Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic

Reconﬁgurability Approach

Osama Mohammed Dighriri

1,2 a

, Priyadarsi Nanda

2 b

, Manoranjan Mohanty

3 c

and Ibrahim Haddadi

4 d

Department of Computer Science, College of Engineering and Computer Science, JAZAN University, K.S.A.

School of Electrical and Data Engineering, Faculty of Engineering and IT, University of Technology Sydney, Australia

School of Information System, Carnegie Mellon University, Qatar

Department of Computer Engineering, College of Computer Science and Engineering, TAIBAH University, K.S.A.

Keywords:

Internet of Things (IoT), Hybrid and Secure Clustering, Dynamic Reconﬁgurability, AES-128 Encryption.

Abstract:

Current Internet of Things (IoT) networks face signiﬁcant challenges in scalability, energy efﬁciency, and

security within resource-constrained environments. This paper proposes a hybrid clustering framework com-

bining BIRCH with DBSCAN algorithms while integrating AES-128 encryption for secure communication.

Our proposed scheme is implemented using Contiki-NG simulator and analyzed using Python 3. Our approach

demonstrates a 20% improvement in resource utilization, 43.26% reduction in latency, and 99.91% message

success rate conducted across 2,154 test runs, with security overhead below 5%. This reduces cluster recon-

ﬁguration frequency and improves system stability, addressing limitations in adaptability, data integrity, and

optimization for dynamic IoT infrastructures.

1 INTRODUCTION

The Internet of Things (IoT) is rapidly expanding,

with connected devices expected to reach 30 billion

by 2027 (Mansour et al., 2023). However, IoT net-

works face critical challenges, such as; scalability, in-

teroperability, and security issues due to constrained

resources and heterogeneous devices (Nasri et al.,

2021; Canavese et al., 2024).

Industrial IoT deployments such as; smart grids,

factory automation, and remote monitoring require

real-time secure clustering to ensure timely data ag-

gregation, prevent unauthorized access, and maintain

low latency despite limited energy and hardware ca-

pabilities (Kumar et al., 2021; Gyamﬁ and Jurcut,

2022; Zhang et al., 2023). Traditional encryption

methods, although effective, often introduce signiﬁ-

cant power consumption, reducing device life span.

Existing clustering and edge computing methods of-

fer partial solutions but, frequently fall short in ad-

https://orcid.org/0000-0002-1917-0275

https://orcid.org/0000-0002-5748-155X

https://orcid.org/0000-0002-0258-4586

https://orcid.org/0000-0002-7557-4258

dressing adaptability, security, and efﬁciency simulta-

neously (Bouguettaya et al., 2021; Xiang et al., 2023;

Chiang et al., 2023). Moreover, static clustering lacks

ﬂexibility, leading to instability and reduced perfor-

mance in dynamic environments.

Motivated by these challenges, this study pro-

poses a hybrid clustering framework by leveraging

BIRCH for scalability and DBSCAN for reﬁnement,

enhanced with adaptive QoS metrics for dynamic

reconﬁguration and AES-128 encryption for secure,

low-overhead communication. The framework ad-

dresses the need for secure, efﬁcient, and adaptive

IoT clustering, enabling real-time adjustments, opti-

mizing performance, and improving resilience.

Our key contributions in this paper include but not

limited to:

• Design of hybrid clustering method addressing

scalability, security, and adaptability limitations

in dynamic IoT environments.

• A QoS-driven reconﬁguration approach that dy-

namically optimizes network latency, throughput,

and energy efﬁciency.

• A lightweight AES-128 encryption framework

ensuring robust security for IoT communication

412

Dighriri, O. M., Nanda, P., Mohanty, M. and Haddadi, I.

Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconﬁgurability Approach.

DOI: 10.5220/0013525600003979

In Proceedings of the 22nd International Conference on Security and Cryptography (SECRYPT 2025), pages 412-419

ISBN: 978-989-758-760-3; ISSN: 2184-7711

with minimal computational overhead.

The remainder of this paper as follows: Section 2

reviews relevant literature; Section 3 details the pro-

posed method; Section 4 outlines the system model;

Section 5 discusses results and comparative analysis;

and Section 6 concludes future research directions.

2 BACKGROUND AND RELATED

WORK

IoT clustering techniques have evolved signiﬁcantly,

addressing challenges like resource constraints, scal-

ability, and heterogeneity. These techniques primar-

ily aim to optimize energy efﬁciency, extend network

longevity, and enhance security while balancing com-

putational overheads.

Evolutionary algorithms, including Genetic Algo-

rithms (GA) and Fuzzy Logic (FL), have improved

IoT clustering efﬁciency but face scalability chal-

lenges in large-scale deployments (Ajayi et al., 2021;

Mohamed et al., 2020). Similarly, Swarm-based

methods such as Particle Swarm Optimization (PSO)

and Ant Colony Optimization (ACO) have shown

promise in adaptive clustering but, struggle with con-

sistent reconﬁguration in dynamic settings (Sadrisho-

jaei et al., 2022).

Security introduces critical trade-offs, often in-

creasing computational overhead. Lightweight AES-

based encryption approaches have emerged, sig-

niﬁcantly reducing resource use (Ferrag et al.,

2020). Nevertheless, studies indicate frequent secu-

rity breaches in deployments prioritizing energy efﬁ-

ciency over robust encryption, highlighting the neces-

sity of adaptive security solutions (Li et al., 2023).

Dynamic reconﬁguration methods are essential

to maintain IoT clustering efﬁciency in ﬂuctuat-

ing environments. Techniques like Predictive Edge

Computing (PEC) offer improved responsiveness

but, face scalability limitations (Jere et al., 2020).

Context-aware reconﬁguration approaches demon-

strate adaptability yet, struggle in highly dynamic,

large-scale conditions (Wang et al., 2022). Software-

Deﬁned Networking (SDN)-based solutions also suf-

fer throughput limitations when deployed in dynamic

infrastructures (Abuarqoub, 2020).

Recent hybrid frameworks integrating clustering

and security show promise but, often lack compre-

hensive optimization in heterogeneous deployments.

Building on these insights, our study introduces a uni-

ﬁed framework integrating hierarchical BIRCH and

DBSCAN clustering methods, combined with AES-

128 encryption. Our approach addresses existing gaps

by dynamically optimizing security, latency, energy

efﬁciency, and scalability simultaneously.

3 PROPOSED FRAMEWORK

Our work integrates three major components as il-

lustrated in Figure 1: (1) A hybrid clustering frame-

work combining BIRCH’s coarse-grained classiﬁca-

tion with DBSCAN’s reﬁnement capabilities for en-

hanced accuracy and noise detection; (2) QoS-driven

dynamic regulation where cluster centers optimize

network efﬁciency under changing conditions; and

(3) Secure communication using lightweight AES-

128 encryption.

The motivation for combining BIRCH and DB-

SCAN stems from the need to balance scalability,

adaptability, and clustering precision in dynamic IoT

environments. BIRCH efﬁciently processes large-

scale datasets with minimal memory overhead, while

DBSCAN provides robustness against noise and out-

lier detection. Together, this hybrid approach enables

responsive and resilient clustering under varying net-

work conditions and data distributions.

Figure 1: Framework Architecture.

The architecture operates across three distinct lay-

ers: The edge node (top layer) manages monitor-

ing, hybrid clustering, and dynamic reconﬁgurabil-

ity. Cluster centers (middle layer) execute local-

ized clustering, QoS monitoring, and data aggrega-

tion. Client nodes (bottom layer) report performance

metrics and communicate securely via AES-128 en-

cryption. The implementation uses Contiki-NG for

cluster-based logic and adaptive reconﬁguration, with

simulation outputs processed through Python 3 for

evaluation.

3.1 Mathematical Formulas

This section presents the problem formulation within

four essential components:

Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconﬁgurability Approach

413

• Node Metrics: Performance parameters govern-

ing clustering decisions

= (l

, t

, e

, q

) (1)

• Clustering and Reﬁnement: BIRCH-inspired

feature vectors enabling real-time adaptation

= (N

, LS

, SS

) (2)

Node similarity uses a normalized distance metric:

D(n

, n

) =

∑

d∈{l,t,e}



[d] − n

[d]

max

− min



(3)

DBSCAN identiﬁes core points using:

CorePoint(p) = |{q : D(p, q) ≤ ε}| ≥ MinPts (4)

• Dynamic Cluster Management: QoS-driven

size adjustment and load balancing, where cluster

size is dynamically determined by:

Size(C

) = min

MAX Ck,

∑

i=1

QoS Threshold

(5)

Thresholds adapt to QoS changes:

T (t) = T

· (1 +α · ∆QoS) (6)

Load balancing uses load factors:

∑

i=1

MAX RANGE

(7)

• Performance Metrics: QoS evaluation integrat-

ing reliability and resource utilization

Cluster QoS combines message success rate with

resource utilization:

QoS(G

) =



mesgs acked

mesgs sent

active nodes

MAX CLUSTER SIZE



(8)

Energy efﬁciency is maintained by node proximity

enforcement:

total

(k) =

∑

i=1



1 + α ·

MAX RANGE



(9)

System performance is evaluated through cluster

stability and global efﬁciency:

S(C

) =



1 −

reconf count

MAX RECONFS



· QoS(C

) (10)

global

∑

k=1

| · S(C

)

∑

k=1

(11)

3.2 Proposed Algorithm

Our framework operates in three phases: hybrid clus-

tering, dynamic reconﬁguration, and secure commu-

nication. Algorithm 1 synthesizes these components,

where BIRCH provides initial coarse-grained classiﬁ-

cation, DBSCAN reﬁnes cluster boundaries, and con-

tinuous QoS monitoring enables adaptive optimiza-

tion. AES-128 encryption ensures data protection

with minimal overhead.

Input: IoT Dataset D, Parameters T

, Thresholds

, L

, T

}

Output: Optimized Clusters C, QoS Metrics,

Performance P

global

begin

Initialize system parameters and communication

links;

// Hybrid Clustering Phase

Apply BIRCH clustering for initial clusters using

(Eq. 2);

Reﬁne using DBSCAN with metric D(n

, n

) (Eq.

3);

// Dynamic Reconfiguration Phase

while System is Active do

foreach Cluster Center C

Monitor metrics and recalculate QoS(C

)

(Eq. 8);

if metrics exceed thresholds then

Update thresholds using Eq. 6;

Reassign nodes based on updated

parameters;

end

Apply AES-128 encryption for secure

communication;

Validate data integrity and update

conﬁgurations;

end

Compute P

global

(Eq. 11);

end

return Optimized Clusters, QoS Metrics, and

global

;

end

Algorithm 1: Hybrid Secure Clustering with Dynamic Re-

conﬁgurability.

4 IMPLEMENTATION AND

SYSTEM MODEL

This section outlines the implementation strategy and

system model tailored for constrained IoT environ-

ments, optimizing IPv6-based communication lever-

ages IEEE 802.15.4, 6LoWPAN compression, and

RPL-Lite routing. The edge node coordinates cluster

centers, each managing client nodes. Cluster centers

aggregate data, balance workloads, and ensure consis-

tent messaging. Client nodes, distributed across ap-

plications continuously report performance metrics.

SECRYPT 2025 - 22nd International Conference on Security and Cryptography

414

4.1 Experimental Setup

The simulation was conducted within a 100× 100 m

area, as depicted in Figure 2, with a communication

radius of 100 meters. The system achieves complete

connectivity, a latency threshold of 60 ms, and a data

rate above 1 Mbps. Energy usage ranges between

0.1 and 7.0 watts based on node-speciﬁc constraints.

The network consists of an Edge Node (Green Mote)

serving as the central node for data aggregation and

synchronization, Cluster Centers (Yellow Motes) that

conduct adaptive clustering, and routing, and Client

Nodes (Purple Motes) that transmit real-time metrics

(latency, throughput, energy, QoS) securely. The sim-

ulation uses Cooja motes and Contiki-NG’s internal

power proﬁling, accurately modeling resource efﬁ-

ciency and scalability.

Figure 2: Simulated topology showing mote connectivity.

4.2 Dynamic Reconﬁgurability

To maintain optimal performance, the framework dy-

namically adapts based on real-time metrics. As de-

picted in Figure 3, reconﬁguration is triggered when

energy consumption drops below 20%, mobility ex-

ceeds 50%, or throughput falls below 80% of base-

line values. Otherwise, current conﬁgurations remain.

This appraoch is crucial in IoT environments, where

rapid changes in mobility, energy, or trafﬁc can de-

grade performance. By continuously adjusting clus-

tering parameters, the system improves resilience, re-

duces unnecessary reconﬁgurations, and sustains QoS

under dynamic conditions.

Tigger Reconfiguration

No Action

Mobility > 50% Range

Monitor Metrics

Start

Throughput < 80%

Energy < 20%

Figure 3: Dynamic Reconﬁguration Decision Flowchart.

4.3 Assumptions

The framework assumes performance feasibility

where nodes can manage clustering, security, and

IPv6 communication, scalability and Resilience that

supports over 2,000 simultaneous transmissions, self-

adjusting QoS where devices autonomously adapt to

changing network conditions, and distributed coordi-

nation in which the edge node oversees real-time clus-

tering and synchronization.

4.4 Design Requirements

Key operational requirements include integrated clus-

tering with initial segmentation through BIRCH and

reﬁnement by DBSCAN, real-time adaptability offer-

ing adaptive reconﬁguration for efﬁciency and sta-

bility, and secure communication utilizing AES-128

encryption with minimal overhead, and performance

consistency optimization.

5 RESULTS AND DISCUSSION

5.1 Hybrid Clustering Algorithm

This section evaluates our hybrid clustering method-

ology that integrates BIRCH with DBSCAN to in dy-

namic IoT environments. Performance metrics are

improved through data parsing to extract relevant

metrics, with cluster labels normalized using Python

3. Figure 4 compares the three clustering approaches

that manage dynamic network conditions efﬁciently.

Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconﬁgurability Approach

415

Figure 4: Comparison of clustering results on IoT metrics: the left plot shows clear clusters with few outliers, the center

detects anomalies well but is noise-sensitive, and the right achieves a balanced, structured grouping.

BIRCH forms well-separated clusters with high

computational efﬁciency, suitable for large datasets,

but lacks outlier detection capabilities. DBSCAN

excels at anomaly detection (gray-labeled instances)

but often produces larger, less precise clusters with

increased likelihood of misclassifying valid pat-

terns. Our hybrid approach combines BIRCH’s struc-

tured clustering with DBSCAN’s anomaly detection,

achieving clear separation while effectively distin-

guishing valid patterns from anomalies—making it

ideal for IoT environments requiring both accuracy

and adaptability.

Figure 5: Connection distribution across clusters.

The dataset comprises 2,154 communications dis-

tributed across clusters 0–9 as shown in Figure 5.

Cluster distribution demonstrates specialized func-

tionality: high-volume exchange (Cluster 0, 15.2%),

connectivity aggregation (Cluster 2, 13.4%), trans-

mission intermediation (Cluster 1, 13.0%), network

trafﬁc balancing (Clusters 3 and 5, 11.6% and 9.8%),

specialized tasks (Cluster 4, 8.1%), occasional inter-

actions (Cluster 7, 7.0%), and general-purpose com-

munication (Clusters 6, 8, 9, 7.0–10.2%).

Figure 6 provides an integrated view of latency,

throughput, energy consumption, and QoS distribu-

tions across various clusters. The plot shows perfor-

mance metrics across 10 clusters (0-9). Cluster 7 has

high latency ( 48ms) and throughput ( 45 Mbps) with

moderate energy use ( 3.4W). Cluster 3 offers maxi-

mum throughput ( 47 Mbps) but highest energy con-

sumption ( 6.0W). Cluster 8 provides minimal latency

( 10ms) ideal for time-critical applications but lim-

ited throughput ( 2 Mbps). Cluster 0 excels in energy

efﬁciency ( 0.2W) but with performance trade-offs.

Cluster 6 achieves the highest QoS ( 0.61) with bal-

anced performance metrics, while Cluster 4 shows the

lowest QoS ( 0.40). Clusters 1, 2, 5, and 9 maintain

intermediate performance proﬁles across all metrics,

offering balanced operation for general-purpose ap-

plications. These distinct proﬁles enable selecting ap-

propriate cluster conﬁgurations based on speciﬁc IoT

application requirements.

Our hybrid framework outperforms alternatives,

maintaining a mean QoS of 0.52 with optimized

energy use (2.74W), supports bandwidth-intensive

operations (23.35 Mbps throughput) with low la-

tency (28.38ms), and achieves exceptional reliability

(99.95% message success rate).

5.2 Security Implementation

Our IoT clustering security solution integrates the

hybrid BIRCH-DBSCAN method with AES-128 en-

cryption to ensure robust data protection while main-

taining performance efﬁciency. The system adapts

dynamically to environmental variability, with secu-

rity embedded through two key strengths: (1) per-

fect data authenticity via bit-to-bit matching after

encryption-decryption cycles, and (2) operational re-

silience with stable, low-latency processing suitable

for IoT deployments. Figure 7 illustrates the encryp-

tion and decryption results, demonstrating the frame-

work’s resilience and energy efﬁciency. Encrypted

data shows high entropy with no visible patterns. Us-

ing secure random keys and PKCS7 padding, the

dataset size increased from 25,379 to 34,230 bytes

SECRYPT 2025 - 22nd International Conference on Security and Cryptography

416

Figure 6: Integrated View of Cluster Metrics: Latency, Throughput, Energy Consumption, and QoS Distributions.

to support 2,154 connections. Encryption took only

0.0020 seconds, while decryption restored the data

in 0.0012 seconds with 100% integrity veriﬁcation.

Despite a 34.9% increase in size due to padding and

headers, processing latency decreased from 2ms to

1.2ms. Bandwidth usage dropped slightly (from 250

to 240 kbps), preserving real-time execution. En-

cryption throughput reached 12.36 MB/s, and decryp-

tion 20.60 MB/s, conﬁrming the framework’s suitabil-

ity for lightweight, secure data exchange in resource-

constrained IoT environments.

# Encryption and Decryption Results

## 4.1 Original Data

Length: 25379 bytes

Sample: b"\xeb\xedg\x3d\xbe\xa3\xfb\x0eU\xd2\xaf\xbc\xcOV\xxBA\xa7\x97\xf8

\xd\x03\xc0\xax\xbs\xda\xba\x2\x19\xaa\xal\xbecB\xdF\xBs\xBeb3\xFa(\xcl"

## 4.2 Encrypted Data

Length: 34230 bytes

Sample: b"gARAAABNiajtev«popgzeTSFFjjGyDUIVeLTAoNEp7OHDFGCNAB"

## 4.3 Decrypted Data

Length: 25379 bytes

Sample: b"\xeb\xedg\xod\xbe\xa3\xfb\x0eU\xd\xa\xbc\xcOV\xa\xa7\x97\xf\xd\x03

\xc0\xa\xb\xca\xba\xd2\x19\xaal\xbeCB\xdFBS\xBeb3\xFa(\xct"

## 4.4 Performance Analysis

Encryption Time: 0.0020 seconds

Decryption Time: 0.0012 seconds

Data Integrity Check: Successful

Figure 7: Encryption and decryption results.

The cluster analysis reveals three distinct de-

vice proﬁles based on performance characteristics as

shown in both ﬁgures 8 and 9. The security-critical

devices (clusters 0, 5, 9 - represented by circles) prior-

itize data protection with higher energy consumption

( 2.5W) and superior QoS (0.575-0.650), operating

at moderate latency (22.5-30ms) and throughput (22-

30 Mbps). Energy-constrained devices (clusters 4, 7,

8 - marked with ”x”) optimize for power efﬁciency

with minimal energy usage ( 0.5W), moderate QoS

(0.450-0.550), lower latency (10-15ms), and reduced

throughput (2-5 Mbps). General-purpose devices

(clusters 1, 2, 3, 6 - shown as squares) maintain bal-

anced performance with moderate energy consump-

tion ( 2.0W), stable QoS (0.475-0.550), and balanced

latency (17-22.5ms) and throughput (15-25 Mbps).

These patterns demonstrate how the clustering frame-

work effectively segments IoT devices according to

their operational requirements, enabling optimal re-

source allocation for diverse application scenarios.

Figure 10 illustrates the security-performance

trade-offs across clusters, where security scores (nor-

malized between 0 and 1) are derived from encryp-

tion overhead, communication latency, and opera-

tional complexity. Lower scores (0.24–0.29) indi-

cate high-security clusters with stringent encryption

policies (e.g., frequent AES-128 encryption and re-

duced ﬂexibility), while higher scores (0.64–0.66)

reﬂect energy-efﬁcient clusters with minimal con-

straints. General-purpose clusters maintain moder-

ate scores (0.32–0.39), balancing security with per-

formance requirements.

Figure 11 reveals performance metric correla-

tions. Processing time strongly correlates with data

exchange rates ( 0.96), while power utilization per-

fectly correlates with data transmission ( 1.00). Se-

Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconﬁgurability Approach

417

Figure 8: Cluster Distribution (Latency vs Throughput)

Figure 9: Energy vs QoS by device type

Figure 10: Security Score Distribution across clusters.

curity measures show negative correlation with per-

formance metrics (-0.93 to -0.94), conﬁrming the

security-performance trade-off. QoS and security ex-

hibit inverse relationship (-0.64), indicating that oper-

ational optimization may affect resilience.

5.3 Comparative Analysis

Table 1 highlights our framework’s superior perfor-

mance compared to recent approaches. It achieves

Figure 11: Correlation heatmap showing trade-offs.

28.37 ms latency (43.26% improvement over base-

line and 20% better than DLGA), 92% energy efﬁ-

ciency (exceeding Trust-HC, D-ACSM, and DLGA

at 76.5%, 74.2%, and 81.7% respectively), and

23.35 Mbps throughput (compared to Trust-HC at

14.8 Mbps, D-ACSM at 18.1 Mbps, and DLGA at

19.9 Mbps). Network stability reaches 98% ver-

sus competitors’ 81.3-87.9%, with a 99.91% mes-

sage delivery success rate. Unlike Trust-HC’s trust-

based mechanism and D-ACSM’s partial security, our

framework implements comprehensive AES-128 en-

cryption, making it ideal for time-critical, energy-

constrained IoT deployments.

6 CONCLUSION AND FUTURE

WORKS

This paper proposed a hybrid clustering framework

combining BIRCH, DBSCAN, and AES-128 encryp-

tion to enhance security, adaptability, and perfor-

mance in IoT networks. The system achieved no-

table improvements in power efﬁciency (20%), la-

tency reduction (43.26%), and message success rate

(99.91%). The framework presents certain limita-

tions: increased implementation complexity, due to

reliance on simulation. Future work will address these

challenges by extending the framework to real-world

IoT testbeds, improving interoperability with existing

protocols and standards, and assessing its robustness

across diverse network topologies. We also plan to re-

lease the implementation as open source and explore

lightweight enhancements to further reduce security

overhead in resource-constrained environments.

SECRYPT 2025 - 22nd International Conference on Security and Cryptography

418

Table 1: Detailed Performance Comparison of Recent Approaches.

Feature Our Work Trust-HC D-ACSM DLGA

(2025) (Lapegna et al., 2023) (Balakrishna, 2022) (Merah et al., 2024)

Latency (ms) 28.37 46.1 41.5 35.8

Energy Efﬁciency (%) 92 76.5 74.2 81.7

Throughput (Mbps) 23.35 14.8 18.1 19.9

Security Implementation AES-128 Trust-based Limited SDN-based

Network Stability (%) 98 81.3 84.7 87.9

Message Success Rate (%) 99.91 84.9 87.5 89.8

ACKNOWLEDGMENTS

This work was supported by the Department of Com-

puter Science, College of Engineering and Computer

Science, JAZAN University, KSA; Department of

Computer Engineering, College of Computer Science

and Engineering, TAIBAH University, KSA; and the

School of Electrical and Data Engineering, Faculty of

Engineering and IT, University of Technology Syd-

ney, Australia.

REFERENCES

Abuarqoub, A. (2020). A review of the control plane scala-

bility approaches in software-deﬁned networking. Fu-

ture Internet, 12(3):49.

Ajayi, O., Rafferty, J., Santos, J., Garcia-Constantino, M.,

and Cui, Z. (2021). Beca: A blockchain-based edge

computing architecture for internet of things systems.

IoT, 2:610–632.

Balakrishna, S. (2022). D-acsm: A technique for dy-

namically assigning and adjusting cluster patterns for

iot data analysis. The Journal of Supercomputing,

78:12873–12897.

Bouguettaya, A., Sheng, Q. Z., Benatallah, B., Neiat, A. G.,

Mistry, S., Ghose, A., Nepal, S., and Yao, L. (2021).

An internet of things service roadmap. Commun.

ACM, 64(9):86–95.

Canavese, D., Mannella, L., Regano, L., and Basile, C.

(2024). Security at the edge for resource-limited iot

devices. Sensors, 24(2).

Chiang, Y., Zhang, Y., Luo, H., Chen, T.-Y., Chen, G.-H.,

Chen, H.-T., Wang, Y.-J., Wei, H.-y., and Chou, C.-T.

(2023). Management and orchestration of edge com-

puting for iot: A comprehensive survey. IEEE Internet

of Things Journal, PP:1–1.

Ferrag, M. A., Shu, L., Yang, X., Derhab, A., and Maglaras,

L. (2020). Security and privacy for green iot-based

agriculture: Review, blockchain solutions, and chal-

lenges. IEEE Access, PP.

Gyamﬁ, E. and Jurcut, A. (2022). Intrusion detection

in internet of things systems: A review on design

approaches leveraging multi-access edge computing,

machine learning, and datasets. Sensors, 22:3744.

Jere, S., Fan, Q., Shang, B., Li, L., and Liu, L. (2020).

Federated learning in mobile edge computing: An

edge-learning perspective for beyond 5g. ArXiv,

abs/2007.08030.

Kumar, U., Verma, P., and Abbas, S. (2021). Bringing edge

computing into iot architecture to improve iot network

performance. pages 1–5.

Lapegna, M., Mele, V., and Romano, D. (2023). Cluster-

ing algorithms for enhanced trustworthiness on high-

performance edge-computing devices. Electronics,

12:1689.

Li, X., Song, Y., and Li, Y. (2023). An approach to de-

tect iot botnet attacks based on hierarchical clustering.

page 75.

Mansour, M., Gamal, A., Ahmed, A. I., Said, L. A., Elbaz,

A., Herencsar, N., and Soltan, A. (2023). Internet of

things: A comprehensive overview on protocols, ar-

chitectures, technologies, simulation tools, and future

directions. Energies, 16(8).

Merah, M., Aliouat, Z., and Mabed, H. (2024). Dynamic

load balancing of trafﬁc in the iot edge computing en-

vironment using a clustering approach based on deep

learning and genetic algorithms. Cluster Computing,

28.

Mohamed, A., Saber, W., Elnahry, I., and Hassanien, A. E.

(2020). Coyote optimization based on a fuzzy logic

algorithm for energy-efﬁciency in wireless sensor net-

works. IEEE Access, 8:185816–185829.

Nasri, M., Lamiri, A., Maaref, H., and Mghaieth, R. (2021).

Adaptive dynamic multi-hop technique for clustering

protocol in wireless sensor networks assisted-internet

of things applications. IET Networks, 11.

Sadrishojaei, M., Navimipour, N., Reshadi, M., Hossein-

zadeh, M., and Unal, M. (2022). An energy-aware

clustering method in the iot using a swarm-based al-

gorithm. Wireless Networks, 28.

Wang, J., Ping, G., Wang, H., Zhang, W., Sun, C., Zhao, B.,

and Chen, e. (2022). A right transfer access control

model of internet of things based on smart contract.

Security and Communication Networks, 2022:1–11.

Xiang, Z., Wang, D., He, M., and Chen, Y. (2023). Dy-

namic system reconﬁguration in stable and green edge

service provisioning. Mobile Networks and Applica-

tions, pages 1–17.

Zhang, Y., Yu, H., Zhou, W., and Man, M. (2023). Applica-

tion and research of iot architecture for end-net-cloud

edge computing. Electronics, 12(1).

Secure and Hybrid Clustering for IoT Networks: An Adaptive Dynamic Reconﬁgurability Approach

419