. For
Blocki wins. the compression of blockchain, in the existing literature, the
else if Timestampi > Timestampj then RSA accumulator as a data structure, which functions simi-
Blockj wins. larly to that of a Merkle tree, can be used to compress the
else blockchain [39]. Another approach to compress blockchain
if H(Blocki ) < H(Blockj ) then is to use the chameleon hash function to replace the tradi-
Blocki wins. tional hash function of the blockchain [40]. Here, using the
else trust value, we propose a compression method called as trust-
Blockj wins. based compression (TBC). The node with the highest current
end if trust value will be authorized to compress the blockchain. To
end if be specific, such a node will first extract the account tree in
end if the last block as the body of the new block, calculate a hash
value for the body, and finally combine the obtained nonce,
miner’s signature, and timestamp to form the block header.
where βn denotes the base difficulty of block n. TVi ∈ [0, 1] The obtained block is the new genesis block. Since each node
is the trust value of node i. The initial difficulty is denoted stores the original blockchain, it is easy to check whether the
as β0 which can be determined by evaluating the computing account status has been tampered by the selected node during
power of actual IoT devices. How to determine β0 will discuss the compression process. After a node verifies the new genius
this problem in the simulation section. block, it will clear the original blockchain.
We denote the timestamp of block n as Tn , the ideal time
interval of two consecutive blocks is T0 . Then, we have
 D. Privacy Protection Mechanism
Tn−1 − Tn−2
q= (7) The location where a sensing node senses the spectrum
T0 bands is useful for the fusion center to cluster the sensing
where q is defined as adaptive adjustment factor of base diffi- nodes and to improve the cooperative sensing accuracy [41].
culty, and g is defined as the difficulty adjustment granularity, Therefore, alongside the sensing result, the location is needed
which can be written as to be uploaded by a sensing node. However, it is difficult to
 protect the location information of a sensing node from being
βn−1
g= . (8) leaked when a smart contract is used as the fusion center. This
128
is because a smart contract account is not equipped with a key
The updating of βn can be denoted as pair which can be used by sensors to encrypt their upload data
packet.
βn = βn−1 − q · g. (9)
The privacy protection issue in this case is to hide the source
Forking Solution: As the mining speed increases, there may of a sensing packet, i.e., from which sensing node the sensing
be multiple blocks mined at nearly the same time. Because packet comes. However, the sensing node cannot be allowed to
of the communication latency, the block that is first mined be totally anonymous when submitting sensing packet because
might not be the first to be received by all the nodes in the this will make the cooperative sensing system vulnerable to
blockchain network. Instead, the block that is first received malicious attacks. To this end, we propose the use of ring
and recognized by different nodes might be different. In this signature [19] to hide the source of a sensing packet in a group
case, the blockchain forking occurs. It is harmful and thus of valid sensing nodes. Also, the smart contract as the fusion
needs to be solved. Due to the timeliness of spectrum, we center can identify the validity of each received sensing packet.
expect that the spectrum sensing and allocation information Moreover, since sensing packets are unencrypted, fusing these
can be recorded in time, which becomes a checkpoint in the sensing results can be carried out directly and automatically
blockchain ledger at each sensing round. This means that our in the smart contract.
consensus algorithm has deterministic finality, i.e., PoT will In the following, we illustrate the procedures of one sensor,
not fork. To avoid forking, a unique legal block will be deter- denoted by Sensors , in generating the ring signature.
mined in time according to Algorithm 1 at each sensing round, 1) Sensors selects n − 1 legal sensors to form a group and
and miners are required to validate every transaction of can- collect their public keys. The format of a sensing data
didate blocks. In Algorithm 1, we propose a trust value-based packet, denoted as msg, is given as follows:
forking solution. It first compares the trust values in the head  
of blocks, and then validates the block with the highest trust msg = H(msgID), SR, time, location (10)
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13270 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

where H(msgID) is used in the update of the trust value

and SR is the sensing result denoted by one bit.
2) Sensors uses one-way hash function to compute k =
H(msg), which is a symmetric key of the symmetric
encryption function Ek .
3) Sensors generates a random value for each of the other
members in the group where it belongs to. Specifically,
the random number xi is first generated for the ith node
in the group. It then calculates corresponding yi = gi (xi )
using corresponding public key, where the function gi (·)
is the encryption function encrypted with the public Fig. 4. Two-stage commitment scheme.
key pki .
4) Sensors finds the solution to the ring (11) and gets the the predefined conditions are satisfied. Moreover, smart con-
undetermined parameter ys , where v is a random value tracts are not controlled by any third party. Therefore, we
chosen by Sensors . Then, Sensors calculates xs using propose to use smart contracts to realize the automated oper-
its own privacy key: xs = g−1 ation of cooperative spectrum sensing and spectrum auction
s (ys ). To find solution, a
private key of a sensor in this group is needed, anyone on blockchain. In this section, we will give the design of
who is not in the group cannot generate a legal ring the corresponding smart contracts, and then propose the DSA
signature protocol based on these smart contracts.
Ck,v (y1 , y2 , . . . , yn )
A. Smart Contract Design
= Ek (yn ⊕ Ek (yn−1 ⊕ Ek (· · · ⊕ Ek (y1 ⊕ v) · · · ))) = v.
We first discuss the design of our smart contracts for
(11) cooperative sensing and spectrum auction. In the following,
Finally, a valid ring signature, denoted as Ringsig , can be we describe the parameters and functions in the two smart
generated by Sensors contracts, respectively.
Cooperative Sensing Contract (CSC): The parameters in
Ringsig = (pk1 , pk2 , . . . , pkn , v, x1 , x2 , . . . , xn ). (12)
CSC that need to be predefined are Tddl , ds , N1 , and TVthr ,
The verification process involves the following three steps. which are introduced as follows.
1) Calculate all yi using corresponding public key pki 1) Tddl : The deadline for a sensor to send its sensing packet
and xi . to CSC.
2) Calculate the symmetric key k = H(msg). 2) ds : The deposit that a sensor needs to pay before par-
3) Verify that if Ck,v (y1 , y2 , . . . , yn ) = v holds. ticipating in sensing, which is used as a guarantee for
If the verification succeeds, the smart contract recognizes acting honestly. The deposit can be withdrawn only if
that msg is sent by a valid sensing node from the group the sensor uploads the sensing result which is consistent
{Sensor1 , Sensor2 , . . . , Sensorn }. In this way, we can cut the to the cooperative sensing result.
connection between the data packet and its corresponding 3) N1 : The maximum number of sensors needed in coop-
owner. erative sensing. If there are more than N1 sensors who
However, for the trust value update and payments assign- apply to participate in cooperative sensing, those with
ment process, it is necessary to know the mapping of the N1 highest trust value will be selected.
sensing result and its corresponding sensor. To this end, we 4) TVthr : The minimum trust value of a sensor is needed
design a two stage commitment scheme as follows. to participate in cooperative sensing.
This scheme consists of two stages, including Commit Stage The functions in CSC are introduced as follows.
and Reveal Stage. As shown in Fig. 4, at the commit stage, 1) SensorRegister(·): The inputs to this function are address
each sensor participating in the cooperative sensing needs to Addr, deposit dpt, and trust value TV of the node who
submit the msg whose privacy is protected by the ring signa- invokes this function. This function will decide whether
ture, and the field “hash of msgID,” i.e., H(msgID), is used this node can register successfully by checking its trust
as a commitment. msgID is an identifier which is the input value.
of hash function, and can be any message such as “I am 2) CheckRegisterQuality(·): This function is used to check
User 3” or “I like apples.” At the reveal stage, each sensor node’s quality of participating in sensing by checking
directly uploads the unhashed msgID for miners to verify the the parameters passed into it, i.e., node’s trust value
consistency. This checking mechanism is effective since hash TVi and deposit depositi . For new users, participating
function is preimage resistance and second-preimage resis- in sensing is the only way to improve their trust value.
tance [42], i.e., given a hash output R = H(msgID), it is This function offers users a way to convert their tokens
difficulty to find the input msgID or another input msgID to additional trust value, which enables the users whose
such that H(msgID) = R or H(msgID ) = R. trust value is under threshold to participate in sensing
by paying more deposit.
VI. S MART C ONTRACTS AND DSA P ROTOCOL D ESIGN 3) Fusion(·): The input to this function is msgList,
Smart contracts are computer codes that run on the which is a list consisting of all the sensing results
blockchain platform, which are automatically executed when from the registered sensors. This function outputs
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13271

the final sensing result according to the predefined Algorithm 2 Cooperative Sensing Contract
fusion rule. 1: function INIT (TVthr , ds , N1 )
4) UploadSensingData(·): The inputs to this function are 2: require(msg.sender==ContractOwner);
msg and RSIG , which are the sensing packets in (10) 3: initialize TVthr , ds , N1 ;
and the ring signature of the node, respectively. This 4: end function
function is invoked by legal sensors to upload their 5:
sensing result. 6: function S ENSOR R EGISTER(Addr, dpt, TV)
The pseudo-code of the CSC is summarized in Algorithm 2. 7: sensornum ← 0;
Spectrum Auction Contract (SAC): The parameters in SAC 8: if CHECK R EGISTER Q UALITY(TVi , addri ) then
are as follows. 9: sensorMap ← sensori ;
1) CSCid : The identity of CSC, which is used to identify 10: sensornum ← sensornum + 1;
the connection between SAC and CSC since every SAC 11: EMIT event("Registration success!");
is associated with a particular CSC. 12: else
2) N2 : The maximum number of bidders, which is used to 13: EMIT event("Registration failed, please checkout
prevent too many nodes from sending bidding message. the trust value and deposit.");
3) Tself−d : The time that SAC conduct the self-destruct 14: end if
operation to release memory. 15: if sensornum > N1 then
4) da : The amount of tokens that bidders need to deposit 16: Select top N1 sensors according to trust value;
before the final sensing result is released. 17: end if
The functions in SAC are introduced as follows. 18: end function
1) BidderRegister(·): Bidders invoke this function to regis- 19:
ter in SAC. 20: function F USION(msgList)
2) Commit(·): The input to this function is bldBid, which is 21: Decision fusion in majority rule;
the commitment of bidderi . Note that since the account 22: return Cooperative sensing result;
balance is transparent, nodes can identify others’ bid 23: end function
by checking their balance [17]. Therefore, everyone 24:
makes several commits in order to prevent others from 25: function CHECK R EGISTER Q UALITY(TVi ,depositi )
inferring your bid based on changes of their account 26: if depositi > ds then
balance. 27: TVi ← convert(depositi − ds )
3) Reveal(·): The inputs of this function include Dps, 28: if (TVi + TVi ) > TVthr then
Bools, and RNDs. Dps is the list of deposit (Dp) the 29: if totalNum < N1 or TVi > lowestTV then
bidders make; Bools is the list of boolean values which 30: Register Successful;
indicate whether these bids are valid or not; RNDs 31: end if
are random values to make the hash of {Dp, Bool, 32: else
RND} hard to guess. A Commit is the hash of these 33: Register Failed;
three parameters, i.e., Commit = H{Dp, Bool, RND}. 34: end if
bidsList[msg.sender] is the list of the invoker’s commits. 35: end if
By comparing the reveal messages and the commit- 36: end function
ments, this function can identify which bids are revealed 37:
correctly. Valid bids are added together as the bidder’s 38: function U PLOAD S ENSING DATA(msg, RSIG)
total bid, invalid but correctly revealed bids are allowed 39: if RSIG is legal then
to be withdrawn, bids that are not correctly revealed will 40: msgList ← msg;
not be returned. 41: EMIT event("Upload successfully!");
4) Win(·): The input to this function is bidderList, which 42: else
is the list of all valid bids. Here, we adopt the second- 43: EMIT event("Upload failed, illegal sensor!");
price sealed-bid auction for better economic profit [43]. 44: end if
Accordingly, this function selects the bidder with the 45: end function
highest bid as the winner, who only needs to pay the
second-highest bid.
5) EndOfAuction(·): SAC will execute self-destruct opera- 1) Phase 1 (Spectrum Sensing Request): To ask for the
tion at the preset time Tself−d . channel state, SUs need to send a request message to
The pseudo-code of SAC is shown in Algorithm 3. Task Issuer (TI) through the control channel. TI is played
by the node whose trust value is the highest at present.
The role of TI is set to prevent multiple contracts from
B. DSA Protocol appearing in the network, which results in that users may
In this part, we illustrate our proposed blockchain-based participate in different contracts. TI will then creates
DSA protocol. To make it clearer, we elaborate this protocol and deploys the CSC and corresponding SAC in the
in Fig. 5. blockchain.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13272 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

Algorithm 3 Sealed Spectrum Auction Contract

1: Bid ← bldBidi , msg.value;
2: bidsList ← mapping(address => Bid[]);
3: refund ← 0
4:
5: function C OMMIT(bldBid)
6: bidsList[msg.sender].push(bldBid);
7: end function
8:
9: function R EVEAL(Dps, Bools, RNDs)
10: len ← bidsList[msg.sender].length;
11: while len > 0 do
12: Commit ← H{Dps[len], Bools[len], RNDs[len]};
13: if Commit! = bidsList[msg.sender][len] then
14: EMIT event("Illegal Reveal!");
15: Continue; Fig. 5. Workflow of our proposed blockchain-based DSA.
16: else if Commit == bidsList[msg.sender][len] and
Bools[len] == true then
17: refund ← refund + Dps[len]; 3) Phase 3 (Players Selection): Players (including sensors
18: else if Commit == bidsList[msg.sender][len] and and bidder) are selected based on SensorRegister(·) and
Bools[len] == false then BidderRegister(·), respectively. Finally, an event will be
19: withdraw invalid but reveal correctly bid; triggered to provide an appropriate notification to every
20: end if selected nodes.
21: len ← len − 1; 4) Phase 4 (Sensing Phase): Sensors upload the sensing
22: end while data by issuing the transaction
23: end function  
msgtx = CSCid | H(msgID) | TS | SR | loc (17)
24:
25: function W IN(bidderList) with the ring signature generated by (12) before Tddl , and
26: return Bidder with the highest bid; TS denotes timestamp, SR denotes the sensing result. In
27: end function this process, we use the ring signature to protect the
28: SU’s identity. Meanwhile, these sensors need to make
29: function E NDOFAUCTION(Tself−d ) commitments mentioned in Section V-D by sending the
30: Execute self-destruct operation; transaction
31: end function  
Commiti = CSCid | H(SR, RNDi , msgID) | pki (18)

with Sigski (Commiti ). After the sensing deadline, data

2) Phase 2 (Smart Contracts and Nodes Register): The fusion is done, and the final sensing result will be
CSC and SAC are instantiated when TI issues the published in the blockchain network.
transaction 5) Phase 5 (Auction if Necessary): If the cooperative sens-
CSC = {CSCid | Tddl | N1 | TVthr | Fusion(·) | ds } (13) ing result indicates that the corresponding spectrum
bands are idle, the SAC will be invoked. Users who
with the signature SigTI (STC), and the transaction have been selected will reveal their bids for the spec-
SAC = {CSCid | SACid | N2 | Tself-d | Win(·) | da } (14) trum by sending bids and random numbers. Then the
winner will be published on the blockchain.
with the signature SigTI = (AC). To register to CSC, a 6) Phase 6 (Trust Value Updating): In this phase, the trust
sensor issues the transaction value of each node is updated using (5) by miners
in the network. It is worth mentioning that the trust
DepositCSCid = {pki | TVi | CSCid | ds } (15)
value updating process does not compromise the decen-
with the signature Sigski (DepositCSCid ) to make deposit. tralization of the system because the smart contract is
Similarly, an SU who intends to participate in SAC decentralized.
makes its deposit by issuing
  VII. I MPLEMENTATION AND P ERFORMANCE A NALYSIS
BtySACid = pki | TVi | SACid | da | H(bidi , RNDi )
(16) In this section, we first introduce and analyze the work-
flow of our designed smart contracts, and then we analyze the
with the signature Sigski (BtySACid ). The H(bidi , RNDi ) performance of the PoT consensus mechanism on the different
denotes the bidding commitment this SU makes for the type of nodes. Next, we analyze the trust evaluation mecha-
sealed auction. nism to improve the accuracy of cooperative spectrum sensing.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13273

TABLE II TABLE III

R ELATED I NFORMATION OF S ENSOR ACCOUNTS R ELATED I NFORMATION OF B IDDERS

Fig. 6. Initialization of smart contract.

Fig. 7. Commit made by bidder 2.

Finally, we provide the analysis of the incentive mechanism

and security issues of our system.

A. Implementation of Smart Contract

We implement the proposed smart contracts using
Solidity [17] in the Ethereum virtual machine (EVM). The
smart contracts are compiled and deployed in Remix IDE [44]
which is used for testing. After that, nodes in the blockchain
network can send transactions to deploy the smart contracts
and invoke functions of them. In the cooperative sensing
phase, five sensors are considered and the information of their
accounts in the blockchain network are listed in Table II. In
the following, we describe the implementation procedures in
detail.
1) Smart Contract Preparation: The smart contract is cre-
ated and deployed by TI, i.e., the fifth account in
Table II. The relevant parameters and functions which
need to be predefined are set as follows: N1 is set as
3, the trust value threshold TVthr is set as 900 where
we magnify the trust values by 100 times and make
they as integers, since fixed-point numbers are not fully
supported by Solidity yet. ds is set as 100 wei. Fig. 8. Reveal made by bidder 2.
2) Smart Contract Deployment: Fig. 6 shows the details
about the deployed smart contract. The field “from” is true bid with 100 wei and the other one is a false bid with
the account address of TI. The field “decoded input” 200 wei. The second bidder also has two bids: one is true with
shows all of our preset parameters. 150 wei and the other is false with 300 wei.
3) Sensor Registration: All the five sensors send their regis- 1) Commit: Bidders invoke Commit(·) function to make
tration request to the smart contract. Then, three sensors commitments. Fig. 7 shows the log of commitment made
are selected by the function SensorRegister(·). by bidder 2. There are address and bid information about
4) Sensing Results Fusion: The function Fusion(·), where this commit, thus others can know that bidder 2 make a
the majority rule is selected, is used to fuse the sensing bid with 150 wei, but they can not identify whether this
results from legal sensors. is a real commit or not.
If the spectrum is detected as idle, the auction phase initi- 2) Reveal: After the commit phase, bidders invoke the
ates. In this phase, we consider two bidders: 1) each bidder function Reveal(·) to reveal its commitments. The field
has one true bid and 2) one false bid. The related information “decoded input” in Fig. 8 shows the whole information
is listed in Table III. The first bidder has two bids: one is a about this commit. The smart contract will automatically
Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13274 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

Fig. 9. Running time (second) with increasing number of leading zeros.

Fig. 10. Curve of trust value.

verify the information uploaded in the two phase and the

final bidding result can be accessed by all nodes in the
system.

B. Performance Analysis of the Proposed PoT Consensus

Algorithm
In this part, we evaluate the performance of the proposed
PoT consensus algorithm. We first discuss how to set the ini-
tial mining difficulty β0 at the beginning when every node’s
trust value is 0. We simulate the mining process in our per-
sonal computer as a reference to explain how to select a
suitable mining difficulty. We use a given string to represent
the transactions of a block. The result is shown in Fig. 9.
The running time increases exponentially when the number of
leading zeros is larger than 20. Supposing that the hashrate of Fig. 11. Expected mining consumption of different nodes.
the IoT devices is similar to that of our personal computer, the
interval between two blocks T0 is about 1 s, then the mining in sensing infrequently. OOnode is a node who conducts mali-
difficulty should be less than 18 leading zeros, the β0 is about cious behavior periodically. We assume that OOnode will act
218 = 262 144. maliciously after every two normal sensing rounds. The Lnode
It is worth mentioning that even under the same mining is a node who randomly provides binary sensing result with-
difficulty and experimental environment, the mining time may out sensing. There are 12 Rnodes with probability of detection
be different. For example, given the mining difficulty where pd = 0.90 and probability of false alarm pf = 0.15 [45]; three
32 leading zeros are needed in the target hash, it sometimes OOnodes whose pd = 0.90 and pf = 0.15; three Lnodes with
take tens of seconds, and sometimes take more than an hour. pd = 0.5 and pf = 0.5; and two UAnodes with pd = 0.90 and
In order to eliminate the impact of luck, we convert the min- pf = 0.15. Besides, it is assumed that they only have a 50%
ing difficulty into the expected mining cost and evaluate the chance to sign up in CSS each sensing round.
expected mining cost instead of running time. Theoretically, Fig. 10 shows the curve of trust values of different types of
assuming that the output of each hash calculation is unpre- nodes over time. It can be seen that the trust values of Rnodes
dictable, each time when we increase the leading zero by one, are significantly higher than the trust values of other types
the success probability of each hash trial will be halved. The of nodes. Therefore, the mining difficulty required by Rnodes
trust value is associated with the mining difficulty. Therefore, will also be lower than that of other types of nodes. In Fig. 11,
we first convert the trust value into the number of leading we can see that the mining cost of Rnodes is only one third
zeros, and then the number of leading zeros is converted of other nodes. Thus, they can be three times faster than PoW
into the expected mining cost proportionally. Simulation is in the mining process, where the mining difficulty is the same
conducted for 1000 time slots, and the average expected min- for all types of nodes. This proves the effectiveness of our
ing cost of every type of nodes is calculated in Fig. 11. To proposed consensus algorithm.
be specific, we simulate a network with 20 nodes and four
types of nodes are considered. Rnode means a node with high C. Performance of Cooperative Sensing
performance and always act honestly; UAnode means a node In this part, we evaluate the performance of coopera-
with high performance and act honestly, however it participates tive spectrum sensing with the proposed trust evaluation

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13275

(a) (b)

Fig. 12. Performance of cooperative sensing under three selection schemes. (a) Pd of cooperative sensing. (b) Pf of cooperative sensing.

mechanism. In the traditional blockchain network, there is This incentive mechanism can not only encourage the nodes
no trust among nodes hence anyone can record the sens- to participate in spectrum sensing, but also encourage them to
ing information into blockchain, including nodes with poor behave honestly and accurately in spectrum sensing. This is
performance or malicious behavior. However, due to the intro- because, on the one hand, an honest and accurate sensing node
duction of the trust evaluation mechanism, the CSC can is more likely to derive a sensing result that is consistent to
exclude bad nodes according to nodes’ trust value, thus the the final cooperative sensing result and obtain tokens rewarded
system’s sensing performance can be improved. For compar- for spectrum sensing. On the other hand, with the proposed
ison, we consider three kinds of selection schemes to select consensus algorithm, the node with a higher trust value is eas-
the candidate nodes for the cooperative sensing: 1) random ier to mine successfully so that they have a higher probability
selection scheme; 2) select according to register time; and of obtaining tokens for mining. On the contrary, the dishonest
3) select according to sensor’s trust value. We simulate a 20 nodes will be discouraged since they are less likely to obtain
nodes network with the same setup in Section VII-B. It can rewards since they need to spend more computing resources
be seen from Fig. 12 that when the number of needed sensors on mining.
N1 is small, pd and pf of cooperative sensing in the first two
schemes which do not take advantage of trust value perform
worse than that of the last selection scheme. This indicates that E. Security Analysis
our proposed trust value mechanism can effectively improve 1) Punish rules: First, PoT is highly depends on the trust
the cooperative sensing performance of the system. Moreover, evaluation mechanism. If a node behaves dishonestly,
in the last selection scheme, pd is close to 1 when N1 is about its trust will drop because of the negative factor Ni,w (n)
one fourth of all network nodes. The fewer nodes involved, in (1). Second, the deposit mechanism is also a part of
the smaller the total monetary reward that the network needs our punish rules. If a node behave maliciously in sensing
to pay to sensors, and the smaller the economic burden for or auction process, the corresponding deposit will be
spectrum buyers. As N1 grows, the performance of different confiscated.
schemes will gradually close, since most of the network nodes 2) Distributed Denial of Service (DDoS) Attack: The DDoS
(including bad nodes) will participate in cooperative sensing. attack here means that malicious users try to make the
sensing service unavailable to other users. Our system
D. Incentive Mechanism Analysis is resist to this attack since we adopt the deposit mech-
Since both sensing and mining will consume SU’s com- anism. Thus, under our scheme, the cost of launching
puting power, an incentive mechanism is needed to reward large-scale DDoS attack is very high since the attack-
nodes for their work. The users who upload a sensing result ers need to obtain lots of tokens in order to launch the
that is consistent to the final cooperative sensing result will attack.
be rewarded with Rs tokens; the users who successfully mine 3) Sybil Attack: One node can create many accounts in pub-
will be rewarded with Rm tokens. The tokens can be used in lic blockchain. However, the initial trust of each account
auction to bid for spectrum resources. In implementation, the is zero, if new accounts want to compete with high-trust
tokens rewarded for accurate spectrum sensing and success- accounts in being selected into sensing group or mining,
ful mining, i.e., Rs and Rm , need to be designed based on the they must accumulate enough trust through providing
evaluation of computation consumption of spectrum sensing correct sensing results. In this way, we increase the cost
and mining. of Sybil attacks.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13276 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

(a) (b)

Fig. 13. Trust value of different types of nodes under two model. (a) Existing model. (b) Proposed model.

4) Collusion Attack: Malicious nodes must undergo a pro- proposed model, as show in Fig. 13, the unintentional
cess of trust accumulation to become a trusted user, mistakes made by sensors can be compensated by mak-
during which they need to consume resources and under ing the right sensing decision. When the parameters τ
the risk of losing deposit. However, collusion in sens- and η satisfy ρ > [η/(1 − e−η )] − η, our proposed trust
ing cannot make them dominate spectrum directly, and evaluation mechanism is considered to be resist to the
collusion in mining also cannot launch forking attack. on-off attack, the proof is given in the Appendix.
Hence, nodes do not have sufficient motivation to accu-
mulate high trust and then launch such a kind of
attack. VIII. C ONCLUSION
5) Spoofing Attack: Spoofing attack means someone tries
In this article, we have proposed a blockchain-based
to masquerade others to create forged transactions.
dynamic spectrum sharing protocol. This protocol mainly con-
The secure elliptic curve digital signature algorithm
sists of three parts: 1) the first part is the trust evaluation
(ECDSA) [46] used in our blockchain can prevent this
mechanism, which is designed to evaluate the credibility of
attack on the premise that attack does not have the user’s
nodes; 2) the second part is the PoT consensus algorithm,
private key.
which makes the mining difficulty of malicious nodes greatly
6) Free-Riding Attack (Lazy Node): Free-riding attack
increased. The combination of a node’s trust value and its min-
means lazy users may directly copy others’ sensing
ing difficulty can motivate it to behave honestly; and 3) the
results at the phase of uploading sensing data. First,
third part is the privacy protection mechanism, in which we
there is no motivation for sensors to submit the sensing
combine the ring signature and the commitment scheme to
result before the sensing deadline. Thus, when the lazy
solve the problem of privacy leakage in the process of coop-
nodes get the sensing result, it is difficult for them to
erative spectrum sensing. Finally, a prototype of our proposed
repack the sensing result and submit before the deadline.
smart contracts has been implemented, and the performance of
Second, even if a few sensors submit sensing results in
the PoT consensus algorithm and the improvement in coop-
advance, the connection between a user’s identity and
erative sensing has been analyzed. Security analysis of the
sensing data is cut by ring signature, thus the lazy users
system shows that our framework can resist many types of
cannot determine the owner of the sensing data, and thus
attacks which are common in trust management mechanism
the credibility of data cannot be guaranteed. Thus, it is
and blockchain-based IoT systems.
no better than submit a sensing result randomly.
7) On-Off Attack: On-off attack means that a node performs
malicious behaviors periodically, and pretends to be a
normal node who “unintentionally” sends sensing result A PPENDIX
incorrectly from time to time. If a trust management P ROOF OF THE R ESISTANCE OF O N -O FF ATTACK
mechanism satisfy the condition that the dropping rate Theorem 1: The proposed trust model given in (2) is resis-
of trust value is larger than its increasing rate, it is con- tant to on-off attack when ρ > [η/(1 − e−η )] − η.
sidered to be resist to the on-off attack [38]. However, Proof: Let f (Nr , Nw ) = e−ρ·Nw · (1 − e−η·Nr ) denotes the
the model in [38] may cause the trust value drop a lot trust value update function in (2). Because when Nr = 0,
even the misbehavior is unintentional, since the increas- f (Nr , Nw ) can only increase, this situation is not discussed
ing rate is much smaller than the dropping rate. In our here. At any other points (Nr , Nw ) in this function, the

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
YE et al.: TRUST-CENTRIC PRIVACY-PRESERVING BLOCKCHAIN FOR DYNAMIC SPECTRUM MANAGEMENT 13277

decreasing rate should be larger than the increasing rate, that is [14] T. Ariyarathna, P. Harankahadeniya, S. Isthikar, N. Pathirana, H. M.
    N. D. Bandara, and A. Madanayake, “Dynamic spectrum access via
 ∂f   ∂f 
 >  (19)
smart contracts on blockchain,” in Proc. IEEE Wireless Commun. Netw.
 ∂N   ∂N  Conf. (WCNC), 2019, pp. 1–6.
w r
[15] S. Bayhan, A. Zubow, and A. Wolisz, “Spass: Spectrum sensing as
where a service via smart contracts,” in Proc. IEEE Int. Symp. Dyn. Spectr.
  Access Netw. (DySPAN), 2018, pp. 1–10.
 ∂f   
  = ρ · e−ρ·Nw · 1 − e−η·Nr (20) [16] X. Kang, Y.-C. Liang, and J. Yang, “Riding on the primary: A new
 ∂N  spectrum sharing paradigm for wireless-powered IoT devices,” in Proc.
w
IEEE Int. Conf. Commun. (ICC), 2017, pp. 1–6.
and [17] “Solidity Webpage.” 2020. [Online]. Available: https://
 
 ∂f    docs.soliditylang.org/en/v0.7.5/
  = η · e−ρ·Nw · e−η·Nr (21) [18] N. Szabo, “Formalizing and securing relationships on public networks,”
 ∂N  First Monday, vol. 2, no. 9, Sep. 1997. [Online]. Available:
r
https://firstmonday.org/ojs/index.php/fm/article/download/548/469
then we have [19] R. L. Rivest, A. Shamir, and Y. Tauman, “How to leak a secret,” in
   
 ∂f   ∂f  −η·Nr Proc. Int. Conf. Theory Appl. Cryptol. Inf. Security, 2001, pp. 552–565.
 > ⇔ ρ > e (22) [20] O. Goldreich, Foundations of Cryptography, vol. 2. Cambridge, U.K.:
 ∂N   ∂N  η 1 − e−η·Nr Cambridge Univ. Press, 2004.
w r
[21] K. Kotobi and S. G. Bilen, “Secure blockchains for dynamic spectrum
which can also be denoted as
    access: A decentralized database in moving cognitive radio networks
 ∂f   ∂f  ρ 1 enhances security and user access,” IEEE Veh. Technol. Mag., vol. 13,
 > 
 ∂N   ∂N  ⇔ η > 1 − e−η·Nr − 1. (23) no. 1, pp. 32–39, Mar. 2018.
w r [22] S. Zheng, T. Han, Y. Jiang, and X. Ge, “Smart contract-based spec-
trum sharing transactions for multi-operators wireless communication
When Nr ≥ 1 networks,” IEEE Access, vol. 8, pp. 88547–88557, 2020.
1 1 [23] X. Fan and Y. Huo, “Blockchain based dynamic spectrum access of non-
−1≤ − 1. (24) real-time data in cyber-physical-social systems,” IEEE Access, vol. 8,
1 − e−η·Nr 1 − e−η pp. 64486–64498, 2020.
Therefore, as long as it is satisfied (ρ/η) > [1/(1 − e−η )] − 1, [24] H. Zhang, S. Leng, and H. Chai, “A blockchain enhanced dynamic spec-
trum sharing model based on proof-of-strategy,” in Proc. IEEE Int. Conf.
the above inequality holds. Commun. (ICC), 2020, pp. 1–6.
[25] X. Ling, J. Wang, T. Bouchoucha, B. C. Levy, and Z. Ding, “Blockchain
radio access network (B-RAN): Towards decentralized secure radio
R EFERENCES access paradigm,” IEEE Access, vol. 7, pp. 9714–9723, 2019.
[1] Global Mobile Data Traffic Forecast, “Cisco visual networking index: [26] S. King and S. Nadal, “PPcoin: Peer-to-peer crypto-currency with proof-
Global mobile data traffic forecast update, 2017–2022,” Cisco, San Jose, of-stake,” self-published paper, vol. 19, no. 1, Aug. 2012.
CA, USA, document, 2019. [27] “NXT Whitepaper.” 2014. [Online]. Available:
[2] P. Kolodzy and I. Avoidance, “Spectrum policy task force,” Federal https://wiki.nxtcrypto.org/wiki/Whitepaper:Nxt#Proof_of_Stake
Commun. Comm., Washington, DC, USA, Rep. ET Docket 02-155, [28] I. Bashir, Mastering Blockchain: Distributed Ledger Technology,
2002. Decentralization, and Smart Contracts Explained. Birmingham, U.K.:
[3] I. F. Akyildiz, B. F. Lo, and R. Balakrishnan, “Cooperative spectrum Packt Publ. Ltd., 2018.
sensing in cognitive radio networks: A survey,” Phys. Commun., vol. 4, [29] “BitShares Delegated Proof of Stake.” 2014. [Online]. Available:
no. 1, pp. 40–62, 2011. https://github.com/BitShares/bitshares/wiki/Delegated-Proof-of-Stake
[4] J. Wang, M. Li, Y. He, H. Li, K. Xiao, and C. Wang, “A blockchain based [30] I. Grigg, “EOS—An introduction,” EOS, Krailling, Germany, docu-
privacy-preserving incentive mechanism in crowdsensing applications,” ment, 2017. [Online]. Available: https://whitepaperdatabase.com/eos-
IEEE Access, vol. 6, pp. 17545–17556, 2018. whitepaper
[5] L. Xiao et al., “A reinforcement learning and blockchain-based trust [31] G. Wood, “Polkadot: Vision for a heterogeneous multi-chain frame-
mechanism for edge networks,” IEEE Trans. Commun., vol. 68, no. 9, work,” Ethereum, Zug, Switzerland, document, 2016.
pp. 5460–5470, Sep. 2020. [32] M. Salimitari and M. Chatterjee, “An overview of blockchain and
[6] G. Liu, H. Dong, Z. Yan, X. Zhou, and S. Shimizu, consensus protocols for IoT networks,” 2018, arXiv:1809.05613.
“B4SDC: A blockchain system for security data collection in
[33] M. A. A. Careem and A. Dutta, “SenseChain: Blockchain based repu-
MANETs,” IEEE Trans. Big Data, early access, Mar. 17, 2020,
tation system for distributed spectrum enforcement,” in Proc. IEEE Int.
doi: 10.1109/TBDATA.2020.2981438.
Symp. Dyn. Spectr. Access Netw. (DySPAN), 2019, pp. 1–10.
[7] W. Feng and Z. Yan, “MCS-Chain: Decentralized and trustworthy
[34] A. Vosoughi, J. R. Cavallaro, and A. Marshall, “Trust-aware consensus-
mobile crowdsourcing based on blockchain,” Future Gener. Comput.
inspired distributed cooperative spectrum sensing for cognitive radio
Syst., vol. 95, pp. 649–666, Jun. 2019.
ad hoc networks,” IEEE Trans. Cogn. Commun. Netw., vol. 2, no. 1,
[8] M. Li et al., “CrowdBC: A blockchain-based decentralized framework
pp. 24–37, Mar. 2016.
for crowdsourcing,” IEEE Trans. Parallel Distrib. Syst., vol. 30, no. 6,
pp. 1251–1266, Jun. 2019. [35] H. L. J. Ting, X. Kang, T. Li, H. Wang, and C.-K. Chu, “On the
[9] M. Yang, T. Zhu, K. Liang, W. Zhou, and R. H. Deng, “A blockchain- trust and trust modeling for the future fully-connected digital world: A
based location privacy-preserving crowdsensing system,” Future Gener. comprehensive study,” IEEE Access, vol. 9, pp. 106743–106783, 2021.
Comput. Syst., vol. 94, pp. 408–418, May 2019. [Online]. Available: [36] V. Buterin. “Thoughts on UTXOs.” 2016. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0167739X18320909 http://timmurphy.org/2009/07/22/line-spacing-in-latex-documents/
[10] M. B. H. Weiss, K. Werbach, D. C. Sicker, and C. E. C. Bastidas, “On [37] G. Wood, “Ethereum: A secure decentralised generalised transaction
the application of blockchains to spectrum management,” IEEE Trans. ledger,” Ethereum, Zug, Switzerland, document, 2014.
Cogn. Commun. Netw., vol. 5, no. 2, pp. 193–205, Jun. 2019. [38] X. Kang and Y. Wu, “A trust-based pollution attack prevention scheme
[11] E. D. Pascale, J. McMenamy, I. Macaluso, and L. Doyle, “Smart contract in peer-to-peer streaming networks,” Comput. Netw., vol. 72, pp. 62–73,
slas for dense small-cell-as-a-service,” 2017, arXiv:1703.04502. Oct. 2014. [Online]. Available: http://www.sciencedirect.com/science/
[12] G. Qiao, S. Leng, H. Chai, A. Asadi, and Y. Zhang, “Blockchain empow- article/pii/S1389128614002667
ered resource trading in mobile edge computing and networks,” in Proc. [39] D. Boneh, B. Bünz, and B. Fisch, “Batching techniques for accumulators
IEEE Int. Conf. Commun. (ICC), 2019, pp. 1–6. with applications to IOPs and stateless blockchains,” in Proc. Annu. Int.
[13] J. Qiu, D. Grace, G. Ding, J. Yao, and Q. Wu, “Blockchain-based Cryptol. Conf., 2019, pp. 561–586.
secure spectrum trading for unmanned-aerial-vehicle-assisted cellular [40] G. Ateniese, B. Magri, D. Venturi, and E. Andrade, “Redactable
networks: An operator’s perspective,” IEEE Internet Things J., vol. 7, blockchain—or—Rewriting history in bitcoin and friends,” in Proc.
no. 1, pp. 451–466, Jan. 2020. IEEE Eur. Symp. Security Privacy (EuroS P), 2017, pp. 111–126.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.
13278 IEEE INTERNET OF THINGS JOURNAL, VOL. 9, NO. 15, 1 AUGUST 2022

[41] S. Li, H. Zhu, Z. Gao, X. Guan, K. Xing, and X. Shen, “Location Ying-Chang Liang (Fellow, IEEE) received the
privacy preservation in collaborative spectrum sensing,” in Proc. IEEE B.S.E.E. and Ph.D. degrees in electrical engineer-
INFOCOM, 2012, pp. 729–737. ing from Jilin University of Technology, Changchun,
[42] P. Rogaway and T. Shrimpton, “Cryptographic hash-function basics: China, in 1989 and 1993, respectively.
Definitions, implications, and separations for preimage resistance, He is currently a Professor with the University
second-preimage resistance, and collision resistance,” in Proc. Int. of Electronic Science and Technology of China,
Workshop Fast Softw. Encryption, 2004, pp. 371–388. Chengdu, China, where he leads the Center for
[43] Y.-C. Liang, Blockchain for Dynamic Spectrum Management. Intelligent Networking and Communications. He
Singapore: Springer, 2020, pp. 121–146. [Online]. Available: was a Professor with the University of Sydney,
https://doi.org/10.1007/978-981-15-0776-2_5 Sydney, NSW, Australia, a Principal Scientist and
[44] “Remix-Ethereum IDE.” 2020. [Online]. Available: https://remix- a Technical Advisor with the Institute for Infocomm
ide.readthedocs.io/en/latest/ Research, Singapore, and a Visiting Scholar with Stanford University,
[45] X. Kang, Y.-C. Liang, H. K. Garg, and L. Zhang, “Sensing-based spec- Stanford, CA, USA. His research interests include wireless networking and
trum sharing in cognitive radio networks,” IEEE Trans. Veh. Technol., communications, cognitive radio, symbiotic communications, dynamic spec-
vol. 58, no. 8, pp. 4649–4654, Oct. 2009. trum access, the Internet of Things, artificial intelligence, and machine
[46] D. Johnson, A. Menezes, and S. Vanstone, “The elliptic curve digi- learning techniques.
tal signature algorithm (ECDSA),” Int. J. Inf. Security, vol. 1, no. 1, Prof. Liang is a recipient of numerous paper awards, including the IEEE
pp. 36–63, 2001. Jack Neubauer Memorial Award in 2014 and the IEEE Communications
Society APB Outstanding Paper Award in 2012. He received the Prestigious
Engineering Achievement Award from the Institution of Engineers, Singapore,
in 2007, the Outstanding Contribution Appreciation Award from the IEEE
Standards Association in 2011, and the Recognition Award from the IEEE
Communications Society Technical Committee on Cognitive Networks in
2018. He has been recognized by Thomson Reuters (now Clarivate Analytics)
as a Highly Cited Researcher since 2014. He is the Founding Editor-in-
Jingwei Ye (Graduate Student Member, IEEE) Chief of the IEEE J OURNAL ON S ELECTED AREAS IN C OMMUNICATIONS :
received the B.S. degree in communication engi- C OGNITIVE R ADIO S ERIES, and the Key Founder and currently the Editor-
neering from the University of Electronic Science in-Chief of the IEEE T RANSACTIONS ON C OGNITIVE C OMMUNICATIONS
and Technology of China, Chengdu, China, in 2019, AND N ETWORKING . He is also serving as an Associate Editor-in-Chief
where he is currently pursuing the master’s degree. for China Communications. He was a Guest/Associate Editor of the IEEE
His current interests include blockchain technolo- T RANSACTIONS ON W IRELESS C OMMUNICATIONS, the IEEE J OURNAL
gies, dynamic spectrum management, and wireless OF S ELECTED A REAS IN C OMMUNICATIONS , the IEEE Signal Processing
communication systems. Magazine, the IEEE T RANSACTIONS ON V EHICULAR T ECHNOLOGY, and
the IEEE TRANSACTIONS ON SIGNAL AND I NFORMATION P ROCESSING
OVER N ETWORK. He was also an Associate Editor-in-Chief of the World
Scientific Journal on Random Matrices: Theory and Applications. He was a
Distinguished Lecturer of the IEEE Communications Society and the IEEE
Vehicular Technology Society. He was the Chair of the IEEE Communications
Society Technical Committee on Cognitive Networks and served as the TPC
Chair, and an Executive Co-Chair of the IEEE Globecom’17. He is a Foreign
Member of Academia Europaea.
Xin Kang (Senior Member, IEEE) received the
B.Eng. degree in electrical engineering from Xi’an
Jiaotong University, Xi’an, China, in 2005, and
the Ph.D. degree in electrical and computer engi-
neering from the National University of Singapore,
Singapore, in 2011. Sumei Sun (Fellow, IEEE) received the B.Sc. degree
He was a Research Scientist with the Institute (Hons.) from Peking University, Beijing, China,
for Infocomm Research, A*STAR, Singapore, from the M.Eng. degree from Nanyang Technological
2011 to 2014. After that, he joined Shield University, Singapore, and the Ph.D. degree from
Laboratory, Huawei Singapore Research Center, as the National University of Singapore, Singapore.
a Senior Researcher. He is currently a Full Professor She is a Principal Scientist, an Acting Deputy
with the University of Electronic Science and Technology of China, Chengdu, Executive Director (Research), and the Head of
China. He has published over 70 top-tier journal and conference papers and the Communications and Networks Department,
over ten of them are listed as SCI highly cited research papers. Up to now, he Institute for Infocomm Research, Singapore. She is
has filed more than 50 patents, and contributed more than 30 technical propos- also holding a joint appointment with the Singapore
als to 3GPP SA3, and 17 of his proposals have been accepted by 3GPP SA3. Institute of Technology, and an adjunct appointment
He has more than ten years’ research experience and his research interests with the National University of Singapore, both as a Full Professor. Her cur-
include but not limited to optimization, wireless communications, network rent research interests are in next-generation wireless communications, joint
security, trust modeling, digital identity, blockchain, security protocol design, sensing-communication-computing-control design, and Industrial Internet of
and applied cryptography. Things.
Prof. Kang has received the Best Paper Award from IEEE ICC 2017 and Dr. Sun is the Editor-in-Chief of IEEE O PEN J OURNAL OF V EHICULAR
Best 50 Papers Award from IEEE GlobeCom 2014. He is also very active T ECHNOLOGY, a Distinguished Speaker of the IEEE Vehicular Technology
in standardization. He is one of the key contributors to the newly published Society from 2018 to 2024, a Member-at-Large with the IEEE
ITU-T standard X.1365, and newly established work item X.ztd-iot. He is Communications Society from 2021 to 2023 and the IEEE Vehicular
also one of the key contributors to Huawei 5G security white papers. Technology Society Board of Governors from 2022 to 2024.

Authorized licensed use limited to: North West University. Downloaded on June 09,2023 at 09:12:10 UTC from IEEE Xplore. Restrictions apply.