Fast user authentication in 5G heterogeneous networks using RLAC-FNN and blockchain technology for handoff delay reduction

Abstract

In the fifth generation (5G), ultra-dense heterogeneous network (UDHN) is considered as a prominent technology to resolve network system problems. It is challenged to provide secure access since the UDHN contains access points (APs), user equipment (UE) which are characterized with the nature of dynamic, temporary as well as autonomy. The coverage of AP is less when compared to that of classical base station and the problem arises with the interaction between APs and UE during the mobility of UE. In order to attain efficient key agreement with fast and subsequent authentication, a new consensus mechanism has been proposed. By integrating the reinforcement learning method with the actor-critic learning based on fuzzy neural network (RLAC-FNN), blockchain-enabled handover authentication is enabled in 5G heterogeneous Networks. In the proposed method, the user can establish a secure and quick connection by excluding re-authentication and handover operators between heterogeneous cells with less delay. In the proposed approach, all the secondary peers have been further divided into several secondary peer groups based on credit value. The credit of all the secondary peers will be assigned based on their probability of successful participation in consensus. The secondary nodes will summarize all their results to return to the local service centre (LSC). Finally, LSC will identify the trusted and malicious peers by confirming the consensus adaptively by integrating the RLAC-FNN. The performance of the proposed method has analyzed by implementing it in Python platform and compared with existing approaches. The simulation outcomes showed that the proposed method could efficiently reduce the authentication frequency, handover delay, consensus delay, etc., than existing approaches.

Full text

ORIGINAL PAPER
Fast user authentication in 5G heterogeneous networks using
RLAC-FNN and blockchain technology for handoff delay reduction
Shivanand V. Manjaragi
1,2
•S. V. Saboji
2
Accepted: 27 April 2023
The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023
Abstract
In the fifth generation (5G), ultra-dense heterogeneous network (UDHN) is considered as a prominent technology to
resolve network system problems. It is challenged to provide secure access since the UDHN contains access points (APs),
user equipment (UE) which are characterized with the nature of dynamic, temporary as well as autonomy. The coverage of
AP is less when compared to that of classical base station and the problem arises with the interaction between APs and UE
during the mobility of UE. In order to attain efficient key agreement with fast and subsequent authentication, a new
consensus mechanism has been proposed. By integrating the reinforcement learning method with the actor-critic learning
based on fuzzy neural network (RLAC-FNN), blockchain-enabled handover authentication is enabled in 5G heterogeneous
Networks. In the proposed method, the user can establish a secure and quick connection by excluding re-authentication and
handover operators between heterogeneous cells with less delay. In the proposed approach, all the secondary peers have
been further divided into several secondary peer groups based on credit value. The credit of all the secondary peers will be
assigned based on their probability of successful participation in consensus. The secondary nodes will summarize all their
results to return to the local service centre (LSC). Finally, LSC will identify the trusted and malicious peers by confirming
the consensus adaptively by integrating the RLAC-FNN. The performance of the proposed method has analyzed by
implementing it in Python platform and compared with existing approaches. The simulation outcomes showed that the
proposed method could efficiently reduce the authentication frequency, handover delay, consensus delay, etc., than existing
approaches.
Keywords Local service centre Ultra-dense heterogeneous networks Blockchain-enabled authentication handover 
APs group And actor-critic learning
1 Introduction
The fast development of mobile devices as well as the
applications and their processing requirements tend to the
advent 5G networks. Compared to 4G networks, the 5G
networks are enabled with properties such as high bit rate
than 10 Gb/s, improved network coverage and minimum
latency. The 5G networks can extend overlay coverage and
operate via heterogeneous cells [1]. The 5G users namely
internet of things (IoT) devices, mobile nodes, vehicles
create the handover procedure activated when moving from
one cell to another. It can tend to cause delay in 5G net-
work if the handover process is frequently happened. In the
heterogeneous network, the handover management is
considered as one of the primary issue. Besides, the han-
dover fulfills ultra-reliable communications requests and
very high availability as well as reliability in 5G networks
[2].
The handover management compacts with each active
link of an user device, which transfers the linkage between
user device as well as the counterpart from a single net-
work point to other network point. Therefore, the handover
decision can define the finest acces network as well as
&Shivanand V. Manjaragi
shiva.vm@gmail.com
S. V. Saboji
saboji_skumar@yahoo.com
1
Department of Computer Science and Engineering, Hirasugar
Institute of Technology, Nidasoshi, Belagavi, India
2
Department of Computer Science and Engineering,
Basaveshwar Engineering College, Bagalkot, India
123
Wireless Networks
https://doi.org/10.1007/s11276-023-03371-z(0123456789().,-volV)(0123456789().,-volV)

decide whether the process of handover has carried out.
The authentication process becomes more intricated and
can improve delay time, by denying the 5G intentions.
Among 5G heterogeneous cells, handover process can tend
to provide low performance through ineffective authenti-
cation process [3]. In 5G networks, cell resource and power
restrictions among the APs need minimum complexity and
high effective handover authentication events among
homogeneous and heterogeneous cells [4]. The 5G network
provides benefit in communication, while considering the
technical aspects, privacy protection, authentication beha-
viour and the existence of heterogeneous cells [5].
5G requires a standard level of security in the network
structure and application scenarios, particularly in authen-
ticating the services and offering their access level for 4G.
An effective approach is needed to develop 5G networks to
confirm and protect privacy in faster, safer and efficient
manner [6]. The security requirement is higher in 5G than
other networks in which solutions are offered with intellec-
tual control across heterogeneous cells for dependable con-
trivances. In current decades, the new technology of
blockchain received great attention to develop the next
generation of wireless networks [7]. The consensus algo-
rithm is a core portion of the blockchain which is considered
as a significant element. It has termed as fundamental to
confirm the effective collaboration of blockchain network.
The main problem of the blockchain is how each node is
created to preserve their data consistent via interaction rules.
A consensus algorithm is developed to solve this problem to
accomplish correctness and consistency of ledger data on
various nodes [8,9]. It needs learning from prior approaches
for accomplishing state consensus in a distributed structure.
Developing the approach for choosing accounting nodes in a
network, as well as how to guarantee that the ledger data
creates a proper consensus in the whole network [10].
Different kinds of consensus algorithms like Proof of
Elapsed Time (PoET), Casper, Proof of Stack (PoS), dele-
gated Proof of Stale (dPoS), Practical Byzantine Fault Tol-
erance (PBFT), and Proof of Work (PoW) are introduced
based on the development of several applications. The
commonly used algorithms are Consortium blockchain as
well as Public blockchain. These approaches have various
benefits, drawbacks, and application in UDHN network
scenarios [11,12]. The POW approach has consumed more
energy and is easily affected by the force of higher consensus
cycle [13]. Howeever, for a communication system along
with time delay, POW is not valuable. Enhancing the density
of APs in unit zone and creating a Ultra Dense Network
(UDN) is a significant means to solve the challenge of
improving the traffic of network by 1000 times and
enhancing the speed of user experience for 10–100 times.
UDN is the most efficient network to handle the fast
development of higher traffic in 5G network, particularly in
the hotspot region. It has projected that the distribution
density of the small APs will reach more based on the
different Radio Access Technology (RAT). The huge APs
can serve for high density UE, and each AP creates a Peer-
to-Peer (P2P) uncentered network [14,15]. The AP has
small coverage and low power in UDN. The user can often
change between APs and decrease access stability and
speed for highly moving mobile users. The APs are no
longer in 5G, but it consumes more business collaboration
with UE. It can offer data services as well as control
according to the variations of practical needs. AP is the
crucial portion of the UE accessing the mobile internet.
The enumerated data verifying users by AP can face a
security problem that can disturb the privacy of user
transaction data. Hence, providing secure access as well as
effective authentication in UDN networks is termed as a
new challenge for the upcoming 5G network structure and
security framework.
1.1 Motivation
In 5G network, the ultra-dense small cell networks have
employed to offer data rates, security as well as reliability
concerns are introduced by low latency in the network. Due
to the availability of various malicious attacks, security
problems may occur and also the frequent handover tends
to become a reliability issue. Hence, offering a secure and
reliable connection is significant however, at the same time
it is challenging for 5G networks. The blockchain is con-
sidered as a most encouraging technology that emcompass
the prospective to transform the way in which services are
provided to 5G communication network. Besides, to permit
end-to-end services delivery over the whole 5G network,
blockchain has the ability tocobine the authentication
process with 5G network. Nowadays, the main challenge
for the recent 5G platform is the necessity to assure an
transparent, open as well as a fair system in the unexpected
number of resources and various malicious users. With its
distinctive decentralization feature, high level of security,
data privacy, immutability and transparency, blockchain
has turn out to be an evident choice. Hence, blockchain is
needed to combine into a 5G network. However, the main
problem of the blockchain is how each node is created to
keep its data consistent via a rule. To solve this problem, a
consensus mechanism has been presented for enabling
correctness and consistency of the ledger data on various
nodes.
1.2 Contribution
•The new authentication approach is developed with the
utilization of blockchain-enabled consensus mechanism
and RLAC-FNN to avoid re-authentication during
Wireless Networks
123

repeated handover of heterogeneous cells. It resolves
the technical challenges related to user privacy protec-
tion, authentication and resource management in 5G
UDHN.
•The proposed approach was modelled to assure low
delay in which the users are replaced with least delay
among heterogeneous cell. It can be accomplished with
the use of private and public keys provided by block
chain to ensure privacy and to enhance user authenti-
cation with minimal handoff delay in 5G networks.
Quick and secure connection during handoff operations
among heterogeneouscell is obtained with low delay.
•Using the block chain propagation model, the authen-
tication outcomes of UE have transmitted in the Access
Point Group (APG) through directional trust transfer.
Thereby, the members of APG will share the authen-
tication outcomes of UE, minimize the frequency of
authentication while the UE moves between APs, and
enhance the user experience and access efficiency.
•The APs has been organized into a secured, trusted
chain as APG, so the reliability and security of APG
have been enhanced. By using proposed RLAC-FNN
approach, better energy consumption with less sigan-
lling overhead than existing approaches are achieved.
The paper’s organization is illustrated as follows: In
Sect. 2, the related works done by various researches are
discussed. In Sect. 3, the proposed methodology is dis-
cussed in detail. In Sect. 4, the experiments accomplished
to determine the performance of the proposed method is
illustrated. Finally, the conclusion and future work is
specified in Sect. 5. The list of abbreviations is given in
Table 1.
2 Related works
The work related to the proposed handover authentication
is described as follows.
Sangeetha et al. [16] suggested a trust-based handover
authentication scheme in Software-Defined Networking
(SDN) 5G heterogeneous network to improve 5G mobile
communications security. The three-way handshaking
protocol was used to attain mutual authentication. Further,
the security of the trust-based handover authentication
scheme has been determined utilizing a trust value algo-
rithm along with a clustering process.
Divakaran et al. [17] suggested an enhanced handover
authentication model in 5G communication networks by
employing fuzzy evolutionary optimization. In this
authentication model, fuzzy evolutionary optimization was
utilized to manage handover as well as maintain key
management to enhance the effectiveness in 5G networks.
Moreover, this method was modelled to reduce complexity
and delay during network authentication in 5G networks.
Man Chun and Maode Ma [18] suggested a 5G
authentication approach and key agreement protocol based
on blockchain against Denial-of-Service (DoS) attacks.
The suggested method had employed the private block-
chain to offer a distributed database for storing each
authentication record. Moreover, it also explored the trap-
door collision property of the Chameleon hash function,
where the blockchain entries could check incoming
authentication requests. The device anonymity was pro-
tected using a SUbscription Concealed Identifier (SUCI),
and Elliptic-Curve Diffie-Hellman (ECDH) had utilized to
establish a session key.
Xinghua Li et al. [19] suggested a Fast and Universal
Inter-Slice (FUIS) handover authentication model using
ring signature, chameleon hash and blockchain. For inter-
slice handover, a service-oriented authentication protocol
was introduced and a key agreement by generating an
anonymous ticket using a property of trapdoor collision. A
privacy-preserving ticket validation along with a ring sig-
nature was modelled to minimize computation overhead
during the authentication process, mainly for completing
the consensus phase of the blockchain.
Xudong Jia et al. [20] suggested a blockchain-based
decentralized authentication model called A
2
chain for 5G-
assisted IoT. The processing of authentication requests was
initially decentralized by using edge computing in the A
2
chain and avoiding the burden on the network and
authentication services. Next, blockchain and sidechain
technologies were utilized in the A
2
chain to execute the
cross-domain identity of IoT devices and securely share the
identity verification data. Further, to remove the manage-
ment overhead instigated by the centralized authentication
scheem, A
2
restores Public Key Infrastructure (PKI) with
the Identity-Based Cryptography (IBC) algorithm.
Zaher Haddad et al. [21] suggested a blockchain-based
new efficient, secured authentication and key agreement
protocol in 5G network. The network’s capital entity,
known as the Home Network (HN), was accountable for
initiating the blockchain as well as bootstrapping the net-
work. The HN network would share the Blockchain over
the entire network nodes. The authentication as well as
registration process should be performed between the
serving network and the UE if the new UE is switched on
or comes to a new coverage area. Moreover, this protocol
had secured and endured major attacks like DoS, a man in
the middle, distributed DoS, compromising and Hijacking
attacks.
In SDN-based 5G networks, a blockchain-based
authentication handover and privacy protection were pre-
sented by Abbas Yazdinejad et al. [22]. SDN-based and
Blockchain-based authentication approaches were
Wireless Networks
123

Table 1 List of abbreviations Abbreviation Explanation
AC Actor-critic
APG Access point group
AUC AUthentication center
APs Access points
B Beyond
BAHEPP Blockchain-enabled authentication handover with efficient privacy protection
BS Base station
CNs Candidate nodes
CRBFT Credit reinforcement byzantine fault tolerance
C/S Client-to-server
DoS Denial-of-service
dPoS delegated proof of stale
ECDH Elliptic-curve diffie-hellman
FNN Fuzzy neural network
FUIS Fast and universal inter-slice
HN Home network
IBC Identity-based cryptography
ID IDentifier
IoT Internet of things
IT2-FS Interval type 2 fuzzy set
LSC Local service centre
ME Management engine
MIMO Multi-input multi-output
MN Master node
NN Neural network
PBFT Practical byzantine fault tolerance
PDR Packet delivery ratio
PKI Public key infrastructure
PoET Proof of elapsed time
PoS Proof of stack
PoW Proof of work
P2P Peer-to-peer
RAT Radio access technology
RL Reinforcement learning
RLAC Reinforcement learning with the actor-critic
RTT Round trip time
SCI Secure context information
SDH Software-defined handover
SDN Software-defined networking
SHA Secure hash algorithm
SN Sub-node
SUCI SUbscription concealed identifier
UAV Unmanned aerial vehicle
UDHN Ultra-dense heterogeneous network
UDN Ultra dense network
UE User equipment
5G Fifth generation
6G Sixth generation
Wireless Networks
123

developed to remove re-authentication in frequent han-
dovers amid the heterogeneous cells. The system was
accomplished to guarantee the minimum delay, suitable for
5G network. Here, users could be interchanged with the
minimum delay between heterogeneous cells by private
and public keys offered through invented blockchain ele-
ment while keeping their privacy.
Amina Gharsallah et al. [23] suggested SDN based
handover management process in5G UDN. This process
could utilize Software-Defined Handover (SDH) for opti-
mizing the handover in 5G networks. Besides, a
SDHManagement Engine (SDH-ME) to manage the han-
dover control process in 5G UDN. SDH-ME was utilized to
apply the SDN structure that had been accomplished
through the control plane to orchestrate the data plane.
Jing Yang et al. [24] suggested a fast and unified han-
dover authentication model in 5G SDN heterogeneous
networks based on a link signature. To establish fast and
unified handover authentication, a distinctive wireless
channel characteristic between serving AP and user were
extracted as a Secure Context Information (SCI) and for-
warded to the target AP. Further, the latter could evaluate
whether the user was a legitimate one, who had previously
authenticated corresponding to the forwarded SCI.
A blockchain-based security authentication approach
had been presented by Zhonglin Chen et al. [25] for 5G
UDN. Here, an APG was employed with PBFT approach
based on blockchain consensus mechanism. In APG-PBFT,
the consensus mechanism would be optimized, as well as a
new reverse screening approach would be adapted. More-
over, along with the APs, a trusted chain APG could be
created using the APG-PBFT algorithm. By using block-
chain message propagation, the results of authentication
could be shared in the APG.
Muhammad Nabeel et al. [26] had presented the
deployment of TurboRAN testbed for evaluating the per-
formance of 5G& Beyond (B) cellular network. The
deployment challenges were discussed and highlighted
with the details of hardware components as well as the
reason for selecting the hardware components were pro-
vided. Case study for the working procedure of TurboRAN
was provided with the impact of mobility parameters. It is
essential to highlight the TurboRAN testbed focusing on
sub-6 GHz and Multi-Input Multi-Output (MIMO) capa-
bility features.
Shidrokh Goudarzi et al. [27] had proposed a approach
based on cooperative game theory for selecting Unmanned
Aerial Vehicle (UAV) during handover and optimized with
the decrease of handover latency, end-to-end delay and
signal overhead. In addition to that the standard modelling
of software-defined network software-defined network with
media-independent handover which are utilized as a for-
warding switches for obtaining seamless mobility. Table 2
shows the Contribution and Limitations of existing
methods.
The consensus mechanism is considered a major process
in blockchain-based 5G heterogeneous networks. Recently,
several consensus protocols have been proposed by dif-
ferent authors and applied in several decentralized plat-
forms. Still, it is not suitable for 5G environments because
of handover delay, re-authentication, etc. Therefore, a new
blockchain-based handover and consensus mechanism is
necessary in 5G heterogeneous networks to resolve these
problems and provide fast authentication.
3 Proposed methodology
In our proposed method, a new consensus mechanism of
blockchain-enabled authentication handover approach in
5G Heterogeneous Networks is proposed to enhance the
user authentication. In the proposed method, users acquire
a secure and quick connection through avoiding re-au-
thentication amid handovers operators among heteroge-
neous cells with minimum delay. Delay reduction is
considered as one of the primary intentions and features of
5G, is of great significance that can occur with solid
structure. The proposed consensus mechanism will opti-
mize the consensus node partition structure and adapt the
consensus period. The LSC will use a new learning
approach to detect the trusted AP in the block generation
process. Our proposed algorithm improves the node sepa-
ration structure to reduce the delay of the consensus
algorithm. In the proposed approach, all secondary peers
are further subdivided into several secondary peer groups
based on credit value. All secondary peers are credited
based on their likelihood of successfully participating in
the consensus. Each secondary peer group contains some
Candidate Nodes (CNs). Here, the CNs will return their
results to secondary nodes. The secondary nodes will
summarize all their results to return to LSC. Finally, LSC
will identify the trusted peer and the malicious peers by
adaptively confirming the consensus by integrating the RL
method with the AC learning based on FNN (RLAC-FNN).
In addition, the block chaining methods usually set up a
certain time interval to remove the block generated during
the consensus process. After this interval, the consensus
result must be recomputed to form an APG trusted block-
chain. This period will be adjusted automatically by sens-
ing the number of nodes and their transactions in the
proposed method.
3.1 System model
In 5G, UDHNs have been determined as a key mechanism
for managing orders of magnitude rise in the volume of
Wireless Networks
123

Table 2 Contribution and Limitations of the existing methods
Authors Methods Contributions/unique characteristics Results Limitations
Sangeetha et al.
2022 [16]
Trust-based handover
authentication
scheme
To improve the security in SDN 5G
heterogeneous network. Mutual
authentication has been obtained with
three way handshaking
Better throughput, delay,
packet delivery ratio (PDR)
had attained
Need to improve the
handover process
Divakaran et al.
[17]
Enhanced handover
authentication model
based on fuzzy
evolutionary
optimization
To reduce complexity and delay in 5G
networks during the network
authentication process. Hanover
authentication process is enhanced with
fuzzy evolutionary process
Better in handling,
authentication and mitigation
against different attacks
Complexity still exists
Man Chun and
Maode Ma 2021
[18]
5G authentication
approach and key
agreement protocol
based on Blockchain
To establish a secured 5G network from
DoS attacks. Authentication and key
agreement was provided against DoS
with SUCI, and ECDH
Attained mutual
authentication, accurate
forward secrecy, key
agreement and device
anonymity
Security issues still exist
Xinghua Li et al.
[19]
FUIS handover
authentication model
To support inter-slice handover. Ticket
validation with privacy-preservation is
modelled with a ring signature for
minimizing overhead
Minimized handover over Need to minimize the time
Xudong Jia et al.
2020 [20]
A
2
chain To establish a secure authentication
information sharing process.
Authentication process is decentralized
with edge computing process
Minimized authentication
time, storage space and
communication cost
Need to improve the
authentication time
reduction and delay
Zaher Haddad
et al. 2020 [21]
Blockchain-based new
efficient, secured
authentication and
key agreement
protocol
To provide a secure and authentication
scheme for finding major attacks.
Secure authentication and registration
process is accomplished against DoS, a
man in the middle, distributed DoS,
compromising and Hijacking attacks
Secured and counter measured
major attacks
Need to minimize the
consensus delay
Abbas Yazdinejad
et al. 2019 [22]
Blockchain-based
authentication
handover model
To avoid the re-authentication in frequent
handover amid the heterogeneous cells
in SDN-based 5G networks. Privacy
protection is guaranteed with minimum
delay between heterogeneous cells
Attained less signalling, less
delay, and less energy
consumption
Minimize signalling
overhead and enhance the
performance
Amina Gharsallah
et al. 2019 [23]
SDN based handover
management process
To improve the handover process in 5G
UDNs. Handover control process is
managed with SDH-ME
Minimized handover delay and
handover failure ratio
Increased the risk
Jing Yang et al.
[24]
Fast and unified
handover
authentication
mechanism
To establish a fast and unique
authentication process in 5G
heterogeneous networksby considering
the channel characteristics
Minimized overhead and
latency
Susceptible to the channel
condition
Zhonglin Chen
et al. 2018 [25]
APG-PBFT consensus
mechanism
To organize APs into a secured, trusted
chain as APG for enhancing the
reliability and security of APG.
Consensus and message propagation is
utilized for authentication
Eliminated the frequency of
authentication if UE moved
amid the APs as well as
improved the access
efficiency
Need to improve the
consensus mechanism
Muhammad
Nabeel et al. [26]
TurboRAN testbed
deployement
Summarized relevant testbeds and
deployment of turboRAN testbed.
Handover procedures were investigated
with capability and functionality
By tuning the parameter
configurations, several
handovers occured for
achieving highest datarates
It is required to be
incorporated for massive
MIMO for enabling Sixth
Generation (6G)
communication
ShidrokhGoudarzi
et al. [27]
cooperative game
theory
An entire view of the network was
provided with SDN for end-to-end
policy formulation with high
computational power which is not
available at the UAVs
For large range of mobile
nodes, higher amount of
handoff strategy was
obtained
Frequent handover leads to
high energy consumption
Wireless Networks
123

data traffic. For users, UDHN is terermed as a direct access
network that uses APs to connect users to the 5G network.
However, in UDHNs, users require to secure access as well
as guarantee that access to the network is secured and UE
has not associated to an illegal or fake AP. So, it needs the
UDHN to guarantee that each AP that communicates with
UE should also be secured.
Figure 1illustrates the architecture of 5G UDHNs. In
this 5G UDHNs architecture, the user-centric UDHN has
been explored to constitute a assured range of APs around
the UEs, and APG accomplishes user service to the UE.
When the user moves, the members of the APG have been
dynamically updated so that the user can experience that
there contains a mobile network coverage linked with it.
Thus, the demand of mobile traffic can be effectively
addressed, and the user’s experience can be improved. In
the UDHNs architecture, the APGs are accountable for
accessing the UE connection. So as long as it is ensured
that APGs are secured (for instance, a rogue or fake AP
does not active in APG or not included in APG), it has been
guaranteed that the access of UE is secured without having
to need that every APs are secured nodes that will effec-
tually minimize the complexity of security protection. In a
5G heterogeneous network, either the members of APG or
the regular access nodes in UDHN, all APs are entirely
equivalent between one another, is termed as an organi-
zation that does not contain a center.
3.2 Problem formulation
The cyclic additive group is represented as G1and the
cyclic multiplication group is represented as G2. Both
groups are having the same primary order qand prepre-
sents the generator of G1. The bilinear mapping is formed
with G1G1!G2an dhteone way hash functions are
denoted as H1,H2and H3in which H1:f0;1g!G1,
H3:f0;1g ! Zq, and H2:f0;1g ! Zq. The public key
system parameters are initialized and the master key is kept
as secret. When the AP is having ID to joint with the
system, the AUC calculates the private key which is
securely sent to the APs. It is assumed that the UE with
ID 2f1;0gsends ID to AUC to start registration process.
If the ID is valid then AUC chooses collection of identities
for the computation of relevant private keys.
In the initial phase, a request for APG generation has
been accomplished. After providing a UE access request to
the network, the APs from a particular range around the UE
can accept the message and send a request. If the request
message is subsequently reached at LSC, the LSC can be
called to organize an APG for providing service to the UE.
LSC enquires the AUthentication Center (AUC) for the
APG key as well as other parameters to provide corre-
sponding data like APG unique IDentifier (APG-ID).
Moreover, the UEs have registered in LSC to construct
encryption materials depending on the possessions and
further accept the encryption possessions as well as key.
LSC forwards a vector consisting of UE information of the
cell simultaneously. For each UE, the LSC assigns two
Fig. 1 Architecture of 5G
UDHNs
Wireless Networks
123

keys, namely a private and a public key. The UE forwards
the join request to the LSC to enter the specified domain or
cell. This request was acknowledged through the authen-
tication control unit and forwarded to the UE after the
confirmation. The authentication control is applied to
determine the unique UE information like direction, iden-
tity, Round Trip Time (RTT), location, as well as private
and public key assignment.
The set of information from users registered in LSC,
namely the public key, has been forwarded to other cells,
then UE joints the cell. If the UE needs to hand over the
current AP to another AP in the identical cell, the UE
forwards the associate request to the target AP and dis-
connects with the current AP. Normally, the distinctive
behavior of UE has been shared amid neighboring and
adjacent cells, so there is no necessity for a re-authenti-
cation process while sending over the heterogeneous cells.
The LSC checks the AUC to guarantee that the UE is a
trusted cell. The UEs are registered in the accessible cell
and aims to move to the adjacent heterogeneous cell, which
forwards the request to AP of the identical cell. The private
key Qhas been known between the AP and UE, and it can
switch between the AP of a cell. Then, the public key Rhas
used in signing transactions as well as decoding the
information for privacy protection. The UE forwards the
message to other neighboring public key cells to avoid the
necessity for a frequent re-authentication process in the
heterogeneous cells. This process stimulates authentication
while transiting through APs and some heterogeneous
networks.
3.3 Blockchain-based handover and fast security
authentication
In order to provide a handover and fast authentication
process, a blockchain-enabled authentication handover
approach based on RLAC-FNN has been proposed. Block
chain approach is needed to combine the concept of
hashing algorithm, public key encryption, peer to peer
protocols, and consensus algorithm. It is basedon decen-
tralized network in which the main task is the protection of
stored list of records against tampering. Block chain
manages the database by enhancing the speed of opera-
tions, and information security level by reducing the
required time and individual error. The concept of block
chain mechanism does not require centralized data storage
or central supervisor. There is no additional requirement of
organizational authority instead of consensus algorithm for
managing the decentralized network.
Then, for authentication, the LSC sends an instruction
for consensus computing to all AP providing service to UE.
The new consensus computing process is based on RLAC-
FNN. Based on the consensus outcomes, the trusted APs
are selected since there contain one or more fake or
untrusted AP.
3.4 New consensus computing process based
on RL with the ACbased on FNN
In the blockchain-based 5G networks, the main objective is
to attain a consensus on the blockchain transaction infor-
mation in complete network. A new consensus computing
algorithm using RLAC-FNN has been proposed the process
of consensus computation. Relating to the blockchain
decentralized scheme, the Client-to-Server (C/S) paradigm
has shifted to the P2P paradigm. Thereby, the system
contains no client. The consensus node is divided into three
categories: Master Node (MN), CN and Sub-Node (SN).
For each consensus node, the credit attribute has been set
so that the system will dynamically split the consensus
node, and the nodes will leave and join the system
dynamically. Then some of the elementary parameters
applied in the system can be provided as:
•Set an attribute credit Eas an index to determine the
reliability of consensus node’s, which means the
probability of participating in consensus successfully.
It is considered as a significant basis for dividing the
consensus nodes.
•The node with the highest credit value is assigned as the
MN. Psignifies the number of consensus nodes, where
P3Gþ1 and Gresembles the maximum number of
malicious nodes a system will bear. Normally, Puses
3Gþ1.
•The value of admission credit EBAS has been preset.
P1 CNs with credit EEBAS are chosen as SNs. The
nodes with credit value less than EBAS or recently
included nodes are chosen as CN.
During system initialization, the credit value of each
consensus node is set to EBAS and Pconsensus nodes have
chosen randomly as SN. Then, the MN has elected from the
SN, and the number qis allocated. Because the blockchain
consensus system is decentralized and P2P-based, every
consensus node must initiate in an identical state. This
resembles that the information deposited in each consensus
node requires consistency. To guarantee consistency needs
verification and data backup. After completing the backup
as well as verification process, the model initiates the
consensus process. In the proposed consensus process,
message delivery applies digital signature schemes as well
as Secure Hash Algorithm-256 (SHA-256) to guarantee the
authenticity and message integrity. The consensus process
is initiated by the MN and set as the interval for MN
consensus as U. Then the particulars of the proposed
algorithm are provided below as follows.
Wireless Networks
123

•The initiator starts the transaction through signing the
transaction as well as broadcasting it to the consensus
node.
•The legality of the received transaction is verified by
the consensus node. When the received transaction is
illegal, the transaction is discarded directly. Otherwise,
it is considered legal, so the transaction information has
been cached, as well as the MN constructs a block.
•After U, a proposal message is created by the MN qfor
the constructed block and provide a unique number pto
the proposal. The unique number can maximize with
each newly constructed proposal. Moreover, to request
the SN for participation, the MN qbroadcasts the
proposal message to the SN. During the consensus
process based on credit value, the messaging format is
assigned as \\ Pr oposal message;q;p;Mes:Sig;
e;[;Block [, where, eresembles the message
digest determined using the SHA-256 algorithm and
Mes:Sig indicates the message signature.
•First, the SNs verify the MN number, proposal number,
and proposal’s message signature. The SNs will
forward a verification message to MN if the verification
has passed. Then the format for sending the verification
message is given as \consensus:com;q;p;Mes:Sig;
e;st;l;D[, where, consensuscom represents the
consensus confirmation lindicates the SN number. st
resembles the verification type of SN lto the proposal
message digest. Ddenotes the reputation of the SN l.st
can be either 1 or 1, relating to false or true,
considerably, where true represents the proposal mes-
sage digest eis consistent with cached data through the
SNs. Else, false indicates the inconsistency, and the SN
mistrusts the MN. On the other hand, if the verification
flops, the SN has discarded the proposal message.
•Then, the MN utilizes RL to carry out consensus
confirmation for the received verification message.
When the MN has received 2Gconfirmation messages,
it is regarded that the consensus is attained as well as
published a block. However, if the MN has received 2G
suspicious messages, the consensus node broadcasts a
message to change and reselect the MN. During the
consensus timeout, if sufficient messages are not
received, every node discards the generated block at
the consensus computation process, reselect the MN,
and minimize the credit of the discarded MN. Further,
again re-elect the MN but no longer choose the original
MN. Eventually, it generates a new round of consensus.
When the MN successfully publishes the block, and
after publishing the newly generated block, such node
forwards a credit adjustment message to each SN.
•If the consensus node has received the published block,
this consensus process has finished. Further, such a
node can update the credit corresponding to the
adjusted data, clear the cache, update the consensus
node, as well as establish a fresh round.
The proposed algorithm of RLAC-FNN is given in
algorithm 1 and the SHA-256 based block chain based
approach is given in algorithm 2.
Algorithm 1: RLAC-FNN for consensus confirmation
Input: data YðuÞ
Step 1: The parameters, learning rate and weight vectors are
initialised
Step 2: Then the amount of fuzzy rules n1
jand error are computed
Step 3: Follow step 4 to step 6 when the number of iterations uis
not reached
Step 4: Produce the vectors of Bernoulli variables for updating
truth value
Step 5: Ordinal index of fuzzy rules and output is calculated for
estimating error
Step 6: The process is terminated if the calculated error is less
than the error tolerance or the maximum number of iterations
are reached. Otherwise, the gradient decent of vector is
calculated and the parameters are updated with RLAC
Step 6.1: Collect the parameters stðlÞ,DðlÞand as vðlÞfrom FNN
as state for RLAC for a specific time
Step 6.2: for each fuzzy rule knormalized firing strength Pk
wis
computed
Step 6.3: The critic value function and action output is calculated
Step 6.4: After calculating and invoking the actual action the
reward is received and the state is transmitted
Step 6.5: then the prediction error function fdðlÞand critical value
function FdðlÞis calculated
Step 6.6: Condition for rule unit is checked and if, the conditions
are satisfied then the parameters are updated else continue to the
next step
Step 6.7: Weight, centre and depth are adjusted
Step 6.8: It is checked that whether the units are required to be
merged or not. If the condition is satisfied then similar units are
merged otherwise next step is proceeded
Step 6.9: The number of iterations for RLAC is checked. If the
number of iterations are not completed then it is continued from
8.2
Step 6.10: After updating the parameters, it is repeated from step
4 to step 8 until the number of iterations are attained
Step 8.11: KðlÞis returned to step 8
Output:consencus confirmed or not
Wireless Networks
123

Algorithm 2: SHA-256 based block chain approach
Input: message
Step 1: Som additional bits are added to the input message in
which the message length is 64 bit shorter than the multiple of
512. While adding the bits, the first bit is one and the remaining
bits are filled with zero
Step 2: The charecters for this 64 bits are computed by applying
modulo operation to the original text without padding
Step 3: The default values are initialized for eight buffers
Step 4: Then the entire message is divided into several blocks and
each block contains 512 bits
Step 5: Each block passes through 64 rounds operations and the
output of each block is fed to the input of next block
Step 6: The entire process is completed until the last 512-bit
block is processed., Then the output is considered as final hash
digest
Output: 256 bit hash digest
3.4.1 Consensus confirmation
The consensus confirmation process integrates the AC
model and FNN. It primarily consists of an action network
that generates AC network, which has been employed to
compute the actions. The primary objective of AC is to
decide whether the SNs are in agreement and establish a
source for credit adjustment. Initially, the structural
learning of AC-FNN has applied to determine the number
of AC-FNN rules and assign the initial values for the
parameters of the AC network. Online structural learning is
a process that provides if–then rules depending on the input
data. It has been performed using fuzzy clustering that
determines the fuzzy rule based on the rule firing strength.
The first input data YðuÞis utilized to determine the first
fuzzy rule. The first epileptic Interval Type 2 Fuzzy Set
(IT2-FS) parameters for the first if–then rule are set in
Eq. (1).
n1
j¼yjðuÞ;11
j¼11
jFixed ;j¼1;2;3; ::::zð1Þ
where, n1
jindicates the mean of the membership function,
11
jFixed resembles the pre-defined value that must be greater
than 0, which denotes the interval range of initial IT2-FS.
The firing strength is computed at each iteration u. Further,
the average value of firing strength Pk
wis computed for
each rule k. After this, for subsequent input data YðuÞ, the
determination of the number of rules has computed
depending on J¼Arg Max
1kNhðuÞ
Pk
w, where, Jrepresents the
number of hidden rules,NhðuÞdenotes the number of
immediate rules at iteration u. Moreover, a new if–then rule
has determined at Nhðuþ1Þ¼NhðuÞþ1ifPj
wPth,
where Pth resembles the pre-determined threshold. The
consensus confirmation process based on AC-FNN is given
in Fig. 2.
The architecture of both the actor-network and the critic
network is typically the same. Both adapt a backpropaga-
tion Neural Network (NN) with the hidden layer. Consider,
stðlÞand DðlÞin the lth verification message as input for the
action network, obtain an output vðlÞ. The critic network
uses stðlÞ,DðlÞas well as vðlÞas input and provides the
output KðlÞas the approximate value of return HðlÞ, which
has been applied to evaluate the critic network’s outcomes
considerably. So, the approximation value has been utilized
to simulate the outcome of vðlÞbetter. Further, the action
output vðlÞand stðlÞare compared to choose the rein-
forcement signal SðlÞ. When vðlÞand stðlÞcontain identical
signs, the output is considered as successful and SðlÞis
provided as ‘‘0’’ whereas, if it contains different signs, the
output is considered as unsuccessful and SðlÞis regarded as
‘‘1’’. The schematic diagram of the critic network is
depicted in Fig. 3.
Initially, the weight of the critic network and actor-
network are assigned randomly at the consensus process.
During the NNs learning process, the critic network
employs SðlÞfor updating the weight and estimating the
optimum KðlÞ. Then the actor-network utilizes the opti-
mum KðlÞto update weight and attain the optimal
outcomes.
3.4.2 Critic network
The learning objective of the critic network is to reduce the
error between the actual value and approximation value of
the value function while optimizing for highest return.
Then the prediction error function fdðlÞof the critic net-
work is expressed in Eq. (2).
Fig. 2 Consensus confirmation based on AC-FNN
Wireless Networks
123

fdðlÞ¼bKðlÞ½Kðl1ÞSðlÞ ð2Þ
where, brepresents the constant parameter. The prediction
error has the property of approaching zero over time while
converging actor critic learning. If the distribution error
and average error is zero then the average prediction error
is large. The prediction error in the critical network is
computed by learning the reinforcement signal. The mini-
mized objective function FdðlÞof the critic network is
given in Eq. (3).
FdðlÞ¼1
2f2
cðlÞð3Þ
Besides, the output KðlÞin the critic network can be
expressed in Eqs. (4), (5), and (6).
cjðlÞ¼X
pinþ1
k¼1
xð1Þ
djk ðlÞykðlÞ;j¼1;2;3::::::; Pið4Þ
ejðlÞ¼1expcjðlÞ
1þexpcjðkÞ;j¼1;2;3; :::::; Pið5Þ
KðlÞ¼X
Pi
j¼1
xð2Þ
djðlÞejðlÞ;ð6Þ
where, cjindicates the jth hidden node input for the critic
network, Piresembles the total number of hidden nodes
present in critic network, ejrepresents the corresponding
output. xdsignifies the weight vector in the critic network,
ykðlÞrepresents the input vector, Pin þ1 defines the total
number of inputs provided to critic network, and analog
action value from the action network.
Corresponding to the chain rule and error propagation
equation of the backpropagation algorithm, the gradient of
NN objective function to weight has been determined. It
can be provided below as follows.
For the hidden layer output is represented with Eqs. (7),
(8), and (9).
Dxð2Þ
djðlÞ¼mdðlÞ
oFdðlÞ
oxð2Þ
djðlÞ
2
43
5ð7Þ
¼mdðlÞoFdðlÞ
oKðlÞ
oKðlÞ
oxð2Þ
djðlÞ
2
43
5ð8Þ
¼mdðlÞbfdðlÞejðlÞ
 ð9Þ
The critic parameters are updated based on prediction
error of critic network. The fuzzy interference system has
certain equivalence when the number of conditions are
satisfied. The basis function of hidden layers aresame as
that of the membership function of the input vector. In
addition to that the part of each fuzzy rule is equal to the
connection weight between the output layer and hidden
layer. Each hidden layers are represented with fuzzy rule.
For input to the hidden layer, the representation is given
with Eqs. (10), (11), and (12).
Dxð1Þ
djk ðlÞ¼mdðlÞ
oFdðlÞ
oxð1Þ
djk ðlÞ
2
43
5ð10Þ
¼mdðlÞoFdðlÞ
oKðlÞ
oKðlÞ
oejðlÞ
oejðlÞ
ocjðkÞ
ocjðlÞ
oxð1Þ
djk ðlÞ
2
43
5ð11Þ
¼bmdðlÞfdðlÞxð2Þ
diðlÞ:
1
21e2
jðlÞ


ykðlÞð12Þ
where, mdðlÞ[0 signifies the learning rate of critic net-
work while considring jth message. Normally, this rate can
reduce overtime to less value. At each iteration, critic
output determines discounted total reward to resolve the
issue of infinite horizon.
3.4.3 Action network
As discussed earlier, the output is pre-defined and if the
outcome SðlÞis ‘‘0’’, then the output is successful. That is,
the output ‘‘0’’ is signified as the reinforcement signal
merely for success. To persuade Bellman equation as well
as increase the state value function, the final learning target
of the action network represented as Vd, has set to 0. While
analyzing in action network, it is determined that the
parameter adjustment principle is to indirectly backpropa-
gate the error between Vdand K. The network learning can
beaccomplished through prediction error function of the
network. Then, the prediction error function fbðlÞand
objective function FbðlÞof active network can be defined in
Eqs. (13), and (14).
fbðlÞ¼KðlÞVdðlÞð13Þ
Fig. 3 Schematic diagram of critic network
Wireless Networks
123

FbðlÞ¼1
2f2
bðlÞð14Þ
The action network accommodates the NN model, the
same as that on the critic network. Then, the action network
is characterized in Eqs. (15), (16), (17), and (18).
njðlÞ¼X
pin
k¼1
xð1Þ
bjk ðlÞykðlÞ;j¼1;2;3; :::::; Pið15Þ
ajðlÞ¼1expnjðlÞ
1þexpnjðlÞ;j¼1;2;3; ::::::; Pið16Þ
iðlÞ¼X
Pi
j¼1
xð2Þ
bjðlÞajðlÞð17Þ
vðlÞ¼1expiðlÞ
1expiðlÞð18Þ
where, njindicate the input of jth hidden node for action
network, iðlÞresembles the input of output node, xb
resembles the weight vector, ykðlÞrepresents the input
vector,ajrepresents the corresponding output. Like the
critic network, the parameter updating rule of action net-
work has provided below as:
For hidden to the output layer, the representation is
given in Eqs. (19), (20), and (21).
Dxð2Þ
bðlÞ¼mbðlÞ
oFbðlÞ
oxð2Þ
bjðlÞ
2
43
5ð19Þ
¼mbðlÞoFbðlÞ
oKðlÞ
oKðlÞ
ovðlÞ
ovðlÞ
oiðlÞ
oiðlÞ
ox2
bj
"#
ð20Þ
¼fbðlÞ1
21v2ðlÞ


ajðlÞX
Pi
j¼1
1
2x2
djðlÞð1e2ðlÞÞxð1Þ
dj;pþ1ðlÞ

ð21Þ
The output layer is made up of actor and critic part in
which the critic part contains the estimation for state value
function. Mapping is determined from the state to expected
critic value. The actor part considers as action selector
which provides mapping from state space into action space.
When the resolution of state space is insufficient, the
variation of prediction error is high through learning value
function in space converges with partial observations.
For input to the hidden layer, the representation is given
with Eqs. (22), (23), and (24).
Dxð1Þ
bðlÞ¼mbðlÞ
oFbðlÞ
oxð1Þ
bjk ðlÞ
2
43
5ð22Þ
¼mbðlÞoFbðlÞ
oKðlÞ
oKðlÞ
ovðlÞ
ovðlÞ
oiðlÞ
oiðlÞ
oajðlÞ
oajðlÞ
onjðlÞ
onjðlÞ
oxð1Þ
bjk ðlÞ
2
43
5ð23Þ
¼mbðlÞfbðlÞ1
21v2ðlÞ


x2
bjðlÞ1
21a2
jðlÞ


ykðlÞ:X
Pi
j¼1
1
2x2
djðlÞð1e2ðlÞÞxð1Þ
dj;pþ1ðlÞ

ð24Þ
where, mbðlÞ[0 defines that the learning rate for action
network is considered with jth message. Subsequently, a list
PU and PG is defined to record nodes where stis 1 or 1.
The list PU accumulates where stis 1 whereas the list PG
accumulates where stis 1. If the number of nodes in PU
is equal to or greater than 2G, then the consensus is con-
firmed. On the other hand, if the number of nodes PG is
equal to or greater than 2G, the SN does not trust the MN
respectively.
4 Result and discussion
In this part, the outcomes of the proposed method have
been discussed. The performance of RLAC-FNN can be
determined based on different metrics and compared with
existing algorithms to evaluate its effectiveness. The
implementation of RLAC-FNN is carried out using the
Python platform. Then, the result of the RLAC-FNN is
compared with recent existing methods. The network
consisting of 30 heterogeneous cells with a distance of
200 m between two APs are selected to determine the
comparability of the consensus approach based on RLAC-
FNN among 5G heterogeneous cells. Table 3illustrates the
network parameters used for simulation. All the values of
the parameters are selected from the previos research as
well the performance of the proposed approach with pos-
sible parameter values. By using this parameter values
better performance is obtained in the exiting researches
[28] and our proposed work.
Table 3 Simulation parameters employed for evaluation
Parameters Values
Number of cell 30
Number of transaction 1200
Cell radius 100 m
Distance between two AP 200 m
Number of users 600
User mobility direction Random
Receiving power 1340 mW
Transmit power 1726 mW
Block size 4 byte
Wireless Networks
123

While testing the performance, the rules to adjust the
credit value can be shown below as follows:
•Initially, the credit of all nodes indicated as DBASE ,
which is assigned to 3.The reliability of consescus
nodes are measured with credit value. The node with
higher credit can be participate in the consensus. Hence,
the value DBASE is set as lower than than the possible
credit value of non malicious node from the previous
iteration.
•The credit value of a MN is maximized by 0.2 if it
finishes a consensus. The credit value of the SN is
reduced by 0.1 if the verification type of the SN is false.
Whereas the credit value of the SN is maximized by 0.1
if the verification type of SN is true. In addition, the
credit is minimized by 0.1 to a SN, which is not
responding over time. These values are selected from
the possible credit maximization or minimization
values which are applied to the proposed algorithm.
From the possible values the best one is selected based
on the performance of proposed algorithm.
•The credit of the MN is set to 2 when the SN guess that
the MN is successful.
•The MN will be reduced by 1 if it has not finished the
consensus process within the timeout.
Subsequently, some of the specific settings of NN have
been illustrated as follows:
•The internal cycle of the action network Pbis set to 100.
•The internal cycle of the critic network Pdis set to 50.
•The initial learning rate of the action network mbð0Þis
set to 0.3.
•The initial learning rate of the critic network mdð0Þis
set to 0.3.
•The learning rate of the action network mbðuÞat time u
is minimized by 0.05 until it attains 0.005.
•The learning rate of the critic network mdðuÞat time uis
minimized by 0.05 until it attains 0.005.
•The number of hidden nodes Piis considered as 6.
•For an action network, the internal training error
threshold Ubis set as 0.005 (the threshold of FbðlÞ).
•For a critic network, the internal training error threshold
Udis set as 0.05 (the threshold of FdðlÞ).
Moreover, in this evaluation, the proposed method has
considered 10 consensus, and the value Gis set to 3 and P
to 13. The outcomes of the simulation are depicted in
Fig. 4.
4.1 Performance metrics
This section discusses various performance metrics that
were considered to determine the simulated results of
RLAC-FNN. RLAC-FNN considered various metrics such
as consensus delay, consensus time, signalling overhead,
handover authentication delay and energy consumption for
evaluation.
a. Consensus delay
Consensus delay is considered a significant metric to
compute the speed of the consensus process. The minimum
consensus delay can quickly confirm the transaction,
making blockchain outcomes more practical and stable.
Indeed, the consensus delay Ueevaluated in the proposed
method is the consensus completion time, and it is
expressed in Eq. (25).
Ue¼Uud Uus ð25Þ
where, Uud resembles the transaction’s start time and Uts
indicates the completion time of consensus. Here, the value
of Ghas set to 1, 2, and 3, and the value of Phas set to 4, 7,
10, considerably.
b. Consensus time
It is a taken between the UE, AP and AUC to authen-
ticate the particular UE for accomplishing handover. It
depends on the organization of network structure and the
rules of interaction for measuring the system reliability.
c. Authentication delay
After sending request to access the duration taken for
getting authentication response is refered as authentication
delay. It includes the process of request generation, vali-
dation with AUC, getting response from AUC and sending
authentication message to the UE. Normally, when the
network load is low then the authentication delay is neg-
ligible. With increasing the network load, data transmission
operation and mobility among cells, the authentication
delay is increased.
d. Energy consumption
Fig. 4 Output of simulation
Wireless Networks
123

If the received rate DRX is the received rate and DTX is
the transmit rate, the energy consumption can be computed
in Eq. (26).
F¼½DTX QUT1þ½DRX SYT1þ½QRðT
T1Þ
ð26Þ
where, QUindicates the transmit power, SYresembles the
receiving power, QRsignifies the received power, Trep-
resents the initial time, and T1 represents the connection
time.
e. Signalling overhead
Signalling overhead comprised additional or pattern
information to improve the performance of wireless com-
munication. This overhead relates to the register UE in
LSC. The signalling overhead is compared to network-
based, POW-based and BAHEPP models. Then, the sig-
nalling overhead can be expressed in Eq. (27).
WOH ¼CN
T

þZ
Tð27Þ
where, WOH indicates the signalling overhead, Trepresents
the time, Crepresents the steps count between UE and
LSC. Zresembles the length of packets forwarded to LSC
and Nsignifies the length of packets registered in LSC.
4.2 Performance evaluation
Besides, the result of RLAC-FNN is demonstrated and
compared with recent existing algorithms such as PBFT,
Credit Reinforcement Byzantine Fault Tolerance (CRBFT)
algorithm, network-based model, POW-based model,
blockchain-enabled Authentication Handover with Effi-
cient Privacy Protection (BAHEPP) model Fast and
Universal Inter-Slice (FUIS) [19] and Lightweight and
Secure Handover Authentication (LSHA) [29]
considerably.
The working principle of proposed fast authentication
process is given as follows. Initially all the parameters are
initialised for the proposed 5G UDHN network model.
After network deployment and parameter initialization, the
request is send for APG generation at initial phase. After
granting the access request of UE, the AP accept and send
it to LSC. LSC interacts with the AUC for generating the
parameters such as APG key and APG ID. UE is registered
with LSC which provides private and public key to the UE.
The join request for UE is forwarded to the LSC for
entering into the specific cell. This request is acknowledged
with AUC and the confirmation is sent to the UE. The
authentication control is applied for finding the UE infor-
mation as well as private and public key management.
During the process of authentication, the block chain based
approach with SHA-256 is utilized for fast and efficient
authentication. For authentication LSC sends consensus
computing to all AP with the proposed algorithm RLAC-
FNN.
Then, the consensus delay attained by the proposed
method has depicted in Fig. 5.
In Fig. 5, it is determined that when the number of nodes
maximizes, the consensus delay of the proposed, PBFT and
CRBFT algorithm can also be maximized. However, the
maximization of PBFT and CRBFT is more than proposed
RLAC-FNN. So increasing the consensus node contains a
major impact on PBFT and CRBFT algorithms. Moreover,
for varying numbers of nodes, the consensus delay of
RLAC-FNN is considerably less than PBFT and CRBFT
algorithms. Thus, the proposed method is superior to
existing algorithms.
Table 4illustrates the consensus delay performance
comparison with RLAC-FNN and existing PBFT and
CRBFT algorithms. The consensus is determined for
varying the number of nodes to 4, 7, and 10. The RLAC-
FNN has attained better outcomes of 191.94 ms for 4
nodes, 239.143 ms for 7 nodes, 0.21 s for 270.32 ms for 10
nodes, compared with existing algorithms considerably.
The performance of consensus delay is compared with
the exiting approaches such as CRBFT, PBFT, FUIS and
LSHA in Fig. 5and Table 4. When compared with the
exiting approaches, the performance of the proposed
approach is lower in terms of delay. For the proposed
approach, the delay is in the range of 200 ms. It is due to
the fast consensus process of proposed RLAC-FNN. It
computes the consensus with minimal time duration by
effectively learning the parameters used for the process.
Efficient parameters are chosen with the RLAC based
Fig. 5 Performance evaluation of consensus delay with proposed and
existing algorithms
Wireless Networks
123

approach which provides better weight value with minimal
computational complexity. Hence, the FNN accomplishes
the weight updating process within short duration and it
enhances the computation process of FNN. Meanwhile, the
consensus confirmation efficiency is improved with fast
computation of consensus information.
The performance evaluation of the single consensus time
is given in Fig. 6. In the figure, it is evidently seen that the
proposed method takes more time in the first consensus, and
with the learning of NN, the consensus time minimizes till
the learning finish., For the first consensus, RLAC-FNN has
taken 1606.76 s followed by 670.78, 655.18, 608.38, 389.99,
389.99, 358.79, 358.79, 343.19, and 311.99, considerably.
The handover authentication delay of the proposed
method is related with existing network-based, POW-based
and the BAHEPP model [22]. In the existing POW-based
model, the users should register in the blockchain and upon
continuous displacement of the cells, it becomes re-au-
thentication in the cell. Both network-based as well as
POW-based schemes separate protocol and re-approval
amid heterogeneous cells for the authentication process. In
RLAC-FNN, the user cannot require to be re-authenticated
if being placed in the heterogeneous cells since they are
valid in neighbour cells as well as handover simply by
eliminating the re-authentication delay.
In Fig. 7, the handover authentication delay is shown
with varying rate of network utilization in 5G
heterogeneous network. The proposed delay is compared
with the existing POW based, Network based, BAHEPP
approaches. With low network utilization, the handover
authentication delay is in the range below 0.2 ms. If the
network utilization is increased to 0.6, then the delay is
slightly increased for the existing approaches. For the
proposed approach, it is below 0.2 upto the network uti-
lization rate 0.8. When it is increased above 0.8, the han-
dover authentication delay is slightly increased.
Authentication delay is in the range between 0.5 and 3
when the network utilization is 1. FUIS and LSHA per-
formance is lower when compared to that of the proposed
approach. When compared with the existing approaches,
the handover authentication delay is lower for the proposed
approach.
Table 5compares handover authentication delay per-
formance with RLAC-FNN and existing network-based
model, POW-based model and BAHEPP. The table clearly
shows that the proposed consensus mechanism based on
RLAC-FNN has minimized the handover authentication
delay more than the existing models due to eliminating the
re-authentication process. The handover authentication
delay given in Table 5is lower for the proposed approach
and it is higher for the existing approaches. Upto the net-
work utilization is 0.3 is 0.1 for all the approaches. By
increasing the network utilization, the handover authenti-
cation delay is increased heavily for the existing
Table 4 Comparative analysis
of consensus delay Number of nodes Consensus delay (ms)
PBFT CRBFT FUIS LSHA RLAC-FNN (Proposed)
7 800 400 550 590 239.143
10 1250 600 873 900 270.32
4 500 300 1350 1410 191.94
Fig. 6 Performance evaluation of single consensus time with
proposed and existing algorithms
Fig. 7 Performance evaluation of handover authentication delay with
proposed and existing models
Wireless Networks
123

approaches. But for the proposed approach all the values
are below 0.6. This deviation is due to the proposed fast
handover algorithm which utilizes the weight optimization
process for FNN. Hence efficient weight updating process
is accomplished with minimal duration. During handover
process the consensus confirmation can be accomplished
with the proposed algorithm which provides fast confir-
mation process and it minimizes the time required for
authentication.
Figure 8signifies the performance evaluation of energy
consumption with proposed and existing models. Accord-
ing to Fig. 8, the proposed method has attained minimum
energy consumption than existing models. In contrast with
the existing models, the network-based model has
increased energy consumption due to the frequent re-au-
thentication process.
The performance comparison of energy consumption
with RLAC-FNN and existing network-based model,
POW-based model and BAHEPP are illustrated in Table 6.
The initial energy level is considered as 0:1104and by
increasing the time duration the energy consumption is
increased for all approaches. The comparison shows that
the proposed method has obtained less energy consumption
than existing models. Moreover, BAHEPP has consumed
less energy close to the proposed method, and the network-
based model reached higher energy consumption among all
existing models. The time represents the processing time of
request registration and authentication among cells.
Figure 9depicts the performance evaluation of the sig-
nalling overhead with proposed and existing models such
as network-based, POW-based and BAHEPP models. In
the graph, it is observed that the proposed method achieved
less overhead than all existing models. However, the POW-
based model has applied consists of more overhead.
Besides, the network-based model can need different
authentication servers and third parties in communication
among heterogeneous cells. BAHEPP attained minimum
overhead closed to the proposed method since it has reg-
istered directly on the blockchain centre and interacts with
the cell without communicating with other APs. When
comparing the energy consumption of the proposed
approach, it is higher for the existing approaches. It is due
to the fast computing process which requires less energy.
Since the time required for the authentication process is
low and the authentication delay is minimal, the energy
consumption is lower with the proposed approach. Efficient
computation process involved in the block chain process
with SHA-256 minimizes the energy consumption of the
process.
The performance comparison of energy consumption
with RLAC-FNN and existing network-based model,
POW-based model and BAHEPP are stated in Table 7. The
new consensus mechanism based on RLAC-FNN has
considerably achieved minimum overhead than all other
existing models such as network-based, POW-based, and
BAHEPP. It is noted in the table that increasing the time
maximizes overhead due to more requests for authentica-
tion and registration among the heterogeneous cells.
Table 5 Comparative analysis
of handover authentication
delay
Network utilization Handover authentication delay (ms)
Network-based model POW-based model BAHEPP FUIS LSHA Proposed
0.1 0.1 0.1 0.1 0.1 0.1 0.052
0.2 0.1 0.1 0.1 0.1 0.1 0.036
0.3 0.1 0.1 0.1 0.1 0.1 0.03
0.4 0.21 0.2 0.1 0.1 0.1 0.057
0.5 0.28 0.2 0.1 0.31 0.43 0.05
0.6 0.4 0.2 0.1 0.38 0.52 0.057
0.7 1.69 0.8 0.37 0.47 0.55 0.074
0.8 1.8 0.1 0.4 0.5 0.64 0.13
0.9 2.2 0.2 0.61 0.65 0.73 0.081
Fig. 8 Performance evaluation of energy consumption with proposed
and existing models
Wireless Networks
123

However, the proposed method achieved a minimum
overhead than existing models respectively.
The comparison of signaling overhead was given in
Fig. 9and Table 7. When compared with network based,
BOW based, FUIS, BAHEPP and LSHA models, the per-
formance is lower for the exisitng approaches. For the
existing approaches, the signaling overhead is higher than
21 bytes for all approaches. But in case of RLAC-FNN, it
is on or below 20 and it is not increased with increasing the
time duration. With the utilization of proposed block chain
based approach, the input information is divided into sev-
eral blocks and the fast encryption process is accomplished
with SHA-256. Hence the size for the message send for
each communication is lower for the existing approaches. It
can be possible with the proposed weight updating process
and fast consensus confirmation process. The proposed
algorithm minimizes the number of bits send and by min-
imizing the number of bytes, the overhead is also
minimized.
Subsequently, from the overall analysis, it is observed
that the new consensus mechanism based on RLAC-FNN is
superior to existing methods and suitable for fast authen-
tication and reducing the handover delay in 5G heteroge-
neous networks.
4.3 Computational complexity analysis
The complexity of the proposed algorithm is based on the
handover procedure, authentication algorithm such as
block chain based and SHA-256 encryption, and the
Table 6 Comparative analysis of energy consumption
Time (sec) Energy Consumption ðj104Þ
Network-based model POW-based model BAHEPP FUIS LSHA RLAC-FNN (Proposed)
20 3.2 2.9 1.9 4.5 5.3 1.213
40 5.1 4.7 2.6 6.1 6.5 1.216
60 5 4.9 3.2 6.8 7 1.217
80 7.3 7 3.9 8 8.9 1.217
100 9.3 7.8 4.5 10.5 12.7 1.219
120 11.7 9.1 4.9 11.9 13 1.22
Fig. 9 Performance evaluation of signalling overhead with proposed
and existing models
Table 7 Comparative analysis of signalling overhead
Time (sec) Signalling overhead (bytes)
Network-based model POW-based model BAHEPP FUIS LSHA RLAC-FNN (Proposed)
20 73 49 20 85 92 18.48
40 80 58 30.5 90 96 18.79
60 100 65.3 35.8 105 110 19.12
80 125 70.5 40.3 130 140 19.45
100 145 80.5 45 153 165 19.8
120 175 110 50 182 193 20.16
140 190.5 125.5 55 195 198 20.53
Wireless Networks
123

proposed consensus confirmation process. Based on the
number of AP and the UE and the amount of registration
process the computation complexity of the handover pro-
cess is Oðn2Þ. Then the process of authenticating the UE
during the handover process has the complexity OðnÞ.Itis
assumed that the random variable is represented with J
which denotes the amount of invalid blocks generated
between the consecutive blocks. In addition to the esti-
mated value the complete histogram of Jis generated with
its properties in block chain process. The time required for
each block creation is Tland the length of block chain is hl
in which lrepresents each block. Due to the complexity of
time and matrix with number of iterations has the com-
plexity OðnmÞ. Where, mand nis associated with time and
block length. By adding SHA-256 to the proposed block
chain model, the complexity becomes Oðnm þm2Þ. Then
the consensus confirmation can be verified with the pro-
posed model RLAC-FNN in which the complexity required
is Oðu;N2ÞþOðm2n2Þ. Where, udenotes the number of
layers,Nis the complexity of fuzzification process, and
actor and critic network complexity is represented with m
and n.
5 Conclusion
This paper proposes a new consensus mechanism based on
RLAC-FNN and an authentication handover model to
improve the user authentication and handover delay in 5G
heterogeneous Networks. The credit value has been set to
all APs and RL adjusts it in the proposed approach. RLAC-
FNN can detect the failure AP and the illegal AP in the
consensus network, thereby enhancing consensus delay,
consensus network security, energy-saving etc. Finally, in
contrast with existing PBFT, CRBFT, network-based
model, POW-based model and BAHEPP, the proposed
consensus mechanism based on RLAC-FNN has achieved
minimum consensus delay, minimum signaling overhead,
less energy consumption, less consensus time, and mini-
mum handover authentication delay, considerably.
In the future, a new solution will be determined to
enhance the stability of NN and increase the learning
process, along with security enhancement to apply in var-
ious kinds of blockchain. The proposed handover authen-
tication process is fast and efficient but slightly far away
from complete and perfect. It requires several essential
consideration for making the system robust and practicable.
The proposed approach is affected with noises and channel
conditions and hence higher authentication is required with
the estimation of better channel conditions. The future
work of this research focuses on authentication perfor-
mance, resource consumption and time latency in
unconstrained environment. In addition to that energy
efficient protocols will be modelled with mobility man-
agement for enhancing the handover authentication
through better mobility prediction.
Author contributions All authors have made equal contributions to
this work.
Funding No funding is provided for the preparation of manuscript.
Data availability Data sharing not applicable to this article as no
datasets were generated or analysed during the current study.
Declarations
Conflict of interest Authors have declared that they have no conflict
of interest.
Ethical approval This article does not contain any studies of human
participants or animals performed by any of the authors.
Consent to participate All the authors involved have agreed to par-
ticipate in this submitted article.
Consent to publish All the authors involved in this manuscript give
full consent for publication of this submitted article.
References
1. Alezabi, K. A., Hashim, F., Hashim, S. J., Ali, B. M., & Jama-
lipour, A. (2020). Efficient authentication and re-authentication
protocols for 4G/5G heterogeneous networks. EURASIP Journal
on Wireless Communications and Networking, 2020, 1–34.
2. Zhou, Z., Chen, X., Zhang, Y., & Mumtaz, S. (2020). Block-
chain-empowered secure spectrum sharing for 5G heterogeneous
networks. IEEE Network, 34(1), 24–31.
3. Kumar, A., & Om, H. (2018). Handover authentication
scheme for device-to-device outband communication in 5G-
WLAN next generation heterogeneous networks. Arabian Jour-
nal for Science & Engineering (Springer Science & Business
Media BV), 43(12), 7961–7977.
4. Zhang, Y., Deng, R., Bertino, E., & Zheng, D. (2019). Robust and
universal seamless handover authentication in 5G HetNets. IEEE
Transactions on Dependable and Secure Computing, 18, 858–874.
5. Alquhayz, H., Alalwan, N., Alzahrani, A. I., Al-Bayatti, A. H., &
Sharif, M. S. (2019). Policy-based security management system
for 5g heterogeneous networks. Wireless Communications and
Mobile Computing, 2019, 1–14.
6. Nyangaresi, V. O., Rodrigues, A. J., & Abeka, S. O. (2020).
ANN-FL secure handover protocol for 5G and beyond networks.
In International Conference on e-Infrastructure and e-Services
for Developing Countries, Springer, Cham, 99–118.
7. Yang, H., Liang, Y., Yuan, J., Yao, Q., Yu, A., & Zhang, J.
(2020). Distributed blockchain-based trusted multidomain col-
laboration for mobile edge computing in 5G and beyond. IEEE
Transactions on Industrial Informatics, 16(11), 7094–7104.
8. Mafakheri, B., Subramanya, T., Goratti, L., & Riggio, R. (2018).
Blockchain-based infrastructure sharing in 5G small cell net-
works. In 2018 14th international conference on network and
service management (CNSM), IEEE, 313–317.
Wireless Networks
123

9. Serrano, W. (2021). The blockchain random neural network for
cybersecure IoT and 5G infrastructure in smart cities. Journal of
Network and Computer Applications, 175, 102909.
10. Kausar, F., Sadiq, M. A. K., & Asif, H. M. (2021). Convergence
of blockchain in IoT applications for heterogeneous networks.
Real-time intelligence for heterogeneous networks (pp. 71–86).
Springer.
11. Dai, Y., Xu, D., Maharjan, S., Chen, Z., He, Q., & Zhang, Y.
(2019). Blockchain and deep reinforcement learning empowered
intelligent 5G beyond. IEEE Network, 33(3), 10–17.
12. Feng, C., Liu, B., Guo, Z., Yu, K., Qin, Z., & Choo, K.-K.R.
(2021). Blockchain-based cross-domain authentication for intel-
ligent 5g-enabled internet of drones. IEEE Internet of Things
Journal, 9, 6224–6238.
13. Adat, V., Politis, I., Tselios, C., Galiotos, P., & Kotsopoulos, S.
(2018). On blockchain enhanced secure network coding for 5G
deployments. In 2018 IEEE global communications conference
(GLOBECOM), 1–7.
14. Hojjati, M., Shafieinejad, A., & Yanikomeroglu, H. (2020). A
blockchain-based authentication and key agreement (AKA) pro-
tocol for 5G networks. IEEE Access, 8, 216461–216476.
15. Abdullah, R. M., Abualkishik, A. Z., & Alwan, A. A. (2018).
Improved handover decision algorithm using multiple criteria.
Procedia Computer Science, 141, 32–39.
16. Sangeetha, D., Selvi, S., & Keerthana, A. (2022). A trust-based
handover authentication in an SDN 5G heterogeneous network.
Computer networks and inventive communication technologies
(pp. 841–852). Springer.
17. Divakaran, J., Prashanth, S. K., Mohammad, G. B., Shitharth, D.,
Mohanty, S. N., Arvind, C., Srihari, K., Abdullah, R. Y., &
Sundramurthy, V. P. (2022). Improved handover authentication
in fifth-generation communication networks using fuzzy evolu-
tionary optimization with nanocore elements in mobile healthcare
applications. Journal of Healthcare Engineering,2022.
18. Chow, M. C., & Ma, M. (2021). A blockchain-enabled 5G
authentication scheme against DoS attacks. Journal of Physics:
Conference Series, IOP Publishing, 1812(1), 012030.
19. Ren, Z., Li, X., Jiang, Q., Cheng, Q., & Ma, J. (2021). Fast and
universal inter-slice handover authentication with privacy pro-
tection in 5G network. Security and Communication Networks,
2021, 1–19.
20. Jia, X., Hu, N., Yin, S., Zhao, Y., Zhang, C., & Cheng, X. (2020).
A2 chain: A blockchain-based decentralized authentication
scheme for 5G-enabled IoT. Mobile Information Systems, 2020,
1–19.
21. Haddad, Z., Fouda, M. M., Mahmoud, M., & Abdallah, M.
(2020). Blockchain-based authentication for 5G networks. In
2020 IEEE International Conference on Informatics, IoT, and
Enabling Technologies (ICIoT), 189–194.
22. Yazdinejad, A., Parizi, R. M., Dehghantanha, A., & Choo, K.-
K.R. (2019). Blockchain-enabled authentication handover with
efficient privacy protection in SDN-based 5G networks. IEEE
Transactions on Network Science and Engineering., 8,
1120–1132.
23. Gharsallah, A., Zarai, F., & Neji, M. (2019). SDN/NFV-based
handover management approach for ultradense 5G mobile net-
works. International Journal of Communication Systems, 32(17),
e3831.
24. Yang, J., Ji, X., Huang, K., Chen, Y., Xu, X., & Yi, M. (2019).
Unified and fast handover authentication based on link signatures
in 5G SDN-based HetNet. IET Communications, 13(2), 144–152.
25. Chen, Z., Chen, S., Xu, H., & Hu, B. (2018). A security
authentication scheme of 5G ultra-dense network based on block
chain. IEEE Access, 6, 55372–55379.
26. Nabeel, M., Manalastas, M., Ijaz, A., Refai, H., & Imran, A.
(2022). Investigating handover behavior with 5G and beyond
TurboRAN Testbed. In 2022 seventh international conference on
mobile and secure services (MobiSecServ), IEEE, 1–6.
27. Goudarzi, S., Anisi, M. H., Ciuonzo, D., Soleymani, S. A., &
Pescape, A. (2020). Employing unmanned aerial vehicles for
improving handoff using cooperative game theory. IEEE Trans-
actions on Aerospace and Electronic Systems, 57(2), 776–794.
28. Chen, P., Han, D., Weng, T.-H., Li, K.-C., & Castiglione, A.
(2021). A novel Byzantine fault tolerance consensus for Green
IoT with intelligence based on reinforcement. Journal of Infor-
mation Security and Applications, Elsevier, 59, 102821.
29. Yan, X., & Ma, M. (2021). A lightweight and secure handover
authentication scheme for 5G network using neighbour base
stations. Journal of Network and Computer Applications, 193,
103204.
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds
exclusive rights to this article under a publishing agreement with the
author(s) or other rightsholder(s); author self-archiving of the
accepted manuscript version of this article is solely governed by the
terms of such publishing agreement and applicable law.
Shivanand V. Manjaragi is
working as an Assistant Pro-
fessor in Department of Com-
puter Science and Engineering
at Hirasugar Institute of Tech-
nology, Nidasoshi, Belagavi,
India and he is currently pursu-
ing Ph.D. studies at Basa-
veshwar Engineering College,
Bagalkote Research Centre,
Visvesvaraya Technological
University, Belagavi, India in
the Department of Computer
Science and Engineering. He
received his Master and Bache-
lor of Engineering degrees from the Visvesvaraya Technological
University, Belagavi, Karnataka, India in 2011 and 2002, respec-
tively. He has published papers in journals, International, and
National conferences. His main research interests include Computer
Networks, Wireless Networks, Machine Learning, Deep Learning,
and Block Chain. He is a member of the CSI and ISTE association.
S. V. Saboji is working as a
Professor in the Department of
Computer Science and Engi-
neering at Basaveshwar Engi-
neering College, Bagalkote,
India. He pursued his B.E. in
Computer Science and Engi-
neering from Karnataka Uni-
versity Dharwad in 1997,
M.Tech, and Ph.D. from Vis-
vesvaraya Technological Uni-
versity, Belagavi in 2005, and
2013 respectively. His research
areas include Wireless Mobile
Networks, Wireless Adhoc
Networks, Computer Networks. He has published papers in journals,
International, and National conferences. He is the reviewer for
international journals and conferences. He is a Senior Member of the
CSI and IEEE association.
Wireless Networks
123