This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access

DNSTSM: DNS Cache Resources Trusted

Sharing Model Based on Consortium
Blockchain
ZHONG YU1, DONG XUE1, JIULUN FAN1, CHANG GUO1
School of Communication and Information Engineering, Xi’an University of Posts and Telecommunications, Xi’an 710071, China
1

Corresponding author: Dong Xue (e-mail: dongxue@stu.xupt.edu.cn).

This work was supported in part by the National key research and development plan under Grant 2017YFC0803805, and in part by
the Shaanxi Provincial Key Research and Development Program of China under Grant 2018ZDCXL-GY-04-01.

ABSTRACT The inseparable internet requirements and the endless stream of cyber-attacks have led to
strong demand for trusted IP addresses. However, the existing collaborative DNS security schemes have the
defects of low credibility and imperfect incentive mechanism. Enlightened by the Consortium blockchain
technology, we propose a novel DNS Cache Resources Trusted Sharing Model, which can improve the
credibility of DNS resolution results by establishing a complete chain of trust. Firstly, the consortium
blockchain is introduced as the carrier of the peer-to-peer network to reduce the impact of illegal access and
complicity tampering on the DNS cache credibility; Secondly, the evaluation index of the node credibility
in the DNS cache sharing model is proposed, and the trust-based incentive mechanism is designed to reduce
the impact of free-riding behavior and on the trusted performance of the system. The two indicators of node
abnormal behavior similarity and roundtrip time between nodes are used to comprehensively evaluate the
degree of recommendation of the node and serve as the basis for dynamic scheduling; Finally, we use the
stochastic distributed decentralized storage mechanism to solve the problem of low efficiency in the
consortium blockchain. The simulation results show that the model has certain advantages in ensuring the
credibility of domain name resolution results, and maintains the ideal efficiency while ensuring trust.

INDEX TERMS Blockchain, Consortium blockchain, DNS security, Sharing Model

I. INTRODUCTION the payload and analyzing its authenticity based on the

At present, Domain Name System (DNS), as one of the
essential protocol in the Internet, maintains the interactive
mapping relationship between the Domain Name System
and the IP address space. However, there are two inherent
vulnerabilities in this network protocol, DNS resolvers trust
the response received after sending out queries, and users
usually trust the host-address mapping provided by the
DNS. Those vulnerabilities are easily exploited by
malicious attackers who can insert a fake IP address in the
Recursive Server cache [1], [2]. DNS cache poisoning is
often used as a primary source for developing more serious
secondary threats, such as phishing [3], masquerading
servers for middleman (MITM) attacks [4], malware
injection, and denial of service [5].
In order to improve the credibility of the domain name
resolution results, the researchers proposed a forgery
detection mechanism for DNS cache poisoning attacks. This
mechanism filters out forged DNS responses by examining
auxiliary information [6]. The Domain Name Cross
Referencing (DoX) [7] use the peer-to-peer network domain
name cross-referencing technology to detect cache poisoning
attacks. With this approach, Recursive Servers are linked to
each other through P2P (peer-to-peer) connections and allow
them to share a single piece of data called a verification
cache (vCache). The correct rate of the solution to improve
the resolution results depends on the trust between the nodes,
but the solution inevitably risks the free ride and malicious
collusion between the nodes. HARD-DNS [8] employs
arbitration techniques between peers to increase tolerance for
cache poisoning but has similar problems as DoX. CoDNS [9]
groups nodes that trust each other to assist each other's
queries when local domain name resolution fails, but its
security is weak because individual resolver error results can
easily be propagated throughout the peer-to-peer network.
Although ConfiDNS [10] improves the security of CoDNS,
their security mechanisms are not sensitive to Kaminsky's

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain
DNS cache poisoning attacks. H. Sun et al. [11] proposed the
DepenDNS protocol, but Alfardan et al. [12] found that it
was weaker against DNS spoofing attacks.
After studying the existing advanced solutions, we found
that they did not consider the system may be subject to
internal attacks, and there is room for improvement in the
ability to resist cache poisoning attacks. To deal with these
problems, we propose a DNS Cache Resources Trusted
Sharing Model based on consortium blockchain to achieve
high credibility of domain name resolution results. The main
contributions are as follows:
1) We propose to introduce the consortium blockchain into
the peer-to-peer network collaboration system for forgery
detection, and design an identity-based access control system
to eliminate potential threats caused by untrusted nodes. At
the same time, the model is divided according to the access
interest of the consortium members, which improves the
reuse rate of the cached data.
2) We propose a method for calculating the credibility of
DNS resolution nodes, which can evaluate the behavior of
malicious or later subverted nodes, and avoid them from
further affecting the system. Moreover, based on the level of
credibility, a dynamic scheduling mechanism is implemented
to motivate nodes to perform their own security upgrades.
3) We design a Stochastic Distributed Decentralized
Storage Algorithm (SDDSA) to group members in the same
federation, which provided conditions for further
implementing the multi-channel coroutine processing scheme
of DNS cache resources so that the DNSTSM system
satisfies the domain name resolution scenario for low latency
and high-reliability requirements.
The structure of this paper is as follows: The second part
introduces the consortium blockchain technology and the
related research progress of the current blockchain in the
sharing model field. In the third part, we introduce our
trusted sharing model. The fourth part analyzes security and
performance. We give the conclusion in the fifth part.
II. RELATED WORK
In this section, we present a brief review of the consortium
blockchain technology and applications of blockchain in
Trusted Sharing Model (TSM).
A. CONSORTIUM BLOCKCHAIN TECHNOLOGY
The consortium blockchain is only open to specific
organizational groups, and it is a blockchain that requires
registration permission [13]. The consortium blockchain is
limited to the participation of consortium members, and the
read and write permissions on the blockchain and the
participation accounting rights are specified according to
the consortium rules [14]. The consortium blockchain is
composed of one supernode and several common nodes. In
addition to the functions of common nodes, the supernode
also has the functions of implementing member
management, authority management, and data monitoring
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
in the alliance. Network access is generally accessed
through the gateway nodes of the member institutions, and
the consensus process is controlled by pre-selected nodes.
The public can consult and trade, and a consortium license
is required to verify the transaction or issue a smart contract.
Therefore, to a limited extent, only authorized nodes and
users can participate in the execution of transactions and
smart contracts, and any anonymous node or unauthorized
user is denied service. With the advent of the FISCO BCOS
[15], Ethereum Quorum [16], and Hyperledger Fabric [17],
consortium blockchain has become very popular and has
attracted more interest and investment in the world.
B. OVERVIEW OF APPLICATIONS OF BLOCKCHAIN IN
TRUSTED SHARING MODEL
In this section, we focus on several strategies to build a
trusted sharing model based blockchain and point out their
differences with our trusted sharing model for DNS caching.
Zhang et al. [18] proposed a distributed data security
sharing and storage system solution for vehicle self-
organizing network. In DSSCB, when the vehicle node
uploads the collected data, the integrity and security of the
data can be guaranteed by digital signature technology, and
the data coin is used to motivate the vehicle participating in
the data sharing. This solution effectively improves the
VANET identity validity and message reliability issues.
However, the data coins incentive method in this solution
does not bring intuitive benefits to sharers. In our solution,
trusted sharers can obtain safe, and efficient service.
Li et al. [19] proposed Meta-key, a data-sharing
mechanism that enables users to share their encrypted data
under a blockchain-based distributed storage architecture.
For the blockchain expansion problem, the scheme only
extracts and stores some metadata, and the data and
metadata are encrypted by the data owner. The solution
enables lightweight, secure storage and easy key
management. We also use lightweight data storage, but our
solution validates the data before it is stored.
Yang et al. [20] proposed a Differentially Private Data
Sharing scheme based on the blockchain. The solution
builds different cloud servers into a cloud federation for
interoperability and collaboration, and enables privacy
owners to control the anonymization process and enhance
service security through a privacy technology. The effect of
the privacy control of the data is achieved, but the
management of the cloud server in the alliance is lacking.
Unlike this approach, our shared model provides control
over the management of model members.
The current work on the blockchain in the DNS is mainly
focused on decentralizing the centralized authoritative
domain name servers [21], [22], [23], [24]. However, our
proposed DNSTSM is more concerned with the incremental
improvement of the existing DNS infrastructure, and the
credibility of resolution results is improved by the way that
peer-to-peer networks share trusted DNS cache resources.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain

FIGURE 1. DNSTSM network architecture diagram.

III. DNS CACHE RESOURCES TRUSTED SHARING

MODEL BASED ON CONSORTIUM BLOCKCHAIN
A. OVERALL FRAMEWORK AND THREAT MODEL FIGURE 2. The overall function module of DNSTSM.

The network structure of the solution is shown in Fig.1. TABLE 1. Threat model.
This solution is to maintain a set of validated A Records
between recursive servers in a peer-to-peer network, and to Membership management DNS security
isolate possible threats from trusted networks, ensuring the Cache
Permission out of bounds
tampering
credibility of users' domain name resolution results. Internal
Malicious logout Collusion attack
Fig.2 presents an overview of the function module of the
Free-riding
proposed DNS Cache Resources Trusted Sharing Model
Identity Cache
(DNSTSM), which is categorized into three parts, i.e. data Illegal
forgery poisoning
acquisition module, data storage module, service module. External access to
Evidence
In the data acquisition module section, we designed a set of the network
stealing
cache validation mechanisms. When the response returned
by the local Recursive Server is inconsistent with the
1) Collusion attack refers to the process of multiple
cached information in the cached federation or the domain
retransmission recursive servers jointly pointing the
name is resolved for the first time, the DNS Validator
resolution request to the same wrong IP address under
responds to the Multi-DNS Recursive Servers Lookup.
malicious control so that the forged IP is valid and written
Alert Manager is activated when Cache Validation Manager
into the Blockchain database. 2) Hitchhiking, in this article,
detects inconsistent resolution information to list it as high-
refers to the use of the system to obtain the credible results
risk domain name information. In the data storage module
of its own needs, but as a forwarding party, it does not
section, a peer-to-peer network consisting of different
maintain and upgrade its security.
recursive servers maintains a set of A records stored in
blockchain mode, and a Stochastic Distributed
B. CONSTRUCTION OF DNS CACHE RESOURCES
Decentralized Storage Scheme is designed to improve the TRUSTED SHARING MODEL
read and write efficiency of the system. In the service Member organization structure in DNSTSM network is
module section, the Dynamic Scheduling Mechanism shown in Fig.3. DNSTSM is established between recursive
provides an efficient way to assign forwarding Recursive servers with the same theme of interest. Depending on the
Server to all Recursive Servers. The Time-To-Live Scheme region, entities in the DNSTSM can be divided into several
maintains the real-time and effectiveness of the cached data organizations; depending on the service function, the
in the Blockchain Database. The credit manager is designed entities in the DNSTSM can be divided into the local
for credibility-based incentives to calculate the credibility recursive server and the forwarder recursive server. The
of each node and invoke the Dynamic Scheduling local recursive server is responsible for the functions of the
Mechanism to reward trusted nodes. traditional recursive server, and send the request to the
The threat model of our solution is constructed from the forwarder recursive server under the same model network
perspective of the attacker's possible attack methods. As when the poison alarm is triggered. The forwarder recursive
shown in Table 1, the security threats involved include server provides additional resolution services for the local
threats to member management and threats to DNS security. recursive server.
Under this threat model, our solution is effective. It should This section describes: DNSTSM member role
be specially stated that: assignment, DNSTSM member management.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain
TABLE 2. Entities correspond to roles in the DNSTSM.
Entity
Recursive Server (local)
Recursive Server (local/fowarder)
FIGURE 3. Member organization structure in DNSTSM network.
1) DNSTSM MEMBER ROLE ASSIGNMENT
As shown in Table 2, in the P2P network built by
DNSTSM, two types of recursive servers have five
different logical roles.
There are five logical roles in the DNSTSM: 1. Client:
The client uses the SDK to communicate with the
DNSTSM network. It obtains a valid identity certificate
from the CA and joins the application channel in the
network. 2. Peer (Endorser): Responsible for checking and
endorsing the proposal of the transaction, calculating the
execution result of the transaction. 3. Peer (Committer):
Responsible for rechecking the legality before accepting the
transaction result, accepting the modification of the legal
transaction reconciliation ledger, and writing to the
blockchain structure 4. Orderer: Is the blockchain identity
of the members of the sorting service in the cache sharing
consortium member. It does two things: first, receiving and
sorting the transactions by rules. Second, the completed
transactions are packaged into blocks at regular intervals. 5.
CA: A CA server cluster is formed by some nodes elected
in the DNSTSM network, responsible for managing the
certificates of all recursive servers in the DNSTSM
network, and providing standard PKI services, including
registration and unregister.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
Roles in DNSTSM Description
Client The direct operator of the blockchain
Peer (endorser) Simulated execution
Peer (committer) Verify the transaction
Orderer Sort and package transaction
CA Verify identity information
FIGURE 4. Flowchart of certificate issuance in DNSTSM.
2) DNSTSM MEMBER MANAGEMENT
Credibility is strongly related to identity authentication.
DNSTSM is established on the permissioned blockchain
and only allows authorized members to participate in data
maintenance. It uses the Membership Service Provider
(MSP) to implement member authentication. MSP abstracts
certificate issuance, user authentication, background
encryption mechanisms, and protocols. The MSP's
certificate is issued based on a trusted third-party
organization, the Certification Authority (CA), which
issues the certificate after confirming the identity of the
applicant. The Root Certificate is a self-signed certificate,
and a new certificate can be signed with the private key of
the Root CA Certificate to form a tree structure. The digital
certificate issuance process is shown in Fig.4. Here the
node signature uses the Elliptic Curve Digital Signature
Algorithm (ECDSA), and the digest is generated using
SHA-256. A certificate signed by the private key of the
Root CA Certificate can also issue new certificates. When
the new certificates serve as intermediate certificates, these
intermediate certificates form a list that meets the X.509
standard. Members in the list are called intermediate CAs,
and intermediate CAs can also Issue a new certificate with
own private key. Issuing a certificate is a Trust
endorsement process that forms a Chain of Trust from the
root CA certificate to the end-user certificate. The data
structure of membership information is shown in Table 3.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain
TABLE 3. Membership Data Structure.
Variables Types Description
Membership ID, one-to-one
id string
correspondence with ip
Adopted the public key
cert x.509
certificate standard
pk pk.Key public key
mspid string Subordinate to the MSP
The validity of the management certificate is managed
by using a CRL (Certificate Revocation List) or an OCSP
(Online Certificate Status Protocol) in the PKI system of
DNSTSM. DNSTSM member management includes
registration and unregister.
(a) Registration: Complete registration by applying to a
higher-level CA. There are three types of registered
certificates: ECert: issued to the user or node entity that
provides the registration credentials, long-term valid;
communication certificate (TLSCert): TLS certificate is used
to secure the communication link and control access to the
network layer Access, you can verify the remote entity
identity, prevent eavesdropping; transaction certificate
(TCert): issued to the user, control the permissions of each
transaction, generally for a transaction, short-term effective.
(b) Unregister: A certificate that has been issued can be
revoked due to rights management or other reasons. The
CRL is part of the shared cache and is used to track
revoked identities and certificates. It is shared by all CA
servers in the consortium, and non-CA servers do not have
access. There are several ways to unregister an issued
certificate: 1) Delete the intermediate CA certificate: delete
the certificate in the intermediate certs directory, so that the
certificates issued by the intermediate CA certificate belong
to the invalid certificate. 2) Add CRL list: you can add an
intermediate CA certificate or a single certificate to the
CRL list.
C. CREDIBILITY-BASED INCENTIVE MECHANISM
In this section, we propose a credibility-based incentive
mechanism to address the behavior of rational users “free-
riding” in cache-shared peer-to-peer networks. This
mechanism fine-grained regulate the traditional one-sided
credibility scheme and uses dynamic programming to
allocate node resources, while maximizing system
credibility maintenance efficiency, also enable resource-
providing nodes to maximize rewards after sharing reliable
network resources, in order to encourage rational users in
the network to perform security maintenance and upgrade
of their own domain name servers. Ensure the credibility of
the resources in the network, and achieve the characteristics
of sharing the trusted resources in the true sense of the
sharing model. First, describe and measure the definition of
credibility, and then introduce the implementation scheme
of the incentive mechanism based on node credibility.
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
1) DESCRIPTION AND MEASUREMENT OF
CREDIBILITY
Definition 1 Credibility: Credibility refers to the definition
of trust evaluation of target node based on its direct
interaction with peer node. Let the set of servers be
A, d  A , and the credibility of d is calculated as follows:
 q
1  e1 q , 0  q  1
d cred   (1)
 1 ,q 1
The calculation formula of q is as follows, where the
interaction between the target node and the peer node is
recorded as set B , and the abnormal result set of the server
is detected as C , t p is the time attribute of q .
t
 p.ln now
now  t p
pB t
q  1 (2)
 k
kB
In this paper, the transition parameter q is set as
logarithmic decay with time, that is, the credibility is
affected by previous attacks, and the influence intensity
gradually decreases with time. In practical applications, it
can be understood as a server with an abnormality,
although the vulnerability is fixed, it still has a greater risk
of being attacked than a server without an abnormality.
Definition 2 Similarity of abnormal state: refers to the
similarity degree of abnormal behaviors of nodes and other
nodes in the network. In this paper, the similarity is solved
by the method of cosine similarity. Suppose that the set of
detection results of server attack degree is C , p  C , and
the calculation formula of abnormal state similarity of d sim
is as follows:
n
p vi . pwi
(3)
d sim  i 1
n n
p
i 1
vi
2
p
i 1
wi
2
Definition 3 Roundtrip time ( RRT ): Round trip time
refers to the time it takes for the remote name server to
respond to requests. The internal timer is started whenever
the local server sends a query request to the remote name
server. Stops timing when it receives a response, recording
the time it takes for the remote name server to answer.
2) DYNAMIC SCHEDULING MECHANISM
The incentive mentioned in this section are reflected in the
dynamic scheduling mechanism. The dynamic scheduling
mechanism is composed of a scheduling manager, which is
used to allocate a suitable upstream resolution server for the
forwarding mechanism of the local resolution server. Based
on the credibility-based incentive mechanism, the dynamic
scheduling mechanism uses the credibility of nodes in the
cached alliance network as the allocation criterion. The
selection of the upstream resolution server considers the
node anomaly state similarity and the comprehensive index
of roundtrip time in the network. The higher the node
credibility, the greater the contribution to the total credibility
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain
of the network, correspondingly, the higher the allocation
priority, the mechanism encourages the node to share trusted
network resources for higher contribution.
Definition 4 Selection rate: The selection rate is a description
of the degree to which it can be selected to serve the target
node. The calculation formula is as follows:
a 1  dcred   b 1  d sim  +c 1  d rtt 
d selection _ rate 
  cred   sim   rrt 
hA
a 1  h  b 1  h  b 1  h
The initial value of d sim is 0, and a , b , and c are
constants. The selection coefficient is a comprehensive
consideration in terms of safety and performance. The higher
the similarity degree of the abnormal state, the smaller the
selection coefficient; the larger the RTT , the smaller the
selection coefficient. In the scheduling process, a basic
threshold is firstly set. If the number of times that the server
is selected as the upstream parsing server exceeds this
threshold, the server will be added to the frozen list, and the
server will not be selected in a dynamic allocation
mechanism cycle. After the cycle, the frozen list will be
removed and initialized. The specific selection method is
based on the priority of credibility, and three parsing servers
are selected as the upstream party according to the selection
coefficient. Since the credibility, the attack state similarity,
and roundtrip time are dynamically updated, the selection
factor changes, making the selection dynamic.
D. DNSTSM DATA ACQUISITION AND STORAGE
1) DNSTSM DATA ACQUISITION
This section describes the method of collecting trusted data
in DNSTSM and the process of returning trusted data to the
client.
FIGURE 5. Multi-DNS Recursive Servers Lookup Mechanism.
According to the fact that it is difficult for an attacker to
destroy multiple DNS servers using the same error data at
the same time, we propose a Multi-DNS Recursive Servers
Lookup Mechanism (MDRSLM) to isolate the fake reply
packets. It consists of Cache Validation Manager, TTL
Manger, and Alert Manager. Typically, when a client
program makes a domain name resolution request, the local
Recursive Server performs only one DNS resolution on
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
10.1109/ACCESS.2020.2966428, IEEE Access
behalf of the client. Instead of resolve only onetime,
DNSTSM duplicates the query and forwards them to
multiple recursive servers according to the Dynamic
Scheduling Mechanism. When the forwarder detects that
the request comes from other members of the consortium, it
does not trigger the MDRSL. In addition, a domain name
often corresponds to multiple IP (i.e., dynamic DNS). For
(4)
how to return the appropriate IP for the user, in MDRSLM,
the IP selection is related to the recursive server to which
the user belongs. DNSTSM executes the following actions
when a user's wants to resolve a domain name. The
proposed method does not change the packet length, so the
packet length is the same as the original DNS packet length.
These actions are shown in Fig. 5.
(a) The application's request triggers DNSTSM to perform
a regular domain name resolution to obtain IP data R ;
(b) If a packet with the same TXID is received later, the
Alert Manager is triggered and R is left blank, and write the
exception status record of this recursive server to the
blacklist;
(c) DNSTSM searches the Blockchain Database for
domain information and obtains the IP history value H . If
R  H and R is not empty, it can directly return the packet
where R is. Otherwise, proceed to the next step;
(d) DNSTSM converts the first and last letters of the name
tuple of the request packet question part to uppercase, as the
identifier forwarded by the consortium member, and
forwards it to the t recursive servers assigned by the
member manager. And gets the response Li  i 1,2 ,..., t  ;
(e) If R  Li , return the packet where R is, and
supplementary updates to this domain information are made
using R and RNS _ ID through the DNS Cache Manager;
(f) Based on R , the application will connect to the
trustworthy IP addresses.
Algorithm 1: SDDSA.
Input: Domain name X , Number of channels C
Output: Storage location N
Begin
H _ 32 ←Build a 32-bit hexadecimal number of X
base on CRC 32  X 
H _ 8 ←Convert a 32-bit hexadecimal number to an 8-
bit hexadecimal number base on two fold-and-fold
additions method with overflow
if C  255 then
D ←Convert hexadecimal H _ 8 to decimal
N ←Generates the storage location of X based on the
operation D mod C
else
Exit Algorithm // organizations exceeded
endif
End
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain
FIGURE 6. Cache confirmation flow in DNSTSM.
2) DNSTSM DATA STORAGE
This section accomplishes two tasks. The first is to assign the
channel address stored on the DNSTSM to the verified A
record, and the second is to store the verified A record on the
matching blockchain.
A channel is a private "subnet" for communication
between two or more specific network members for
transactions that require data confidentiality. In Fabric,
establishing a channel is equivalent to establishing a sub-
chain. However, because all peers in the same channel store
and process the same data, this storage structure can affect
the read and write speed of Peer nodes when the amount of
data is very large.
The multi-channel (i.e., multi-chain) storage scheme can
improve the reading efficiency of system data. Based on the
multi-chain deployment structure, smart contracts running on
different chains access different data, improving the
efficiency of parallel processing. Each organization has
members used as sorting service nodes to increase the
throughput of the blockchain system by equally assigning
sorting tasks.
We designed a Stochastic Distributed Decentralized
Storage Algorithm (SDDSA) to select storage locations.
The usage scenarios of the algorithm include the selection
of the storage location when storing data and the
acquisition of the read address when reading data. When
storing or reading data, the position information is obtained
by using two parameters of the converted domain name and
the number of channels (i.e., the initial number of
members). The feature of the algorithm is to ensure the
uniform distribution of the storage location and the
certainty of the storage location.
After obtaining the storage location, the DNSTSM system
initiates the Blockchain-based DNS Caching Mechanism. As
Fig.6 shown, the caching process is as follows: (1) Send TX
Proposal (including write, delete, read); (1.a) Verify Proposal
signature; (1.b) Check if the Channel ACL is satisfied; (1.c)
Simulate execution of transactions and sign results (ESCC);
(2) Reply to Proposal Response; (2.a) Verify signature; (2.b)
Compare the results of multiple Endorsers; (2.c) Check if
enough Endorsement is collected; (3) Send Transaction; (3.a)
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
Submit Transaction; (3.b) Forward; (4) Sort the transactions
and construct the blocks; (5) Send Transaction block; (5.a)
Check transaction structure integrity, signature, repeat; (5.b)
Verify that the transaction complies with the Endorsement
Policy (VSCC); (5.c) Check that the version in the read set is
consistent with the ledger; (5.d) Perform legal transactions in
the block and update the account status.
IV. EXPERIMENTAL EVALUATION
Unless explicitly stated, in our experiments: (1) The node
runs on Fabric v1.1 and performs performance testing
through local login, (2) All nodes use a 2.0GHz 16-vCPU
virtual machine, using Ubuntu16.4 LTS system, with 12G
memory, (3) Use three ZooKeeper nodes, four Kafka brokers
and three Fabric orderers to run multi-channel orderer
service on different virtual machines, (4) There are five peers,
which are composed of different recursive servers and all of
which are CA endorsement nodes, (5) Signature uses the
default 256-bit ECDSA policy.
A. PERFORMANCE TEST
In this section, we evaluate the performance of the proposal
in terms of availability and efficiency. In addition, we
estimate the effects of factors such as cache percent and the
number of members on system efficiency.
1) AVAILABILITY
This test is used to verify the normal domain name
resolution capability of the system under the cache
poisoning attack. Birthday attacks were launched on
traditional DNS servers and DNSTSM systems. The values
of the parameters are shown in Table 4. RT and NoF are
used to set the number of valid packets forged by the
attacker, and q is the number of requests for the same
domain name sent by the attacker in advance. In this case,
the attacker only needs to guess one of them. RT takes 0.1s
to refer to "DNS Security", NoF takes 10000 and q takes
100 is the empirical value. In this configuration, the attack
effect can be made more obvious. The value of n is 2,
which is the simplest case, and the lower limit of the
security scheme provided by the corresponding scheme.
According to the experimental results in Fig.7, the attack
success rate of the DNSTSM system is on the order of 10 11 ,
while that of the HARD-DNS and DependDNS system is
on the order of 10-9 . With the same number of server nodes,
DNSTSM system has better anti-attack capability and
stability than the HARD-DNS and DependDNS. The
results show that the model and method are effective. This
indicates that it is difficult for an attacker to cooperate with
multiple DNS servers, and a single trial-and-error attack
cannot produce a cumulative effect. When the alarm senses
an abnormal response, the DNSTSM system can block the
attack chain by dynamically changing the forwarder to
achieve the high availability of the system.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain

TABLE 4. Experimental parameter configuration. As shown in Fig.8, the average latency of DependDNS and
HARD-DNS queries is 128.93ms and 119.81ms, and the
Variables Value Description average latency of the DNSTSM domain name system is
Number of query requests for the 94.562ms. The DNSTSM mechanism gives a lower latency
q 100 same domain name sent by the
than the control group.
attacker in advance 100
Number of false attack packets that
NoF 10000
an attacker can send per second
RT 0.1s DNS query request reply time 90
Forwarding party recursive server

Response time (ms)

n 2
number
80

70

10

0
0 20 40 60 80 100
Cache percent (%)
FIGURE 9. The relationship between the proportion of the query in the
cache of the DNSTSM system and the average response time.
100

98 With SDDSA
Normal
96
Response time (ms)

94
FIGURE 7. Effect of different attack intensity on availability of existing
advanced methods and DNSTSM system. 92

90

88
8
6
4
2
0
0 3 6 9 12 15 18 21 24 27 30 33
Num
FIGURE 10. Comparison of average response time between the system
with SDDSA and without SDDSA.
Fig.9 shows that the DNSTSM system can bring
cumulative benefits. As the cached data increases, the cache
hit probability is increased to ensure the efficiency of the
service. We use the system with and without the SDDSA
FIGURE 8. Comparison of average response time between DNSTSM
algorithm as a control group to test the average response time
system and existing advanced methods. when all nodes perform domain name resolution at different
2) EFFICIENCY system sizes (number of nodes). As shown in Fig. 10, the
Compared with general domain name queries, DNSTSM system with the SDDSA algorithm has a lower average delay
requires multiple domain name servers and blockchain than the system without the SDDSA algorithm, and as the
databases at the same time, which adds additional latency. system size (number of nodes) increases, the delay of the
The delay mainly includes the user and data validator delay system with the SDDSA algorithm increases very low. These
Tcv , the data validator and the last response. The propagation characteristics show that the scheme we designed improves
delay between the forwarders is delayed by Tvf and the the system's ability to handle multitasking. Combining the
processing delay between the blockchain database Tb . The above two experiments, we can see that increasing the
data validator delay Tv , the forwarder's resolution delay T f . number of members in the consortium only brings a very
little delay, but the number of members is more conducive to

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain

increasing the cache rate, which means that the overall
efficiency of the system has been improved.
Therefore, the DNSTSM prototype system only needs to
add a small processing delay and network communication
delay in the case of providing highly reliable and highly
available services. At the same time, when sending requests
to DNSTSM systems at different stages (referring to
different levels of caching), the network communication
delay can have an important impact on the average time of
the actual domain name query. Different sizes (referring to
the number of members of the cache consortium) have less
impact on response time, but scale effects can lead to more
cache rates. Therefore, the DNSTSM system should be
formed to a certain size in exchange for low latency.
B. SECURITY ANALYSIS
In our scheme, by combining the identity authentication
mechanism and consensus mechanism of the consortium
blockchain, and designing the incentive mechanism
according to the technical characteristics of DNS resolution
service, we obtain a resolution service that is more secure
and trusted than the traditional P2P network sharing model.
In this section, we discuss the security of the program.
As shown in Table 5, we compare our scheme with
existing advanced schemes from the perspective of Defend
against cache poisoning attacks and internal attacks. The key
factors that define internal attacks include: Avoid
information disclosure, Identity management, Access control,
Avoid conspiracy. We take whether to resist Kaminsky
attack and whether to address these four critical factors of
internal attacks as the evaluation criteria in Table 5. The
comparison results show that our solution has more
advantages in terms of functions and features.
1) COMPLETE DATA TRUST CHAIN
Reliable data source: In our solution, we obtain reliable
cached data source in the way of multi-recursive server
parsing, and ensure the credibility of the data from the source;
Security of data storage: The entire system obtains a copy
of the complete database in the form of a sub-database. Each
copy is a chain of identical blocks, which stores the data in a
Table 5. Comparison of the security indicators of our solutions with dox, CoDNS,ConfiDNS, DepenDNS, and HARD-DNS
chained block structure with a timestamp, and the subsequent
blocks record the hash of the previous block (prehash).
Changes to any block data are changed from the changed
block to the last block. Moreover, any new legal block must
be generated through a consensus mechanism to agree on the
final selected block. Ensure that data cannot be tampered
from a storage perspective;
Security of data transmission: Every message transmitted
through Gossip is signed, so forged messages sent by
Byzantine nodes are easily identifiable.
2) COMPLETE MEMBERSHIP TRUST CHAIN
The certificate represents the abstraction of a member entity
in the consortium. In order to solve the problem of member
certificate authority management, there is a Member Service
System (MSP) in the DNSTSM. MSP is based on the
identity authentication framework PKI to achieve
certification and management related to the certificate life
cycle. The issuance of a certificate requires the CA to sign
the validity of the certificate. A higher-level CA can issue the
legitimacy of the CA, and the root CA realizes the basis of
trust by pre-distributing certificates.
3) MOTIVATION FOR CREDIBILITY
The credibility-based incentive mechanism serves as an
important part of ensuring the credibility level of DNSTSM,
stimulating the enthusiasm of members to maintain their own
server security performance, and maintaining the virtuous
circle of DNSTSM domain name resolution.
4) AVOID NODE FAILURE
Avoid single point of failure: In our solution, the consensus
mechanism uses the Kafka message queuing system, and the
sorting nodes are clustered by Kafka, which can effectively
avoid the single network failure and cause the entire network
to collapse;
Node crash tolerance: Each node continuously extracts
blocks from other nodes in the channel. Since there is no
need for a fixed connection to maintain Gossip-based data
propagation, the process can reliably implement node crash
tolerance for shared ledgers.

Defend against cache Avoid information Identity Access Avoid

poisoning attacks disclosure management control conspiracy
Dox High No No No No
CoDNS Medium No No No No
ConfiDNS Medium No No No No
DepenDNS Medium No No No No
HARD-DNS High Yes No No No
Ours High Yes Yes Yes Yes

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain

V. CONCLUSION [13] Q. Feng, D. He, S. Zeadally, M. K. Khan, N. J. J. o. N. Kumar, and

In this paper, we proposed a novel consortium blockchain- C. Applications, "A survey on privacy protection in blockchain
system," vol. 126, pp. 45-58, 2018.
based DNS Cache Resources Trusted Sharing Model [14] D. Puthal, N. Malik, S. P. Mohanty, E. Kougianos, and G. Das,
(DNSTSM). By introducing the consortium blockchain into "Everything You Wanted to Know About the Blockchain: Its
the DNS system, a secure and trusted DNS cache is Promise, Components, Processes, and Problems," IEEE Consumer
Electronics Magazine, vol. 7, no. 4, pp. 6-14, 2018.
maintained and the risk of illegal access and collusion is [15] FISCO BCOS [Online]. Available: www.fisco-bcos.org/
reduced. DNSTSM includes a trust-based incentive [16] Quorum [Online]. Available: https://quorum.com/
mechanism that creates positive feedback of trust for model [17] Hyperledger Fabric [Online]. Available:
https://www.hyperledger.org/
members. We designed a Stochastic Distributed [18] X. Zhang and X. Chen, "Data Security Sharing and Storage Based on
Decentralized Storage Algorithm to improve system a Consortium Blockchain in a Vehicular Ad-hoc Network," IEEE
throughput and load balancing. The experimental results Access, vol. 7, pp. 58241-58254, 2019.
show that DNSTSM meets the good performance [19] D. Li, R. Du, Y. Fu, and M. H. Au, "Meta-Key: A Secure Data-
Sharing Protocol Under Blockchain-Based Decentralized Storage
requirements under the premise of ensuring the security of Architecture," IEEE Networking Letters, vol. 1, no. 1, pp. 30-33,
domain name resolution. All of these benefits are obtained 2019.
without changing any external server-side DNS [20] M. Yang, A. Margheri, R. Hu, and V. Sassone, "Differentially
Private Data Sharing in a Cloud Federation with Blockchain," IEEE
infrastructure, which allows DNSTSM to be incrementally Cloud Computing, vol. 5, no. 6, pp. 69-79, 2018.
deployed, requiring only the local server to join the [21] J. Liu, B. Li, L. Chen, M. Hou, F. Xiang, and P. Wang, "A Data
consortium network and a minimal proxy. Storage Method Based on Blockchain for Decentralization DNS," in
2018 IEEE Third International Conference on Data Science in
In the future, we will explore and design a decentralized Cyberspace (DSC), 2018, pp. 189-196.
root CA certificate mechanism for DNSTSM to improve the [22] X. Wang, K. Li, H. Li, Y. Li, and Z. Liang, "ConsortiumDNS: A
decentralization and fault tolerance of the identity Distributed Domain Name Service Based on Consortium Chain," in
2017 IEEE 19th International Conference on High Performance
authentication system. Computing and Communications; IEEE 15th International
Conference on Smart City; IEEE 3rd International Conference on
REFERENCES Data Science and Systems (HPCC/SmartCity/DSS), 2017, pp. 617-
[1] J. Trostle, B. V. Besien, and A. Pujari, "Protecting against DNS 620.
cache poisoning attacks," in 2010 6th IEEE Workshop on Secure [23] Namecoin [Online]. Available: https://namecoin.org/
Network Protocols, 2010, pp. 25-30. [24] M. Ali, J. Nelson, R. Shea, and M. J. Freedman, "Blockstack: A
[2] L. Fan, Y. Wang, X. Cheng, and J. Li, "Prevent DNS Cache global naming and storage system secured by blockchains," in 2016
Poisoning Using Security Proxy," in 2011 12th International Annual Technical Conference, 2016, pp. 181-194.
Conference on Parallel and Distributed Computing, Applications
and Technologies, 2011, pp. 387-393.
[3] L. Yuan, C.-C. Chen, P. Mohapatra, C.-N. Chuah, and K. Kant, "A
Proxy View of Quality of Domain Name Service, Poisoning Attacks
and Survival Strategies," ACM Trans. Internet Technol, vol. 12, no.
3, pp. 1-26, 2013.
[4] R. Chandramouli and S. Rose, "Challenges in securing the domain
name system," IEEE Security & Privacy, vol. 4, no. 1, pp. 84-87, ZHONG YU received the M.S. and Ph.D.
2006. degrees in electronic science and technology
[5] I. M. M. Dissanayake, "DNS Cache Poisoning: A Review on its from the Xi’an Jiaotong University, China, in
Technique and Countermeasures," in 2018 National Information 2000, and 2005, respectively. He is currently an
Technology Conference (NITC), 2018, pp. 1-6. Associate Professor with the Xi’an University of
[6] A. Herzberg and H. Shulman, "Antidotes for DNS Poisoning by Off- Posts and Telecommunications. His current
Path Adversaries," in 2012 Seventh International Conference on research interests include communication
Availability, Reliability and Security, 2012, pp. 262-267. technology and blockchain technology.
[7] L. Yuan, K. Kant, P. Mohapatra, and C. Chuah, "DoX: A Peer-to-
Peer Antidote for DNS Cache Poisoning Attacks," in 2006 IEEE
International Conference on Communications, 2006, vol. 5, pp.
2345-2350.
[8] C. Gutierrez, R. Krishnan, R. Sundaram, and F. Zhou, "HARD-DNS:
Highly-Available Redundantly-Distributed DNS," in 2010 -
MILCOM 2010 MILITARY COMMUNICATIONS CONFERENCE,
2010, pp. 1343-1348.
[9] K. Park, V. S. Pai, L. L. Peterson, and Z. Wang, "CoDNS:
Improving DNS Performance and Reliability via Cooperative DONG XUE received the B.S. degree in
Lookups," in OSDI, 2004, vol. 4, pp. 14-14. electronic information engineering from Luliang
[10] L. Poole and V. S. Pai, "ConfiDNS: Leveraging Scale and History to University, China, in 2016. He is currently
Improve DNS Security," in WORLDS, 2006, vol. 3, pp. 3-3. pursuing the M.S. degree in communication and
[11] H.-M. Sun, W.-H. Chang, S.-Y. Chang, and Y.-H. Lin, "DepenDNS: information system with Xi’an University of
Dependable mechanism against DNS cache poisoning," in Posts and Telecommunications, Xi’an, China.
International Conference on Cryptology and Network Security, 2009, His research interests include network security
pp. 174-188: Springer. and the blockchain technology.
[12] N. J. Alfardan and K. G. Paterson, "An Analysis of DepenDNS," in
International Conference on Information Security, 2010, vol. 6531,
pp. 31-38.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2020.2966428, IEEE Access
Z. Yu et al.: DNS Cache Resources Trusted Sharing Model (DNSTSM) Based on Consortium Blockchain

JIULUN FAN received the B.S. and M.S.

degrees in basic mathematics from the Shanxi
Normal University, China, in 1985, and 1988. He
received the Ph.D. degrees in signal and
information processing from the Xidian
University, China, in 1998. He is currently a
Professor with the Xi’an University of Posts and
Telecommunications. His current research
interests include information security and
network security.

CHANG GUO received the B.S. degree in

electronic information engineering from
Zhongyuan University of Technology, China, in
2017. He is currently pursuing the M.S. degree in
electronics and communication engineering with
Xi’an University of Posts and
Telecommunications, Xi’an, China. His research
interests include network security and the
blockchain technology.

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.