Content uploaded by Muhammet Çakmak
Author content
All content in this area was uploaded by Muhammet Çakmak on Mar 27, 2024
Content may be subject to copyright.
Research Article
Journal of Intelligient Systems: 7(1) (2024) 1-13
DOI: 10.38016/jista.1225716
_________________________________
* Corresponding Author. Recieved : 28 Dec 2022
E-mail: mcakmak@sinop.edu.tr Revision : 14 Jun 2023
Accepted : 28 Sep 2023
The Impact of Denial-of-Service Attacks and Queue
Management Algorithms on Cellular Networks
Muhammet Çakmak1*
1Sinop University, Department of Computer Engineering, Sinop, Türkiye
mcakmak@sinop.edu.tr
Abstract
In today's digital landscape, Distributed Denial of Service (DDoS) attacks stand out as a formidable threat to organisations all over the
world. As known technology gradually advances and the proliferation of mobile devices, cellular network operators face pressure to
fortify their infrastructure against these risks. DDoS incursions into Cellular Long-Term Evolution (LTE) networks can wreak havoc,
elevate packet loss, and suboptimal network performance. Managing the surges in traffic that afflict LTE networks is of paramount
importance. Queue management algorithms emerge as a viable solution to wrest control over congestion at the Radio Link Control
(RLC) layer within LTE networks. These algorithms work proactively, anticipating, and mitigating congestion by curtailing data
transfer rates and fortifying defences against potential DDoS onslaughts. In the paper, we delve into a range of queue management
methods Drop-Tail, Random Early Detection (RED), Controlled Delay (CoDel), Proportional Integral Controller Enhanced (PIE), and
Packet Limited First In, First Out queue (pFIFO). Our rigorous evaluation of these queue management algorithms hinges on a
multifaceted assessment that encompasses vital performance parameters. We gauge the LTE network's resilience against DDoS
incursions, measuring performance based on end-to-end delay, throughput, packet delivery rate (PDF), and fairness index values. The
crucible for this evaluation is none other than the NS3 simulator, a trusted platform for testing and analysis. The outcomes of our
simulations provide illuminating insights. CoDel, RED, PIE, pFIFO, and Drop-Tail algorithms emerge as top performers in succession.
These findings underscore the critical role of advanced queue management algorithms in fortifying LTE networks against DDoS
attacks, offering robust defences and resilient network performance.
Keywords: DDoS attacks, LTE network, Ns-3 simulation
DDoS Saldırılarının ve Kuyruk Yönetimi Algoritmalarının Hücresel Ağlar
Üzerindeki Etkisi
Öz
Günümüzün dijital ortamında Dağıtılmış Hizmet Reddi (DDoS) saldırıları, dünyanın her yerindeki kuruluşlar için büyük bir tehdit
olarak öne çıkıyor. Bilinen teknolojinin giderek ilerlemesi ve mobil cihazların yaygınlaşmasıyla hücresel şebeke operatörleri,
altyapılarını bu risklere karşı güçlendirme baskısıyla karşı karşıya kalıyor. Hücresel Uzun Vadeli Evrim (LTE) ağlarına yapılan DDoS
saldırıları büyük hasara, yüksek paket kaybına ve yetersiz ağ performansına yol açabilir. LTE ağlarını etkileyen trafikteki
dalgalanmaları yönetmek büyük önem taşıyor. Kuyruk yönetimi algoritmaları, LTE ağları içindeki Radyo Bağlantı Kontrolü (RLC)
katmanındaki tıkanıklığın kontrolünü ele geçirmek için geçerli bir çözüm olarak ortaya çıkıyor. Bu algoritmalar proaktif olarak çalışır,
veri aktarım hızlarını azaltarak ve potansiyel DDoS saldırılarına karşı savunmayı güçlendirerek tıkanıklığı öngörür ve azaltır. Bu
yazıda, Drop-Tail, Random Early Detection (RED), Controlled Delay (CoDel), Proportional Integral Controller Enhanced (PIE) ve
Packet Limited First In, First Out queue (pFIFO) gibi çeşitli kuyruk yönetimi yöntemlerini derinlemesine inceliyoruz. Bu kuyruk
yönetimi algoritmalarına yönelik titiz değerlendirmemiz, hayati performans parametrelerini kapsayan çok yönlü bir değerlendirmeye
dayanır. LTE ağının DDoS saldırılarına karşı dayanıklılığını ölçüyoruz; performansı uçtan uca gecikmeye, üretime, paket dağıtım
hızına (PDF) ve adalet endeksi değerlerine göre ölçüyoruz. Bu değerlendirmenin potası, test ve analiz için güvenilir bir platform olan
NS3 simülatöründen başkası değildir. Simülasyonlarımızın sonuçları aydınlatıcı bilgiler sağlıyor. CoDel, RED, PIE, pFIFO ve Drop-
Tail algoritmaları art arda en iyi performans gösterenler olarak ortaya çıkıyor. Bu bulgular, gelişmiş kuyruk yönetimi algoritmalarının,
LTE ağlarını DDoS saldırılarına karşı güçlendirme, sağlam savunmalar ve esnek ağ performansı sunma konusundaki kritik rolünün
önemini göstermektedir.
Anahtar kelimeler: DDoS atakları, LTE ağı, Ns-3 simülasyonu
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 2
1. Introduction
The rapidly developing technology has increased the
number of mobile phones used in cellular networks.
Cellular network operators invest more in research and
technology to secure against increasing data and
network traffic. It is an important problem for attackers
to cause network disruptions and block user services
(Zenitani, 2023). DDoS attacks that occur in the LTE
network cause mobile network users to disconnect or
decrease the Quality of Service (QoS) (Feng et al.,
2020).
DDoS attacks, which disrupt services on the
network, are challenging to detect. DDoS attacks that
occur in the LTE network affect all network layers but
significantly affect the Radio Link Control (RLC) layer
is a protocol layer in the LTE (Long-Term Evolution)
wireless communication standard. It is responsible for
ensuring the reliable transmission of data between the
source and destination over the wireless link (Çakmak et
al., 2021). One of the LTE cellular networks' main aims
is developing queue management algorithms for RLC.
Queue management method algorithms are used to
ensure resource allocation and quality of service (QoS)
during network congestion (Çakmak et al., 2022).
Active Queue Management (AQM) algorithms enable
efficient use of services such as delay, high bandwidth,
and packet delivery speed. AQM mechanisms are used
in LTE cellular networks under DDoS attacks(Wang and
Wang, 2020). As a result, complete disconnection,
connection slowdowns, disruptions of service, and
potential data loss issues that could occur on the LTE
network are prevented.
A DDoS attack occurring on a cellular LTE network
can cause the entire network to go down, slowing it or
preventing the use of network bandwidth. Choosing a
practical algorithm for the security of the LTE network
and the continuation of its services is crucial. Although
DDoS attacks have been tested on the cellular LTE
network to date, the performance of these attacks on
queue management algorithms has not been assessed.
Unlike other studies, this study evaluated the
performance of queue management algorithms against
DDoS attacks in the RLC layer of the cellular LTE
network. Well-known Drop-Tail, RED, CoDel, PIE,
pFIFO algorithms were compared under DDoS attacks
in LTE networks according to end-to-end delay,
throughput, PDF and fairness index values. PDF is a
metric that measures the percentage of packets
successfully delivered to their intended destination in a
network.
DDoS attacks on LTE networks cause the network to
be temporarily or completely disabled. It also causes
disruption of services. Queue management algorithms
create a preliminary protection area to prevent attacks
and reduce their impact. It prevents congestion in queue
buffers, controls packet drops, and regulates excessive
demands.
Although loss-based, delay-based, rate based, queue-
based, topology based, machine-learning based and
hybrid-based studies have been carried out in LTE
networks to date, the performance of DDoS attacks with
queue management algorithms in LTE networks has not
been examined. In this study, unlike others, the
performances of the queue management algorithms in
the LTE RLC layer under DDoS attacks were compared
according to end-to-end delay, throughput, PDF and
fairness index values using the NS-3 network simulator.
Unlike others in these studies:
• The impact of queue management algorithms on the
cellular LTE network under DDoS attacks was
examined.
• Comparative performance of queue management
algorithms against DDoS attacks in the RLC layer
of the LTE network is shown.
• Well-known Drop-Tail, RED, CoDel, PIE, pFIFO
algorithms were compared according to end-to-end
delay, throughput, PDF and fairness index values
under DDoS attacks in LTE networks.
The remainder of the study is organised as follows.
In the second chapter, background and literature studies
were examined. In the third section, an experimental
framework and performance evaluation are provided.
The results are given in the fourth section.
2. Background and Related Works
2.1. LTE Network Architecture
LTE, which stands for Long-Term Evolution, is a
cellular communication standard that provides high
mobile voice traffic and short messaging services (Wu
et al., 2022). The LTE core network is divided into two
main sections, as indicated in Figure 1, the E-UTRAN
(Evolved Universal Terrestrial Access Network) and the
EPC (Evolved Packet Core). The E-UTRAN disposes of
an eNB (base station), which serves as the gateway
between mobile terminals, radio antennas, and operators
(Israr et al., 2021). eNBs are a base station that controls
cell phones in the cell. It is defined as the eNB serving
when the base station communicates with a mobile
phone. eNBs communicate with mobile phones using
analogue and digital signals via the air interface. eNB
controls the operation of all mobile phones by sending
signal messages (Zidic et al., 2023). eNBs are connected
to the EPC via the S1 interface. The eNB accesses
nearby base stations using the X2 interface for signalling
and packet forwarding during transmission. EPC
consists of three functional modules: Mobility
Management Entity (MME), Service Gateway (S-GW),
and Packet Data Network Gateway (PDN-GW). The
MME delivers a paging message to the base station to
provide service in the Core System (CS) domain. The S-
GW connects LTE nodes and transmits user data
packets. S-GW works like a router. In addition, S-GW
sends the obtained data between the base station and the
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 3
PDN-GW (Mousavi et al., 2017). The Home Subscriber
Server (HSS) is a central database that stores
information about the cellular network operator's
subscribers. The Policy Control and Pricing Rules
Function (PCRF) decides on policy control. It also
controls flow-based charging functions. The LTE
network is connected to the rest of the Internet via PDN-
GW (Oughton et al., 2022). The PDN-GW connects the
EPC unit to external IP networks. PDN-GW forwards
packets to external IP networks. It also allocates IP
addresses to all users. It handles IP user traffic-related
operations, including packet filtering.
LTE network architecture consists of Physical Layer
(Layer 1), Medium Access Layer (MAC), Radio Link
Control (RLC), Radio Resource Control (RRC), Packet
Data Convergence Control (PDCP), and Non-Access
Stratum (NAS) Protocols(Mousavi et al., 2019).
Figure 1. LTE Architecture (Gómez et al., 2014)
Figure 2 shows the data flow in the RLC, PDCP, and
MAC layers.
Figure 2. LTE MAC, RLC and PDCP Layers (Gómez et
al., 2014)
The RLC layer is used for segmentation, efficient
transport, and sequential transmission. RLC analyses the
initial data units containing information of a particular
type and ensures the security of the target data
transmission. More control or security is needed at the
RLC layer. Because the RLC layer is used in all
transmission of data packets such as voice, video, and
FTP, when a DDoS attack or other attack reaches the
RLC level, it can cause packet loss or slow throughput,
long packet delays, traffic congestion, and heavy and
inefficient network performance.
2.2. DDoS Attacks
A DDoS attack is a malicious attempt to disrupt the
regular traffic of the target server, service, or network by
crushing it with a stream of internet traffic. DDoS
attacks provide activity using multiple computer
systems. These attacks can use resources such as
computers, mobile phones, and IoT devices to execute
an attack on the target. During a high-level DDoS attack,
the victim network experiences unexpected traffic
congestion, preventing regular traffic from reaching its
destination. A standard method in a DDoS attack is for
the attacker to send a stream of packets to a victim. This
transmitted stream consumes significant resources,
preventing the real user from accessing the resources
(Ali et al., 2022; Said et al., 2022). Another method is to
send malformed packets, which disable or blocks the
application or protocol on the target machine. Some
attacks block services by overloading the Internet
infrastructure instead of targeting victims. A DDoS
attack is also an effective attack technique to consume
resources so that legitimate clients cannot receive
internal or external services.
2.3. Queue Management Algorithms
Queue Management algorithms have been
developed to detect and protect the network from this
congestion. Queue management algorithms prevent
packet loss in the network, reduce delays, detect
network congestion, and recoverm congestion (Çakmak
and Albayrak, 2018 ; Akhter et al., 2021). Passive queue
management algorithms such as drop-tail only forward
queued packets sequentially, while active queue
management algorithms can pre-identify congestion and
minimise packet transfer rate. This study evaluated
several queue management algorithms, including well-
known options like Drop-Tail, RED, CoDel, PIE, and
pFIFO. Our analysis focused on their effectiveness in
mitigating DDoS attacks on LTE networks.
Drop-Tail is simply the most straightforward queue
management algorithm. It is often used in routers due to
its simple control mechanism. Drop-Tail provides
priority forwarding of the first packets to the queue. New
packets are automatically rejected if the queue is full.
The arrival speed of the packets can be greater than the
output speed. Drop-Tail, the priority levels of all packets
are the same and this causes traffic congestion and
bottleneck.
RED is the first AQM algorithm for congestion. It
calculates network congestion using the average queue
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 4
size to avoid congestion in RED packet switched
networks. The RED algorithm manages the queue size
using four parameters: the minimum threshold (Minth),
the maximum threshold (Maxth), the maximum
probability, and the weighting factor. The minimum and
maximum points determine the queue sizes within
which packets are marked. In contrast, the maximum
likelihood sets the maximum drop probability for
packets exceeding the maximum threshold. The
weighting factor controls the rate at which the average
queue size is computed over time, with higher values
giving greater weight to recent samples. The algorithm
works to maintain an average queue size. The packet
drop probability value, Pd, varies linearly between zero
and Pmax as the mean queue size changes between the
Minth and Maxth values. All incoming packets are
dropped if the average queue size value exceeds Maxth
(Paul et al., 2017). The RED algorithm controls the
congestion structure due to packet drop because the
packet drop mechanism operates according to the
moving average of the past values.
CoDel was developed to solve the bottleneck
problem in the network (Raghuvanshi et al., 2013).
Congestion is detected by CoDel when the packet
transmission time exceeds the set target value. After the
congestion is detected, the signal is sent to drop the
packet to avoid clutter in the queue. CoDel detects
congestion in the network using the packet delay time.
Buffer bloat causes packet losses to increase in the
queue even when the buffer size is large. The CoDel
algorithm efficiently controls the buffer-bloat problem.
Unlike RED, CoDel is independent of parameters such
as queue size and average, queue delay, and drop rate.
pFIFO is an active queue management algorithm
based on the FIFO algorithm (Bisoy and Pattnaik, 2016).
It is classified by considering different channels for
network traffic. High-priority traffic is processed earlier.
Its main purpose is a simple method to support
differentiated service classes. The advantage of pFIFO
is low computational load and traffic transmission
generated in real-time applications. The most critical
problem in the network is that the volume of high-
priority traffic is high, the buffer space allocated for low-
priority traffic is reduced, and overflow occurs. This
causes packet drop on the network and slows down the
network (Low et al., 2002).
PIE is designed to control latency in the network
effectively (Pan et al., 2013). In PIE, the average queue
rate is estimated relative to the non-moving queue. The
speed is used to calculate the available delay. Then, the
delay is periodically used to calculate the probability of
falling. Finally, when the packet arrives at the
destination, a packet is dropped (or flagged) based on
that probability. PIE adjusts probability based on latency
trend. Alpha and beta are statically selected parameters
chosen to control the fall probability increase and are
determined by control theoretical approaches. Alpha
determines how the deviation between the current and
target delay changes the probability. The beta makes
additional adjustments based on the lag trend. PIE is
designed to improve time-sensitive performance and
strives to provide interactive traffic and network
stability while maintaining high connection usage.
Adapting the control parameters in small increments
voids large oscillations leading to unbalance(Pan et al.,
2013).
2.4. Related Works
Researchers and developers have proposed various
queue management algorithms to address congestion
and performance issues in wired and wireless networks.
These algorithms aim to improve network performance
by managing the flow of packets in the network queues.
These algorithms aim to improve features like network
utilisation, packet loss, and adaptability for different
traffic loads.
The proposed AQMs aim to solve problems such as
loss-based congestion control (Verma et al., 2022)
delay-based congestion control and rate-based
congestion control (Amer et al., 2020). For eNB on LTE,
smRED (smart-RED), a smart AQM that works by
adjusting the variance value in RED to prevent buffer
congestion and packet drop in variable traffic loads at
the RLC layer, enables packet programmers to work
through single-cell and multicell handover and no
failover. The study investigated the effects of the queue
management algorithms on the network performance
metrics, including throughput, delay, and deviation
values. The adjusted variance value i of RED took
different values in low and high-load situations,
affecting the drop of packets accumulated in RCL
buffers (Paul et al., 2017).
The authors introduced an innovative machine
learning-based intrusion detection system to mitigate
Distributed Denial of Service (DDoS) attacks within the
LTE-A network. The proposed model effectively
identifies DDoS flows targeting the eNB and utilises the
random forest algorithm for attack classification.
Remarkably, the system achieved an impressive
accuracy rate of 99.95% in accurately identifying and
classifying DDoS attacks (Gong et al., 2019).
In a separate study, the authors introduced the DDoS
Threat Analysis and Response Framework (DTARS),
which encompasses distributed real-time threat
identification, behavioural monitoring, and validation of
control plane operations specifically tailored for LTE
networks (Krishnan et al., 2019).
In another study, the authors proposed an improved
Neural Network (NN) model for intrusion detection,
which identifies authenticated nodes and cluster heads
(CH) using elliptic curve cryptography (ECC) in the
LTE network for Machine Type Communication (MTC)
devices. In case of an attack, the proposed model
imposes a penalty and restricts the affected nodes from
participating in MTC communication. Specifically, the
improved NN was created using a new optimisation
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 5
algorithm called Whale with Three-Level Update
(WTU) (Jyothi and Chaudhari, 2020).
The packet congestion feedback mechanism
proposed by the authors was developed to avoid queue
overflow in the LTE networks and reduce queuing
delays. With the QoS value of the packets in the
network, the packet is defined as TCP or UDP. An
average value was calculated to estimate full queues,
and low ranges for TCP and high ranges were
determined for UDP. Setting the sender congestion
window to high, medium and low according to the
network traffic situationit has been tried to prevent the
queue overflow in Evolved Node B (eNB) (Adesh and
Renuka, 2019).
In another study, the authors classified DDoS attacks
in the LTE network with logistic regression and decision
tree machine learning algorithms (Ashfaq et al., 2022).
The authors examined the performance of AQMs
operating in the RLC layer of the cellular LTE network
according to end-to-end delay, throughput, PDF and
fairness index values. The study has shown that the
selection of the right queue management algorithm
directly affects the performance of the network (Çakmak
et al., 2021; Çakmak and Albayrak, 2022). The authors
proposed an adaptive queue management algorithm
operating at the RLC layer of the cellular LTE network.
The proposed algorithm adaptively adjusts the network's
throughput according to low, medium and high network
load (Çakmak and Albayrak, 2022).
Praveen and Pratap (2021) proposed a congestion-
sensitive resource allocation and routing protocol (CRR)
based on hybrid optimisation techniques for IoT devices
in smart city infrastructure. The proposed method uses a
meta-heuristic algorithm to reduce total congestion and
allocate IoT gateways. It also uses a swarm optimisation
algorithm for the route discovery mechanism. The
proposed ECRR technique was designed and
implemented with the NS-2 simulator. The authors
examined the AQM mechanism based on neural
networks. The study aims to employ machine learning
techniques to learn the behaviour of the Active Queue
Management (AQM) mechanism. Specifically, the
study seeks to develop a model that can predict the
behaviour of the AQM mechanism in response to
changes in network traffic conditions, such as variations
in traffic volume or type, to improve network
performance and reduce congestion. Obtains training
examples considering the similarity of network traffic.
The model uses the Gaussian method. The study
demonstrates the effectiveness of the active queue
management mechanism based on neural networks
(Szyguła et al., 2021). In another study, the authors
proposed the Federated Intelligence (FIAQM)
architecture for AQM, using the Federated Learning
approach. The proposed method dynamically adjusts the
AQM parameters for a multi-domain environment,
which is difficult to achieve with conventional AQM.
The proposed FIAQM method uses a trained
feedforward neural network trained on a network traffic
dataset to predict the behavior of the Active Queue
Management (AQM) mechanism in response to changes
in traffic conditions. In addition, the developed method
improves the performance of FIAQM's inter-domain
connections by reducing congestion in its connections
while keeping the network data private in each
participating domain (Gomez et al., 2021).
The authors propose an algorithm for fair bandwidth
distribution, considering traffic class priority, connected
loads of a node, and average queue size. The average
queue size across the nodes is adjusted based on the
gradient suggested for the Global Priority (GP)
differential, which is a metric that assigns priorities to
different types of network traffic based on their
importance. Specifically, the proposed mechanism uses
the GP differential to determine the relative importance
of different traffic flows in the network and adjusts the
average queue size accordingly. This allows for more
efficient allocation of network resources and improved
Quality of Service (QoS) for critical traffic flows. The
node's speed is calculated based on the node's average
queue size and GP. The proposed routing protocol is
designed for Wireless Sensor Networks (WSN),
networks of small, low-power devices equipped with
sensors that collect and transmit data wirelessly. The
protocol is optimised for WSNs, which typically have
limited resources such as bandwidth, energy, and
computational power and operate in environments
where the network topology may change frequently. The
proposed method is applied to a tree topology that deals
with both Real-Time (RT) and Non-real-time (NRT)
traffic classes (Swain and Nanda, 2021). Table 1 shows
control mechanism literature for LTE networks.
Table 1. Control mechanism literature for LTE networks
Study
Year
Reference
Loss-based congestion control
2022
(Verma et al., 2022)
Delay-based congestion control
2021
(Lin et al., 2021)
Rate-based congestion control
2020
(Amer et al.)
2019
(Paul et al.)
Topology-based mechanism
2021
(Swain and Nanda)
Machine-learning based control
2019
(Gong et al.)
2020
(Jyothi and
Chaudhari)
2021
(Szyguła et al.)
2021
(Gomez et al.)
Hybrid control mechanism
2021
(Praveen and Pratap)
2022
(Çakmak and
Albayrak)
Although loss-based, delay-based, rate-based,
queue-based, topology-based, machine-learning-based,
and hybrid-based studies have been carried out in LTE
networks, the performance of DDoS attacks with queue
management algorithms in LTE networks has not been
examined. A DDoS attack on a cellular LTE network
can cause a temporary disruption and potentially
degrade the network's performance, as it overwhelms
the network with many requests, making it difficult for
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 6
legitimate users to access network resources. However,
it does not permanently turn off the entire network.
Although security studies have been done in the cellular
LTE network, the security of the RCL layer has not been
directly tested. Preventing DDoS attacks that may occur
at the RCL layer will increase the performance of
cellular network security. In this study, unlike others, the
versions of the queue management algorithms in the
LTE RLC layer under DDoS attacks were compared
according to end-to-end delay, throughput, PDF and
fairness index values using the NS-3 network simulator.
3. Experimental Framework and
Performance Assessment
In this section, the network topology and system
parameters for the LTE network are determined. Then
the simulation results were analysed and evaluated
according to the end-to-end average throughput, delay,
PDF and fairness index.
For the experimental environment,
* Firstly, eNB and nodes are created in the NS3
environment.
*Parameter of eNB is set.
*BotNets are placed according to the specified
numbers in accordance with the topology.
*Simulation starts
*Drop-tail, RED, CoDel, pFIFO and PIE algorithms
are run separately at simulation time.
*Test results are saved as end-to-end average
throughput, delay, PDF and fairness index
*Simulation is terminated.
3.1. Network Topology and Simulation
Environment
One of the easiest and most effective ways to
empirically observe traffic on the network is to use
simulation. Using simulation, network nodes,
connections and network traffic can be designed
similarly to the real world (Jevtić et al., 2009) and
(Weingärtner, et al., 2009). In addition, simulation
makes it cheaper and easier to implement systems that
are difficult and very expensive to implement in the real
world. Ns3 network simulator is free open-source
software developed for educational and research
purposes and works on a discrete event basis. Parameter
settings for the Ns3 simulation environment were made
as follows:
*Basic eNB added
*MTU value for SGW and PDN-GW is set to 1500
bytes
*Inter-unit data rate was selected as 100 Gbps, which
is the maximum supported by eNB
* eNB delay set to optimum value 0.01s
* eNB connection latency set to 2ms.
*TCP New Reno module was used as TCP traffic
type
*Attackers' displacement feature was selected as
RandomWalk2D
*For Droptail's package, a maximum of 50 packages,
100 ms interval and 5ms target value were selected
*For RED, the minth value for sending 50 packets of
data is determined as 20 and the maximum value is 50.
Queue weight value was determined as 0.002s
*Limit package setting for pFIFO was set as 50
packages
*Average queue delay value for PIE was set to 0.01s
*Simulation completed between 0.1s and 100s.
This study designed an LTE network structure and
attack scenario suitable for the actual situation. Thus, the
obtained data is more similar to the actual system. In the
experimental environment, 1 eNB is attacked by DDoS
with 10, 20, 40, 60, and 100 attackers, respectively. In
simulation, data traffic starts from 0.1 seconds and the
simulation takes 100 seconds. Drop-tail, RED, CoDel,
pFIFO and PIE algorithms were run in each attack. The
results were compared to the average end-to-end
throughput, delay, PDF and fairness index. The
simulation parameters showed in Table 2.
3.2. Experimental Results
The simulation results were analysed using average
end-to-end delay, throughput, PDF, and fairness index
values. For the real-world simulation environment, a
DDoS attack is carried out on 1 eNB by 10, 20, 40, 60
and 100 attackers, respectively.
Table 2. Simulation Parameters
SGW and PDN-GW Gateway
LTE
Parameters
Value
MTU (Maximum
Transmission Unit)
1500 Bytes
Data Rate
100 Gbps
Delay
0.010 s
Wired Link Capacity
100 Mbps
Data Rate
100 Gbps
Delay
2 ms
Botnet Number
10, 20, 40, 60, and
100
TCP Traffic Type
TCP New Reno
Mobility
RandomWalk2D
Queue Algorithms
Drop-tail
Mode (Bytes,
Packets)
Packets
Max Packets
50
Codel
Mode
Packets
Max Packets
50
Interval
100 ms
Target
5 ms
RED
Mode
Packets
MeanPktSize
50
IdlePktSize
1500*1000bytes
MinTh, MaxTh
20,50
Queue Limit
50
Queue weight
0,002s
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 7
Link Delay
20ms
pFIFO
Limit
50 packets
PIE
Average Queue Delay
0.01 s
Simulation Time
Simulation Start Time
0.1 s
Total Simulation Time
100 s
Figure 3 shows the packet control flow structure of the
Drop-tail RED, CoDeL, PIE and pFIFO algorithms for
the simulation environment.
Figure 3. Simulation Packet Controller
Figure 4 shows the network topology of the simulation.
Figure 4. Simulation network topology
Pseudo code of the system.
1: procedure enodeb (parameters)
2: Initialize variable
3: Arrive packets on RLC
5: Attack on system
4: (Droptail, RED, CoDel, PIE, and pFIFO) apply on RLC
5: Check the packets
6: End the process
3.2.1 Average End-to-end Throughput
Drop-tail, RED, CoDel, pFIFO and PIE algorithms
were compared using 10, 20, 40, 60 and 100 botnets in
the LTE network environment. Table 3 shows the
average throughput values of the LTE RLC layer under
DDoS attacks. The variation of the end-to-end average
throughput values according to the number of botnets is
shown in Figure 5.
Table 3. End-to-end Throughput
AQM
Average Throughput (Kbps)
10 BotNets
20 BotNets
40 BotNets
60 BotNets
100 BotNets
Drop-Tail
2492,157
1593,321
1058,988
310,52
123,963
RED
2793,681
1953,759
1473,846
643,13
317,061
CoDel
2914,524
2047,491
1623,633
730,35
365,868
pFIFO
2554,671
1777,167
1386,537
566,55
217,539
PIE
2756,93
1866,50
1462,62
571,65
275,244
The average end-to-end output throughput values
obtained with the increase in the number of botnets
showed a natural decrease. CoDel algorithm showed the
best performance in terms of end-to-end throughput
value even in high botnet attack. Other best performing
algorithms are listed as RED, PIE, pFIFO, and Drop-
Tail respectively. The CoDel algorithm gives the best
results due to early detection of packet drops and
processing according to the queuing time of packets.
Drop-Tail algorithm, which dropped incoming packets
after the queue is full, gives the worst result among all
algorithms. This shows that Drop-tail is the most
vulnerable algorithm for high traffic.
The presented data outlines the Average Throughput
(measured in Kbps) achieved by various AQM (Active
Queue Management) algorithms across different
scenarios involving varying numbers of BotNets. When
confronted with 10 BotNets, the CoDel algorithm
exhibits the highest throughput at 2914.524 Kbps,
closely followed by the PIE algorithm at 2756.93 Kbps.
The RED algorithm also demonstrates commendable
performance, achieving a throughput of 2793.681 Kbps.
In contrast, both Drop-Tail and pFIFO algorithms yield
lower throughputs at 2492.157 Kbps and 2554.671 Kbps
respectively. As the number of BotNets increases to 20,
40, 60, and eventually 100, a discernible pattern
emerges. CoDel consistently maintains its lead in
throughput across all scenarios, showcasing its superior
ability to handle network congestion, even in high-stress
situations with a substantial number of BotNets. RED
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 8
and PIE algorithms also exhibit competitive
performance, with both consistently achieving
noteworthy throughput values. Conversely, Drop-Tail
and pFIFO face considerable challenges, with their
respective throughputs experiencing significant drops as
the number of BotNets increases. This comparative
analysis underscores the efficacy of CoDel in managing
network congestion and maximising throughput,
particularly in scenarios with a high BotNet count.
Figure 5. Average End-to-end Throughput Under DDoS Attacks
3.2.2 Average End-to-end Delay
Drop-tail, RED, CoDel, pFIFO and PIE were
compared in an LTE network environment using 10, 20,
40, 60 and 100 botnets. Table 4 shows the average
throughput values of the LTE RLC layer under DDoS
attacks.
Table 4. End-to-end Delay
AQM
Average Delay (ms)
10 BotNets
20 BotNets
40 BotNets
60 BotNets
100 BotNets
Drop-Tail
74,686
160,86
343,04
648,45
1809,8
RED
29,193
117,40
246,27
472,62
949,679
CoDel
26,751
104,43
208,85
395,81
766,104
pFIFO
44,6405
131,68
280,81
514,69
1050,6
PIE
41,59
126,50
268,73
488,34
988,104
Figure 6 shows the aveage end-to-end delay based
on the number of botnets. Packets coming from the
PDN-GW router significantly affect parameters such as
modulation, encoding, and packet creation time. These
parameters are adversely affected in a DDoS attack.
CoDel does not use the size of the queue to manage the
queue, it uses the queuing time of the packets. This
ensures that the packet drop value of the CoDel
algorithm is low. Thus, the end-to-end delay value of
CoDel is lower than other algorithms. Dropping more
packets due to a DDoS attack caused more delay. The
CoDel algorithm has the best end-to-end delay
performance, with the lowest packet loss rate under
DDoS attack. The Drop-Tail drops the packets arriving
to the queue from the front if the queue is full. Therefore,
Drop-tail has the worst performance among all
algorithms. This algorithm is followed by RED, PIE,
pFIFO and Drop-Tail, respectively. The CoDel
algorithm is followed by RED, PIE, pFIFO and Drop-
Tail, respectively.
Average Delay (measured in milliseconds)
experienced with various AQM (Active Queue
Management) algorithms across different scenarios
involving varying numbers of BotNets. With 10
BotNets, the CoDel algorithm demonstrates the lowest
delay at 26.751 ms, closely followed by the RED
algorithm at 29.193 ms. In contrast, Drop-Tail
experiences significantly higher delays at 74.686 ms. As
the number of BotNets increases to 20, 40, 60, and
eventually 100, a clear trend emerges. CoDel
consistently maintains its lead in minimising delay
across all scenarios, showcasing its superior ability to
manage network congestion, even in high-stress
situations with a substantial number of BotNets. RED
and PIE algorithms also exhibit competitive
performance, with both consistently achieving lower
delay values. Conversely, Drop-Tail and pFIFO face
0
500
1000
1500
2000
2500
3000
3500
10 BotNets 20 BotNets 40 BotNets 60 BotNets 100 BotNets
End-to-end Average Throughput (Kbps)
Drop-Tail RED CoDel pFIFO PIE
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 9
considerable challenges, with their respective delays
experiencing substantial increases as the number of
BotNets escalates. This comparative analysis
underscores the effectiveness of CoDel in reducing
network delay and enhancing overall performance,
particularly in scenarios with a high BotNet count.
Figure 6. Average End-to-end Delay Under DDoS Attacks
3.2.3 Average End-to-end PDF
Drop-tail, RED, CoDel, pFIFO and PIE were
compared using 10, 20, 40, 60 and 100 botnets in the LTE network environment. Table 5 shows the average
PDF values of the LTE RLC layer under DDoS attacks.
Table 5. End-to-end PDF
AQM
Average PDF(%)
10 BotNets
20 BotNets
40 BotNets
60 BotNets
100 BotNets
Drop-Tail
89
81
64
54
8
RED
97
87
75
69
51
CoDel
98
89
76
72
57
pFIFO
93
83
71
59
38
PIE
95
86
73
61
42
The packet delivery rate, PDF, is calculated as the
ratio of the total packets sent to the total packets
received. The PDF value is an important parameter that
shows the network's performance. Figure 6 shows the
end-to-end average PDF based on the number of botnets.
As the number of botnets increases, the demanded
amount of resources also increases. User packets are
accumulated in the queues of the RLC layer.
Accumulated packets are held for resource allocation.
Due to the DDoS attack, there is a decrease in PDF value
because the resource allocation is not sufficient. With an
increasing number of botnets, all algorithms get lower
PDF values due to excessive packet drop. CoDel, RED,
PIE, pFIFO ,and Drop-Tail get the best PDF values,
respectively. Figure 7 shows the average end-to-end
PDF values. The provided data illustrates the Average
PDF percentages for various AQM (Active Queue
Management) algorithms under different scenarios
involving varying numbers of BotNets. When faced
with 10 BotNets, the CoDel algorithm outperforms its
counterparts, achieving an impressive 98% Average
PDF. Following closely behind, the PIE algorithm
demonstrates substantial efficiency with a PDF of 95%.
Meanwhile, Drop-Tail and pFIFO exhibit decreasing
performance with PDF values of 89% and 93%
respectively. As the BotNet count escalates to 20, 40, 60,
and eventually 100, a clear trend emerges. Across all
scenarios, CoDel consistently maintains its lead,
demonstrating superior PDF percentages compared to
other AQM strategies. RED and PIE algorithms also
exhibit competitive performance, with both showcasing
robust PDF percentages. Conversely, Drop-Tail and
pFIFO face significant challenges as the number of
BotNets increases, with notable reductions in their
respective PDF values. This comparative analysis
highlights the effectiveness of CoDel in managing
network congestion, particularly in high-stress scenarios
with a substantial number of BotNets.
0
200
400
600
800
1000
1200
1400
1600
1800
2000
Drop-Tail RED CoDel pFIFO PIE
Average End-to-End Delay (ms)
10 BotNets 20 BotNets 40 BotNets 60 BotNets 100 BotNets
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 10
Figure 7. Average End-to-end PDF Under DDoS Attacks
3.2.4 Average End-to-end Fairness Index
Drop-tail, RED, CoDel, pFIFO and PIE were
compared using 10, 20, 40, 60 and 100 botnets in the
LTE network environment. Table 6 shows the average
fairness index values of the LTE RLC layer under DDoS
attacks.
Table 6. End-to-end Fairness Index
AQM
Fairness Index
10 BotNets
20 BotNets
40 BotNets
60 BotNets
100 BotNets
Drop-Tail
0,79
0,65
0,41
0,23
0,09
RED
0,88
0,81
0,66
0,58
0,41
CoDel
0,91
0,84
0,78
0,61
0,49
pFIFO
0,89
0,78
0,63
0,51
0,31
PIE
0,87
0,77
0,65
0,56
0,38
As the number of botnets connecting to eNB
increases, the amount of resources also demanded
increases. Normal users expect resource allocation, and
a fair allocation of resources is required for requested
resources. Figure 8 shows the end-to-end average
fairness index values according to the number of
botnets. As the DDoS attack increases, the packets
coming into the queue begin to drop. Excessive packet
drops reduce the fairness index value of the network.
Also, the attack causes network latency. CoDel does not
use the size of the queue to manage the queue; it uses the
queue time of the packets. This ensures that the fairness
index value of the CoDel algorithm is high. The CoDel
algorithm is followed by RED, PIE, pFIFO, and Drop-
Tail, respectively.
89
81
64
54
8
97
87
75
69
51
98
89
76
72
57
93
83
71
59
38
95
86
73
61
42
10 BotNets 20 BotNets 40 BotNets 60 BotNets 100 BotNets
End-to-end Average PDF (%)
Drop-Tail RED CoDel pFIFO PIE
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 11
Figure 8. Average End-to-end Fairness Index Under DDoS Attacks
Packets coming from the PDN-GW router
significantly affect parameters such as modulation,
coding, and packet generation time. These parameters
are negatively affected in a DDoS attack. CoDel does
not use the size of the queue to manage it; it uses the
time it takes for packets to queue. This ensures that the
packet drop rate of the CoDel algorithm is low.
Therefore, the end-to-end delay value of CoDel is lower
than that of other algorithms. More packets dropped due
to a DDoS attack cause more delay. The CoDel
algorithm yields the best results due to its early detection
of packet drops and processing of packets according to
their queue time. As the number of botnets increases, the
amount of requested resources also increases. User
packets are collected in the queues of the RLC layer. The
accumulated packets are retained for resource
allocation. Due to the DDoS attack, the PDF value
decreases because the resource allocation is insufficient.
With the increase in the number of botnets, all
algorithms obtain lower PDF values due to excessive
packet drops. As the DDoS attack escalates, the packets
arriving in the queue begin to drop. Excessive packet
drops reduce the fairness index value of the network.
Additionally, the attack causes network delay. CoDel
does not use the size of the queue to manage it; it uses
the queue time of the packets. This ensures that the
fairness index value of the CoDel algorithm is high.
4. Conclusion
In conclusion, our research underscores the
paramount importance of robust queue management
algorithms in fortifying LTE networks against the
disruptive impact of DDoS attacks. Through a
comprehensive examination of various techniques, our
study extends the existing body of knowledge in
network security beyond conventional paradigms. We
have systematically evaluated the performance of
CoDel, RED, PIE, pFIFO, and Drop-Tail algorithms
under simulated DDoS scenarios, shedding light on their
respective strengths and weaknesses. Notably, the
CoDel algorithm emerges as the standout performer,
leveraging packet waiting time to optimize packet loss,
latency, and end-to-end transmission. The RED
algorithm also demonstrates commendable
performance, strategically regulating packet drop
thresholds. Conversely, PIE, pFIFO, and Drop-Tail
algorithms face notable challenges in packet control
during DDoS attacks, with Drop-Tail exhibiting a
pronounced vulnerability. These findings provide
critical insights for network administrators and security
experts in devising robust defenses against escalating
DDoS threats in LTE networks.
Furthermore, this study contributes significantly to
the broader discourse on network security. By
elucidating the nuanced interplay between LTE
architecture and DDoS attacks, our research highlights
the pivotal role of effective queue management in
thwarting and mitigating such assaults. This
comprehensive evaluation of queue management
algorithms fills a notable gap in the current literature,
offering practical guidance for implementing tailored
security measures. Looking ahead, future research
endeavors could focus on the development of AI-driven
queue management algorithms fine-tuned for LTE
networks, further enhancing the resilience of these
crucial communication systems in the face of evolving
cyber threats.
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
10 BotNets 20 BotNets 40 BotNets 60 BotNets 100 BotNets
End-to-end Average Fairness Index
Drop-Tail RED CoDel pFIFO PIE
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 12
References
Albayrak, Z., Çakmak, M., 2018. A Review: Active
Queue Management Algorithms in Mobile
Communication. International Conference on
Cyber Security and Computer Science 180–184.
Ali, S.M., Çakmak, M., Albayrak, Z., 2022. Security
Classification of Smart Devices Connected to
LTE Network, in: Lecture Notes in Networks and
Systems. https://doi.org/10.1007/978-3-030-
94191-8_91
Amer, H., Al-Kashoash, H., Khami, M.J., Mayfield, M.,
Mihaylova, L., 2020. Non-cooperative game
based congestion control for data rate
optimization in vehicular ad hoc networks. Ad
Hoc Networks 107.
https://doi.org/10.1016/j.adhoc.2020.102181
Ashfaq, M.F., Malik, M., Fatima, U., Shahzad, M.K.,
2022. Classification of IoT based DDoS Attack
using Machine Learning Techniques, in: 2022
16th International Conference on Ubiquitous
Information Management and Communication
(IMCOM). IEEE, pp. 1–6.
https://doi.org/10.1109/IMCOM53663.2022.972
1740
Bisoy, S.K., Pattnaik, P.K., 2016. Design of feedback
controller for TCP/AQM networks. Engineering
Science and Technology, an International Journal
20.
https://doi.org/http://dx.doi.org/10.1016/j.jestch.2
016.10.002
Çakmak, M., Albayrak, Z., 2022. AFCC-r: Adaptive
Feedback Congestion Control Algorithm to Avoid
Queue Overflow in LTE Networks. Mobile
Networks and Applications 27.
https://doi.org/10.1007/s11036-022-02011-8
Çakmak, M., Albayrak, Z., 2020. Performance Analysis
of Queue Management Algorithms Between
Remote-Host and PG-W in LTE Networks.
Academic Platform Journal of Engineering and
Science 456–463.
https://doi.org/10.21541/apjes.662677
Çakmak, M., Albayrak, Z., Torun, C., 2021.
Performance comparison of queue management
algorithms in lte networks using NS-3 simulator.
Tehnicki Vjesnik 28.
https://doi.org/10.17559/TV-20200411071703
F. M. Suaib Akhter, A., F. M. Shahen Shah, A., Ahmed,
M., Moustafa, N., Çavuşoğlu, U., Zengin, A.,
2021. A Secured Message Transmission Protocol
for Vehicular Ad Hoc Networks. Computers,
Materials & Continua 68, 229–246.
https://doi.org/10.32604/cmc.2021.015447
Gomez, C.A., Wang, X., Shami, A., 2021. Federated
intelligence for active queue management in inter-
domain congestion. IEEE Access 9, 10674–
10685.
https://doi.org/10.1109/ACCESS.2021.3050174
Gómez, G., Pérez, Q., Lorca, J., García, R., 2014.
Quality of service drivers in LTE and LTE-A
networks. Wirel Pers Commun 75, 1079–1097.
https://doi.org/10.1007/s11277-013-1409-0
Gong, Y., Cao, J., Fu, Y., Guo, M., 2019. A DDoS attack
detection model for LTE-A network. Journal of
Cyber Security 4.
https://doi.org/10.19363/J.cnki.cn10-
1380/tn.2019.01.03
Israr, A., Yang, Q., Li, W., Zomaya, A.Y., 2021.
Renewable energy powered sustainable 5G
network infrastructure: Opportunities, challenges
and perspectives. Journal of Network and
Computer Applications.
https://doi.org/10.1016/j.jnca.2020.102910
Jevtić, M., Zogović, N., Dimić, G., 2009. Evaluation of
Wireless Sensor Network Simulators.
Proceedings of the 17th Telecommunications
Forum TELFOR 2009 Belgrade Serbia.
Jyothi, K.K., Chaudhari, S., 2020. Optimized neural
network model for attack detection in LTE
network. Computers & Electrical Engineering 88,
106879.
https://doi.org/10.1016/j.compeleceng.2020.1068
79
Krishnan, P., Duttagupta, S., Achuthan, K., 2019.
SDNFV Based Threat Monitoring and Security
Framework for Multi-Access Edge Computing
Infrastructure. Mobile Networks and Applications
24. https://doi.org/10.1007/s11036-019-01389-2
Lin, Y., Li, L., Ren, P., Wang, Y., Szeto, W.Y., 2021.
From aircraft tracking data to network delay
model: A data-driven approach considering en-
route congestion. Transp Res Part C Emerg
Technol 131.
https://doi.org/10.1016/j.trc.2021.103329
Low, S.H., Paganini, F., Jiantao Wang, Adlakha, S.,
Doyle, J.C., n.d. Dynamics of TCP/RED and a
scalable control, in: Proceedings.Twenty-First
Annual Joint Conference of the IEEE Computer
and Communications Societies. IEEE, pp. 239–
248.
https://doi.org/10.1109/INFCOM.2002.1019265
Mousavi, H., Amiri, I.S., Mostafavi, M.A., Choon, C.Y.,
2019. LTE physical layer: Performance analysis
and evaluation. Applied Computing and
Informatics 15.
https://doi.org/10.1016/j.aci.2017.09.008
Mousavi, H., Amiri, I.S., Mostafavi, M.A., Choon, C.Y.,
2017. LTE physical layer: Performance analysis
and evaluation. Applied Computing and
Informatics.
https://doi.org/10.1016/j.aci.2017.09.008
N.D., A., A., R., 2019. Avoiding queue overflow and
reducing queuing delay at eNodeB in LTE
networks using congestion feedback mechanism.
Comput Commun 146, 131–143.
https://doi.org/10.1016/j.comcom.2019.07.015
Journal of Intelligent Systems: Theory and Applications 7(1) (2024) 1-13 13
Oughton, E.J., Comini, N., Foster, V., Hall, J.W., 2022.
Policy choices can help keep 4G and 5G universal
broadband affordable. Technol Forecast Soc
Change 176.
https://doi.org/10.1016/j.techfore.2021.121409
Pan, R., Natarajan, P., Piglione, C., Prabhu, M.S.,
Subramanian, V., Baker, F., VerSteeg, B., 2013.
PIE: A lightweight control scheme to address the
bufferbloat problem. IEEE International
Conference on High Performance Switching and
Routing, HPSR 148–155.
https://doi.org/10.1109/HPSR.2013.6602305
Paul, A., Kawakami, H., Tachibana, A., Hasegawa, T.,
2017. Effect of AQM-Based RLC Buffer
Management on the eNB Scheduling Algorithm in
LTE Network. Technologies (Basel) 5, 59.
https://doi.org/10.3390/technologies5030059
Praveen, K. v., Prathap, P.M.J., 2021. Energy Efficient
Congestion Aware Resource Allocation and
Routing Protocol for IoT Network using Hybrid
Optimization Techniques. Wirel Pers Commun
117. https://doi.org/10.1007/s11277-020-07917-8
Raghuvanshi, D.M., Annappa, B., Tahiliani, M.P., 2013.
On the effectiveness of CoDel for active queue
management. International Conference on
Advanced Computing and Communication
Technologies, ACCT 107–114.
https://doi.org/10.1109/ACCT.2013.27
Said, A.A., Çakmak, M., Albayrak, Z., 2022.
Performance of Ad-Hoc Networks Using Smart
Technology Under DDoS Attacks, in: Lecture
Notes in Networks and Systems.
https://doi.org/10.1007/978-3-030-94191-8_92
Swain, S.K., Nanda, P.K., 2021. Adaptive queue
management and traffic class priority based
fairness rate control in wireless sensor networks.
IEEE Access 9.
https://doi.org/10.1109/ACCESS.2021.3102033
Szyguła, J., Domański, A., Domańska, J., Marek, D.,
Filus, K., Mendla, S., 2021. Supervised learning
of neural networks for active queue management
in the internet. Sensors 21.
https://doi.org/10.3390/s21154979
Verma, H., Chauhan, N., Chand, N., Awasthi, L.K.,
2022. Buffer-loss estimation to address
congestion in 6LoWPAN based resource-
restricted ‘Internet of Healthcare Things’ network.
Comput Commun 181, 236–256.
https://doi.org/10.1016/j.comcom.2021.10.016
Wang, T., Wang, M., 2020. Hyperchaotic image
encryption algorithm based on bit-level
permutation and DNA encoding. Opt Laser
Technol 132, 106355.
https://doi.org/10.1016/j.optlastec.2020.106355
Weingärtner, E., vom Lehn, H., Wehrle, K., 2009. A
performance comparison of recent network
simulators, in: IEEE International Conference on
Communications.
https://doi.org/10.1109/ICC.2009.5198657
Wu, Z., Zhu, M., Li, Q., Xue, L., Yang, J., Chen, Z.,
Cao, Y., Cui, Y., 2022. Design of power
monitoring system for new energy grid-connected
operation based on LoRa and 4G technology.
Energy Reports 8, 95–105.
https://doi.org/10.1016/j.egyr.2022.10.038
Zenitani, K., 2023. From attack graph analysis to attack
function analysis. Inf Sci (N Y) 119703.
https://doi.org/10.1016/j.ins.2023.119703
Zidic, D., Mastelic, T., Nizetic Kosovic, I., Cagalj, M.,
Lorincz, J., 2023. Analyses of ping-pong
handovers in real 4G telecommunication
networks. Computer Networks 227.
https://doi.org/10.1016/j.comnet.2023.109699
