IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO.

4, FOURTH QUARTER 2024 2547

Design Guidelines on Trust Management for

Underwater Wireless Sensor Networks
Rongxin Zhu , Member, IEEE, Azzedine Boukerche, Fellow, IEEE, Libo Long, and Qiuling Yang

Abstract—In recent years, significant advancements in wireless a crucial means to expand human habitation areas and bol-
underwater communication and acoustic sensor technology have ster resource reserves [1], [2]. Governments worldwide have
spurred the exploration and utilization of the ocean’s vast natural placed significant attention on harnessing marine resources and
resources. underwater wireless sensor networks (UWSNs) are
increasingly deployed in unattended and hostile environments, have been actively enhancing their national marine develop-
demanding robust security measures. Secure communication ment strategies. Among the emerging technologies, underwater
environments are essential for a range of UWSN applica- wireless sensor networks (UWSNs) have garnered consider-
tions, including coastal defense, underwater communication, and able interest from academia and industry due to their potential
marine exploration. Trust models have emerged as effective to facilitate reliable and long-term monitoring of underwater
security mechanisms to assess the reliability of individual nodes in
UWSNs during adverse attacks. Unlike Wireless Sensor Networks environments, real-time surveillance of underwater activities,
(WSNs), UWSNs encounter distinct challenges, including con- and the provision of marine disaster warning services [3], [4].
strained resources, harsh underwater conditions, and unreliable As a distinctive subset of Wireless Sensor Networks
acoustic communication, making it crucial to establish a reliable (WSNs), underwater sensor networks share common design
trust-based system. In this paper, we review existing work on objectives with conventional terrestrial WSNs [5]. They
UWSN security, discuss security and trust challenges, and explore
trust-based applications. Furthermore, we evaluate diverse trust involve the deployment of wireless communication-capable
models suited for UWSNs of recent years, categorizing and sensors in designated areas of interest to effectively detect
comparing approaches like weighted sum methods, logic-based events and transmit the collected data back to correspond-
techniques, probability and statistics models, and machine learn- ing data centers for further analysis [6], [7]. However, a
ing paradigms. Finally, we discuss contemporary challenges fundamental difference arises as the primary deployment envi-
and future directions in UWSN trust management. By offering
a systematic overview and classification of trust management ronment for underwater sensor networks is underwater, where
approaches, this paper contributes to the understanding and the unique properties of water, particularly its strong absorp-
development of effective trust mechanisms for UWSNs, ultimately tion effects on electromagnetic/optical (EM/O) signals [8],
enhancing their reliability, security, and successful operation in render traditional data transmission technologies based on
diverse marine applications. EM/O signals ineffective underwater. Indeed, underwater com-
Index Terms—Trust management, UWSNs, underwater wire- munication primarily relies on acoustic signals, which can be
less sensor networks, security. efficiently transmitted over long distances in the underwater
domain. Consequently, underwater wireless sensor networks
based on acoustic communication are often referred to as
Underwater Acoustic Sensor Networks (UASNs) [9].
I. I NTRODUCTION
In underwater applications, the deployment of sensor nodes
N RECENT years, there has been a growing emphasis on
I the development and exploitation of marine resources as
takes place in unattended and potentially hostile environments,
making them susceptible to capture and attacks. Ensuring the
security of these sensor nodes becomes imperative for the
Manuscript received 4 August 2023; revised 27 November 2023 and efficient functioning of UWSNs. While traditional security
4 February 2024; accepted 12 March 2024. Date of publication 16 April 2024;
date of current version 22 November 2024. This work was supported in technologies, such as encryption and identity authentication,
part by NSERC, Canada Research Chairs Program; in part by the National are effective against external intrusions [10], they may prove
Natural Science Foundation of China under Grant 62362026; in part by the inadequate against successful invasions by attackers [11], [12].
Key Project of Hainan Province under Grant ZDYF2023GXJS158; and in part
by the Specific Research Fund of the Innovation Platform for Academicians In recent years, trust mechanisms have shown promise in
of Hainan Province under Grant YSPTZX202314. (Corresponding author: detecting insider attacks. Trust models have been proposed as
Qiuling Yang.) effective security measures for various UWSN applications,
Rongxin Zhu is with the School of Cyberspace Security, Hainan University,
Haikou 570228, China (e-mail: [email protected]). including secure localization, routing, data aggregation, mali-
Azzedine Boukerche is with the Paradise Research Laboratory, EECS, cious node detection, and reliable node selection [13], [14].
University of Ottawa, Ottawa, ON K1N 6N5, Canada (e-mail: boukerche@ Extensive research has focused on modeling the trust of sensor
site.uottwa.ca).
Libo Long is with the EECS, University of Ottawa, Ottawa, ON K1N 6N5, nodes in WSNs and Mobile Ad hoc Networks (MANETs).
Canada (e-mail: [email protected]). Trust is a vital factor in any network system as it dictates the
Qiuling Yang is with the School of Computer Science and Technology reliability, integrity, and security of the transmitted data and the
and the Innovation Platform for Academicians of Hainan Province, Hainan
University, Haikou 570228, China (e-mail: [email protected]). involved nodes [15]. However, the distinctive characteristics
Digital Object Identifier 10.1109/COMST.2024.3389728 of UWSNs render the direct application of these trust models
1553-877X 
c 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2548 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

unfeasible, due to the unique properties and limitations of contextual evidence, which play a pivotal role in the
underwater communication media. establishment of robust trust relationships.
In UWSNs, trust management entails establishing and main- • An overview of existing network security techniques in
taining trust relationships among underwater nodes responsible UWSNs is presented, alongside an examination of the
for data collection and processing [16]. These relationships specific challenges confronting trust management in this
are pivotal in ensuring the accuracy, confidentiality, and avail- domain. Trust-based applications ranging from secure
ability of collected data while preventing unauthorized access routing to identity authentication and intrusion detection
and malicious activities. Several factors contribute to the are discussed, exemplifying the multifaceted applicability
complexity of trust management in underwater environments. of trust management in UWSNs.
Firstly, underwater communication media are highly dynamic • We investigate various trust models suited for UWSNs of
and susceptible to challenges such as signal attenuation, recent years, categorizing and comparing approaches like
multi-path fading, and interference. These physical constraints weighted sum methods, logic-based techniques, probabil-
can impact data transmission’s reliability and timeliness, ity and statistics models, and machine learning paradigms
leading to potential data loss or corruption [17], [18], [19]. in terms of their strengths and limitations.
Secondly, the deployment and maintenance of underwater • Current issues and future challenges are highlighted,
devices and sensor nodes pose logistical and operational chal- including trust bootstrapping, updating trust values,
lenges, necessitating specialized equipment, power sources, addressing node mobility, and more. This provides valu-
and regular maintenance activities. Moreover, the underwater able insights into open research directions for UWSN
environment itself is hostile, with factors like high pressure, trust management.
low temperature, and corrosiveness affecting device durability In summary, this paper provides a comprehensive survey of
and performance [20], [21], [22]. Additionally, UWSNs may trust management in UWSNs, garnering valuable insights into
involve collaborations among different organizations or enti- this crucial network security mechanism. It aids in gaining an
ties, introducing concerns of trustworthiness, accountability, in-depth grasp of the concepts, techniques, applications, and
and privacy. In light of these challenges, effective trust man- research landscape of trust-based approaches for safeguarding
agement mechanisms are critical for the successful operation UWSNs. The article’s organizational structure is depicted in
and utilization of UWSNs. Designing and implementing such Fig. 1.
mechanisms necessitates an in-depth understanding of the Organization: The remaining sections of the survey are
underwater environment’s unique characteristics and limita- structured as follows. Section II provides an extensive review
tions, as well as the specific requirements and goals of UWSNs of related literature in the field. Section III presents an in-depth
applications. analysis of threats, attacks, security requirements, and existing
Contributions: This survey paper focuses on trust manage- security mechanisms in UWSNs. In Section IV, an overview
ment in UWSNs, which has been a subject of increasing of trust computing techniques is presented, followed by an
interest; however, dedicated surveys on this topic are some- exploration of trust concepts and the challenges associated
what limited. Unlike previous surveys, this paper provides with trust management in UWSNs. Trust-based applications
a comprehensive overview of the various trust management are discussed. Section V highlights the challenges faced by
approaches proposed for diverse UWSN applications. We UWSNs. A taxonomy is used to compare various state-of-the-
classify these approaches based on the techniques employed, art Trust Models (TMs) in Section VI. Moreover, Section VII
encompassing threat and attack analysis, fundamental trust delves into the prominent open issues and recent challenges
concepts and phases, security and trust management challenges concerning the implementation of trust-based systems in
in acoustic communications, a review of trust management UWSNs. Finally, Section VIII offers a conclusive summary of
approaches in underwater sensor networks, and a brief dis- this paper.
cussion on future directions for trust management in UWSNs.
The main contributions of this paper are:
• Based on the characteristics of UWSNs, this paper II. R ELATED W ORKS
comprehensively analyzes the security threats and attacks In the realm of trust management in various network
faced by UWSNs and proposes security requirements for environments, a considerable amount of research has been
UWSNs. conducted, as reflected in a range of comprehensive surveys.
• The pivotal role and advantages of trust management However, a glaring gap in the literature is the lack of focus
in ensuring UWSN security are elucidated. The critical on UWSNs, a domain with unique challenges stemming from
steps involved in the trust management process, includ- its operational environment characterized by limited energy
ing trust evidence collection, trustworthiness evaluation, resources, variable node mobility, and intricate underwater
trust propagation, and trust relationship maintenance, are communication dynamics. Our survey aims to fill this gap by
explored. delving deeply into trust management tailored specifically for
• The paper examines the distinctions between direct trust UWSNs.
and indirect trust, emphasizing the importance of har- Over the past few years, trust management research has
nessing both forms of evidence to ensure a resilient been conducted in various areas such as VANET, MANETs,
trust evaluation. Furthermore, it delves into prevalent HetNets, and WSNs [23]. Table I presents related surveys
trust metrics such as communication, reputation, and in the field. For instance, Hbaieb et al. [24], their work

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2549

Fig. 1. Organization structure of this paper.

delves into the intricacies of trust challenges and attacks se. These contributions are significant in contextualizing the
within the Internet of Vehicles (IoV). They establish a novel role and importance of trust management within the broader
taxonomy that categorizes and contrasts approaches to IoV landscape of network security and efficiency.
trust management, providing a foundational framework for this In summary, while these surveys offer rich perspectives
emerging field. Similarly, Wang et al. [25] survey trust models on trust management across various network technologies,
and the Quality of Trust (QoT) in Heterogeneous Networks there remains a noticeable gap in the literature regarding trust
(HetNets), enriching the understanding of trust dynamics in management specifically for UWSNs. By focusing exclusively
diverse network settings. Hussain et al. [26] pivot the focus on UWSNs, we provide insights into the unique challenges
to Vehicular Ad hoc Networks (VANETs), offering insights and requirements of trust management in this domain. Our
into trust establishment and management methods crucial for survey highlights the need for specialized trust management
vehicular communications. strategies in UWSNs, considering the specific operational
In the domain of the Internet of Things (IoT), constraints and environmental factors these networks face.
Arshad et al. [27] present a comprehensive survey on Furthermore, we offer future research directions, emphasizing
Blockchain-based decentralized trust management systems. the development of innovative trust management solutions that
Their exploration emphasizes the requirements and challenges can effectively address the complex and dynamic nature of
of implementing trust systems in IoT environments. This underwater environments. Through this comprehensive exam-
is complemented by the work of Din et al. [28], who ination, our survey aims to provide a valuable resource for
provide an extensive survey on trust management techniques researchers and practitioners in the field of UWSNs, fostering
specifically tailored for IoT, shedding light on various trust further exploration and development in this vital area of study.
management approaches and their applicability in IoT con-
texts. Ahmed et al. [29], on the other hand, delve into the
fundamentals of trust and reputation within IoT, presenting a III. ATTACKS AND S ECURITY R EQUIREMENTS IN UWSN S
taxonomy and addressing open research challenges. In conventional computer networks, mobile networks, and
The literature also includes focused studies on trust applica- WSNs, security threats, and attacks pose significant risks,
tions in wireless communication environments, as evidenced ranging from unauthorized access and data manipulation to
by the works of Khalid et al. [30], Han et al. [31], and network disruption [39], [40], [41]. To safeguard the reliability
Ishmanov and Bin Zikria [32]. These surveys discuss various and security of these networks, it is essential to analyze and
trust applications, ranging from WSNs to MANETs, and offer research various security threats and develop corresponding
comparative analyses of trust models and their applications. strategies and measures for protection.
Moreover, the works of Jiang et al. [8] and UWSNs are rapidly evolving networks that leverage IoT
El-Rabaie et al. [33] provide valuable insights into security technology in underwater environments [42]. They con-
issues, localization, and routing protocols within UWSNs, nect diverse sensors, controllers, communication devices,
although they do not delve deeply into trust models per and underwater equipment, enabling monitoring, control, and

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2550 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE I
C OMPARISON OF R ELATED S URVEYS

data transmission in these challenging environments with information. This amplifies the difficulty of defense and
broad applications [43]. However, the unique characteris- increases security risks for UWSNs.
tics of underwater environments introduce specific security With the growing adoption of UWSNs and the deploy-
and trust issues for UWSNs. Signals in water are sus- ment of numerous underwater devices and sensors, ensuring
ceptible to interference, and attackers can easily remain data protection and the normal operation of these networks
concealed, exploiting vulnerabilities in UWSNs to interfere becomes increasingly crucial. In this section, we com-
with data transmission, disrupt operations, and obtain sensitive prehensively analyze and summarize the various types

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2551

and characteristics of attacks that UWSNs may encounter. • Spoofing: Impersonation of legitimate network nodes
Additionally, we explore the specific requirements for risks the theft of sensitive information and may cause
ensuring security and data protection in underwater envi- unauthorized access or disruptions in UWSN communi-
ronments [44], [45]. We investigate the existing solutions cations.
available for UWSNs to ensure reliability and security, consid- • Sybil Attack: Creation of multiple false identities in
ering the challenges posed by underwater communication and the network exhausts resources, leads to unfair resource
the dynamic nature of the underwater environment. Finally, distribution, and undermines the integrity of UWSN
we address the current challenges faced by UWSNs security, operations.
providing insights for future research and development efforts • Sinkhole Attack: Misdirecting traffic to compromised
in this domain. nodes intercepts and potentially alters data, causing ser-
vice interruptions and routing malfunctions.
• Wormhole Attack: Creating illicit channels for data packet
A. Attack Model interception breaches data security and facilitates data
The primary objective of network security is to safe- tampering and replay attacks in UWSNs [51].
guard the network from various forms of attacks [38], [46]. • On-Off Attack: Irregular traffic patterns disrupt service,
Similarly, UWSNs confront multiple security and trust leading to intermittent service interruptions and resource
challenges due to the unique nature of the underwater envi- wastage in UWSNs.
ronment [47], [48]. In this context, underwater signals are • Bad Mouthing Attack: Spreading false negative
susceptible to interference, and attackers can easily conceal information damages node reputations, leading to
their presence, providing them with opportunities to exploit mistrust and interference in UWSN operations.
UWSN vulnerabilities and disrupt operations, as well as • Conflicting Behavior Attack: Contradictory behavior by
obtain sensitive information. Consequently, defending UWSNs multiple false entities complicates trust assessments, lead-
and mitigating security risks become more complex. In ing to network tampering and unreliable node behavior
trust management for UWSNs, various schemes have been analysis.
established to identify and address the behavior of mis- • Ballot Stuffing Attack: Malicious nodes artificially inflate
behaving nodes, which can be either selfish or malicious. their trust or reputation scores, leading to undeserved
Below, we outline a selection of specific attack instances, high trust levels and potential misallocation of network
which provides a snapshot of common challenges in trust resources in UWSNs [52].
management. • Collusion Attack: Multiple nodes collaborate to dis-
• Interference and Eavesdropping: Unwanted signal over- rupt network operations or manipulate trust evaluations,
lap and unauthorized signal interception, leading to resulting in distorted trust metrics in UWSNs [10].
unreliable data transmission and potential compromise of • Newcomer Attack: Newly joined nodes behave mali-
confidential information in UWSNs. ciously from the outset, exploiting initial trust and leading
• Physical Tampering: Direct physical interference or dam- to immediate network vulnerabilities.
age to UWSN devices disrupts network topology and Table II provides an overview of different attacks, their
sensor functionality, causing data collection gaps and characteristics, and possible mitigation strategies across dif-
network vulnerabilities [49]. ferent layers of UWSNs. These attacks can occur from the
• Sniffing Attack: Utilization of tools to capture data in physical layer to the application layer, often involving data
transit, compromising data privacy and leading to poten- tampering, information theft, and network manipulation, such
tial leakage of sensitive environmental or operational as Greyhole, Spoofing, and Sybil attacks. Some attacks possess
information in UWSNs. specific characteristics, for instance, Denial-of-Service (DoS)
• MAC Flooding: Overloading the network with invalid attacks that overload the target node with excessive data traffic
MAC addresses causes resource exhaustion, disrupting or frequent service requests, disrupting the node’s normal
normal operations and potentially causing system crashes. operation [53]. Table II also highlights trust attacks, which
• DoS Attack: Overwhelming network traffic causing exploit gained trust relationships within the network to deceive
system overload renders network services unavailable, victims and conduct malicious activities. Trust attacks aim
depletes bandwidth, and exhausts resources. to undermine established trust relationships, leading victims
• Selective Forwarding Attack: Selective packet forwarding to make incorrect decisions. An example of a trust attack is
disrupts network communication, resulting in incomplete the On-off attack, wherein attackers alternate between sending
data transmission, increased latency, and potential con- and not sending data packets using normal communication
gestion in UWSNs [31]. patterns to gain the victim’s trust and carry out subsequent
• Blackhole Attack: False routing claims leading to packet attacks [54]. Due to the attackers’ ability to mimic normal
drops result in significant data loss, network congestion, communication patterns, this type of attack is challenging to
and routing dysfunction in UWSNs [50]. detect. To counter such internal attacks, intrusion tolerance
• Greyhole Attack: Selective packet dropping by posing as mechanisms are effective solutions. These mechanisms enable
a reliable node causes data loss and incompleteness, mis- systems to maintain functionality and service availability even
leading network routing, and increasing communication in the presence of attacks. Trust mechanisms, as integral com-
delays in UWSNs. ponents of intrusion tolerance mechanisms, establish reliable

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2552 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE II
ATTACK D ESCRIPTION

evaluation and control mechanisms to limit malicious entities’ the system and network. Security mechanisms encompass
activities within the network. Therefore, trust mechanisms various technologies and measures aimed at fulfilling these
are often more reliable and effective than traditional security security requirements and safeguarding information systems
systems when dealing with internal attacks [55]. and networks by preserving confidentiality, integrity, and
availability. In essence, ensuring the security of UWSNs
is a fundamental system requirement, and comprehending
B. Security Requirements and Security Mechanisms and specifying the security requirements lay the foundation
The increasing and evolving attack methods pose a serious for realizing effective security mechanisms. By implement-
threat to UWSNs. Therefore, a comprehensive understanding ing and utilizing these security mechanisms, the security
of the network’s attack patterns and characteristics is crucial of information systems and networks can be effectively
for formulating appropriate security requirements and mecha- ensured.
nisms to ensure the UWSNs’ security and reliability. Security 1) Security Requirements: Distinct attack types in UWSNs
requirements encompass the essential conditions to safeguard give rise to specific security imperatives. For instance, coun-
the security of information systems and networks, representing teracting data tampering mandates a focus on data integrity
the prerequisites for achieving security objectives [56]. In the and confidentiality. Dos attacks require robust measures for
pursuit of meeting these security requirements, a set of security high availability and resilience. Following are the summarised
mechanisms must be employed to ensure the protection of general security requirements for the network:

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2553

• Confidentiality: Ensuring the confidentiality of data mechanisms to safeguard networks. Adapting measures to
and communications in UWSNs to prevent unautho- address varied demands empowers designers and operators
rized access to sensitive information. This requires the to enhance security, counter diverse threats, and maintain
implementation of appropriate encryption and access network integrity, reliability, and trustworthiness amid evolv-
control measures to prevent data leakage or unauthorized ing security landscapes.
acquisition. 2) Security Mechanisms: Network security techniques are
• Integrity: Protecting data in UWSNs from tampering, essential safeguards ensuring data integrity and confidentiality,
modification, or corruption. Ensuring the integrity of regulating network access, and countering threats to forestall
data during transmission and storage processes to prevent unauthorized breaches. Incorporating technical and procedural
arbitrary alterations or damage to data integrity [49]. measures, these strategies uphold security requisites, curtail
• Availability: Safeguarding the continuous availability and risks, and safeguard systems, networks, and data against
stability of UWSNs, ensuring that the system can provide illicit tampering, unauthorized entry, or compromise. Notable
services on demand, and preventing system unavailability existing security mechanisms in this domain include:
or disruption caused by attacks, failures, or other factors. • Access Control Mechanisms: These mechanisms restrict
• Authentication: Ensuring reliable authentication of users access to system resources, ensuring that only autho-
and devices in UWSNs, preventing unauthorized entities rized users or devices can utilize specific functions or
from accessing system resources or engaging in malicious access sensitive data. By implementing access control,
activities. Common practices include using passwords, potential security vulnerabilities and unnecessary risks
digital certificates, biometric features, and other methods are reduced [60].
for identity verification. • Identity Authentication Mechanisms: These mechanisms
• Access control: Restricting access to system resources verify the identity of users or devices, ensuring that
and functionalities, ensuring that only authorized users only legitimate entities can access system resources.
or devices can perform specific operations or access Common methods include usernames/passwords, digital
sensitive data [57]. Employing appropriate access con- certificates, or two-factor authentication.
trol policies and permission management mechanisms • Encryption Mechanisms: Encryption transforms sensitive
to prevent unauthorized access and potential security data into an unreadable form, preventing unauthorized
vulnerabilities. access. Legitimate users or devices need to decrypt the
• Security auditing and monitoring: Establishing effective data using keys to access the original information.
security auditing and monitoring mechanisms to record • Security Protocols: Security protocols like SSL and TLS
security events and activities in the system, promptly ensure the confidentiality and integrity of data trans-
detecting and responding to potential security threats. mission during communication. These protocols employ
This includes techniques such as logging, anomaly detec- encryption and authentication mechanisms to prevent data
tion, and intrusion detection [58]. leakage or tampering.
• Scalability and recovery capabilities: UWSNs need to • Intrusion Detection and Defense Mechanisms: These
possess a certain level of resilience against attacks, being mechanisms monitor network traffic and system behavior
able to identify and withstand various network attacks to promptly detect potential intrusion activities and imple-
and security threats [59]. Additionally, the system should ment appropriate defense measures to prevent or mitigate
have fast recovery and restoration capabilities to address attacks.
service disruptions or data compromises resulting from • Trust Mechanisms: Trust models and trust evaluation
attacks or failures. mechanisms monitor the behavior of nodes and users
These security prerequisites span various dimensions. within the system to prevent internal attacks and misuse
They encompass safeguarding data confidentiality, upholding of privileges. Trust mechanisms aid in identifying suspi-
data integrity, ensuring uninterrupted availability, establishing cious behavior and taking timely action.
robust authentication protocols, enforcing effective access con- Recently, UWSNs have emerged as an essential domain
trols, enabling vigilant security auditing and monitoring, and that integrates various underwater devices, sensors, and
fostering resilience with recovery capacities against attacks. systems for data collection, transmission, and processing.
They are meticulously designed to ensure the overarching In this context, ensuring secure and trustworthy commu-
security and credibility of the network, safeguarding both nication and data exchange between devices has become
the system and its constituent devices from potential security an increasingly important requirement [61], [62]. Trust, as
vulnerabilities. An illustrative instance of a trust attack, such a fundamental element of IoUT security, entails confidence
as the on-off attack, can engender periodic behavioral patterns in the security and reliability of communication and data
in a node, thereby deceiving the trust management mechanism exchange among underwater devices. In summary, trust plays
into attributing unwarranted trustworthiness [54]. Moreover, a pivotal role in ensuring the security and reliability of
addressing malevolent attacks entails the pivotal roles of trust UWSNs.
management and intrusion detection mechanisms for system
safeguarding.
Each attack type presents distinct challenges, necessitat- IV. OVERVIEW OF T RUST
ing targeted security measures. A thorough grasp of these This section introduces the relevant concepts of trust,
requirements aids in formulating and applying fitting security including the definition and characteristics of trust. It then
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2554 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

contextually relevant trust evaluation in UWSNs, essential for

the network’s security and reliability. Moreover, [64] suggests
categorizing trust evidence into data-based, link-based, and
node-based trust evidence. The trust metrics will be further
introduced in Section IV-C.
2) Trust Evaluation: After trust evidence collection, the
next phase is trust evaluation, where trust models and trust
measurement methods are employed to assess and quantify the
collected evidence. Trust models define the concept of trust
and the calculation methods, while trust measurement methods
convert evidence into quantifiable trust values or indicators. By
considering different types of evidence and their corresponding
weights, an evaluation result of the trust level towards nodes
or entities can be derived. A detailed classification of trust
models will be discussed in Section VI-C.
3) Trust Propagation: Trust propagation is a crucial phase
in trust management, involving the establishment and dissem-
ination of trust relationships among nodes based on generated
relationships and pre-existing trustworthiness values, including
collaboration recommendations [51], [65]. By implementing
Fig. 2. Process of trust management. reliable communication and cooperation mechanisms, nodes
can mutually verify and build trust, leading to the formation
of a broader trust network. Trust transitivity and trust fusion
introduces the various trust evidence and different types of play fundamental roles in trust propagation. This module
trust-based applications. offers key benefits, such as mitigating computational costs by
propagating measured trust values over the network instead
A. Trust Computing of determining individual entity trust. Additionally, nodes
In communication systems, trust management is a vital with a globally established trust may wield better influence
process aimed at ensuring the security, stability, and reliability over services. Trust propagation can be categorized into two
of the system. It involves the implementation of various tech- approaches: centralized and distributed [66]. These approaches
niques and strategies to assess and propagate trust among the provide different perspectives and considerations in managing
participating devices. Trust management is a comprehensive trust dissemination within the network.
and dynamic process, encompassing the following essential Centralized Propagation: Centralized trust propagation
components: trust evidence collection, trust evaluation, trust involves a central node assuming a brokerage role, reaping
propagation, and trust maintenance, as illustrated in Figure 2. advantages, and fostering trust propagation by receiving and
1) Collecting Trust Metrics: The first step in trust man- distributing trust information among nodes. In this type of
agement is trust evidence collection, which involves gathering model, a globally trusted node gathers and shares trust val-
both direct and indirect evidence to assess the trustworthi- ues with other nodes. Some trust models adopt centralized
ness of nodes or entities. In addition to collecting direct trust propagation, as evidenced by the proposed models
and indirect trust evidence, it is essential to consider over- in [67], [68]. This centralized approach offers advantages in
all trust [63] to enhance communication security. Direct terms of trust consolidation and efficient trust dissemination,
evidence includes historical interaction records, behavioral but it also introduces potential single points of failure and
evaluations, and reputation ratings, while indirect evidence security concerns.
comprises recommendations from other nodes or entities and Distributed Propagation: Distributed trust propagation
trust chains. Trust formation in UWSNs is the culmination eliminates central coordination, allowing nodes to interact
of the aforementioned processes. It involves synthesizing directly or via referrals, fostering interactions between them.
trust information from various sources to form a coherent This approach includes transitivity-based and consensus-based
understanding of each node’s trustworthiness. This process methods [69]. Transitivity, a fundamental concept, is often
is particularly challenging in UWSNs due to the network’s seen as a basic trait in trust propagation, derived from classical
dynamic nature and the acoustic channel’s unreliability. Trust social and cognitive models. It characterizes the property of
Formation involves computing the overall trust value and can relationships between actors. For instance, if node A trusts
be either single-trust, focusing on a solitary metric like end- node B and node B trusts node C, based on transitivity,
to-end delay, or multi-trust, which considers multiple aspects node A can infer indirect trust in node C. Such transitive
like intimacy, competence, and honesty. This formation utilizes trust can be sequentially expanded [70]. On the other hand,
threshold-based metrics, weighted sums, and trust scaling consensus-based propagation aggregates trust evaluations on a
by confidence to dynamically adjust trust levels based on target node from multiple nodes to form a group consensus
context, environmental conditions, and entity performance. and disseminate it through the network. This mechanism is
This comprehensive approach ensures a nuanced, adaptive, and widely recognized for fostering trust and cooperation among

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2555

peers in behavior-oriented modeling [71]. While distributed essential property of trust in UWSNs, where older evidence
propagation offers advantages like resilience to failures and carries less weight than recent evidence. Additionally, rein-
adaptability to dynamic networks, challenges can arise in forcement learning-based trust updates have been employed
maintaining consistency and addressing trust disparities among to address hybrid attack problems and improve resource uti-
nodes. lization [55]. However, trust updating based on environmental
While central trust propagation is viable in simple networks, factors may reduce the system’s ability to quickly respond
it may not suit distributed sensor node networks due to to malicious nodes. In clustered networks, agent nodes (e.g.,
security and performance concerns like single points of failure. cluster heads) may maintain and update trust values instead
Distributed and decentralized trust propagation doesn’t fully of individual nodes, as demonstrated by the agent-based
replace central approaches but rather empowers distributed trust and reputation management scheme (ATRM) [74]. Each
propagation with increased responsibility and expertise in trust sensor in ATRM assumes a trusted authority is responsible
management and evaluation. This shift towards decentraliza- for generating mobile agents that perform trust collection
tion offers potential solutions to address the limitations of and update functions. However, the trustworthiness of mobile
centralized methods in more complex and dynamic network agents introduces security risks to the network.
environments. In summary, underwater acoustic networks may introduce
4) Trust Update: In UWSNs, the dynamic nature of trust multiple mechanisms to integrate direct and indirect obser-
relationships necessitates an adaptable and responsive trust vation and verification methods, enhancing the accuracy and
update mechanism. This mechanism is essential for incor- reliability of trust updates and improving the network’s com-
porating new observations and recommendations into the plexity and security.
trust evaluation process. Common trust update strategies in 5) Trust Redemption: Trust redemption mechanisms allow
UWSNs generally follow either an event-driven or time-driven nodes to regain trust after their trust level has dropped due
approach [64]: to misbehavior. In UWSNs, there are cases where normal
Event-Driven Trust Update: This approach updates trust nodes are misclassified as malicious nodes, especially given
data following specific transactions or events. It is reactive the disruptive nature of the acoustic channel, where various
in nature, where trust evaluations are revised post-service factors can cause performance fluctuations. Trust redemption
delivery based on the quality of the service rendered. Feedback approaches, such as those proposed in [75], consider historical
on service quality is typically collected at each participating performance and environmental influences, including unreli-
entity or aggregated at an IoT cloud server. In certain UWSN able acoustic channels and weak link connectivity, to identify
environments, recommendations are exchanged on-demand, instances of misclassification and restore trust. Additionally,
with entities collaborating to share their assessments of others’ time-based and behavior-based trust redemption strategies
behavior [72]. This method ensures that trust evaluations are have been introduced to monitor and adjust trust relationships,
promptly updated in response to recent interactions, maintain- implement risk management measures, and restore trust [55].
ing the relevance and accuracy of the trust data. By continuously updating and maintaining trust relationships,
Time-Driven Trust Update: Contrastingly, the time-driven the trust state in the system can adapt to potential changes
approach involves the periodic collection of trust evidence, and challenges, ensuring the overall robustness and stability
including self-observations and recommendations. During of UWSNs.
periods of inactivity or lack of evidence, trust values may
decay over time, emphasizing the higher reliability of recent
information compared to older data [73]. This approach often B. Related Concepts of Trust
employs an exponential decay function to align with the spe- Correctly understanding the concepts related to trust is
cific needs of the application. It ensures that trust evaluations crucial for comprehending and designing effective trust man-
remain current and reflective of the most recent network agement mechanisms [76], [77]. Trust, being a subjective
conditions and interactions. The time-driven approach, is concept, may have diverse definitions and characteristics in
particularly suited to environments where regular updates different environments and contexts. Therefore, it is essential
are essential for maintaining accurate and up-to-date trust to establish a clear and common understanding of trust
assessments. concepts and characteristics to develop appropriate trust man-
Both approaches have their unique advantages and are agement mechanisms, thereby enhancing the security and
chosen based on the specific requirements and characteristics reliability of the network.
of the UWSN environment. The event-driven approach is 1) Notion of Trust: Numerous researchers have proposed
ideal for networks with frequent interactions and transactions, their own interpretations of trust definitions in trust man-
while the time-driven approach is beneficial in scenarios where agement mechanisms. For instance, some studies have
consistent periodic evaluations are necessary. In UWSNs, the distinguished between individual and system-level trust [11],
choice of trust update strategy plays a critical role in ensuring while others have classified trust as entity trust and data
that trust evaluations accurately reflect the current state and trust [26]. Moreover, researchers have explored node trust,
dynamics of the network. path trust, and service trust from various perspectives related
Various research studies propose different trust update to monitoring node behavior, path connectivity, and service
mechanisms, such as the ARTMM [69], which uses a sliding availability [38]. While trust possesses certain universality
time window concept to update trust values. Time-decay is an and common characteristics, its definition may vary based

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2556 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE III
C OMPARISON OF T RUST AND R EPUTATION

on the research and application perspectives in different conflation of terms underscores the interconnected nature of
disciplines. Hence, defining and understanding trust can incor- these concepts, yet it also necessitates a clear demarcation for
porate discipline-specific considerations. To address this, [78] precise academic discourse and application within the field.
examined trust definitions across various disciplinary fields In summary, while trust is a subjective and context-
and proposed a trust definition applicable to the field of dependent metric based on direct interactions between two
network communication, drawing upon insights from other entities, reputation is a more objective and aggregated measure
domains. According to this definition, trust entails beliefs or of an entity’s trustworthiness as perceived by a community
expectations concerning the reliability and integrity exhibited or group [82]. Both concepts are integral to the design and
by an entity or system in a particular situation. implementation of secure and reliable network systems, each
In this paper, trust is defined as a subjective belief held serving different yet complementary roles in trust management
by a trustor (entity that trusts) about the trustworthiness or frameworks [78]. For example, in a trust management mecha-
reliability of a trustee (entity being trusted) based on past nism for UASNs, a sensor node may possess a good reputation,
interactions and experiences. It is context-specific and can be but in a specific task, other nodes may be more trustworthy.
directly related to the trustor’s willingness to depend on the Thus, reputation can be considered a prerequisite for trust,
trustee in a given situation. Trust in a network setting is often but it does not encompass the entirety of trust. Trust requires
based on the perceived capability, reliability, and intention of a more comprehensive evaluation and may vary in different
the trustee. environments. Specifically, in the routing context, reputation
2) Reputation: Reputation, on the other hand, is an aggre- management often aggregates trust opinions for a global
gated perception held by a group or community about an value. Thus, crafting a reputation-based routing protocol can
entity’s trustworthiness. It is formed based on collective expe- be intricate without incorporating trust management, which
riences and assessments of multiple entities [79]. Reputation lacks a standardized definition [83]. To overcome this, we
is more of a global or community-wide metric that reflects employ trust-based management, encompassing both trust and
the general opinion or view about an entity’s behavior and reputation elements.
trustworthiness. Within reputation models, reputation repre-
sents a subjective anticipation of a node’s behavior formed C. Trust Metrics
through past interactions. In this context, nodes can act as Trust models in communication systems are often based
trustors, seeking opinions, and trustees, providing aggregated on the experiences nodes have with interacting entities. Trust
opinions (reputation). Reputation calculation considers diverse evidence originates from either the nodes’ direct observations
data, like collaboration frequency between nodes [80]. UWSNs or from other nodes (indirect trust or recommendations), or a
utilize reputation systems to enable node rating, nurturing trust combination of both.
establishment, and supporting reliable inter-node communica- Direct Trust refers to evidence that can be directly observed
tion through collective opinions, fostering trust among network or verified. Several common types of direct evidence include:
components. 1) Identity Evidence: This involves verifying the identity of
3) Differences Between Trust and Reputation: In the con- communication entities using digital certificates, identity
text of trust management mechanisms, the distinctions between authentication mechanisms, and access control lists.
trust and reputation are elucidated in Tab. III. While these 2) Authentication Evidence: This verifies the claimed iden-
concepts are frequently interlinked within trust management tity of communication entities through methods such as
frameworks, it is essential to recognize that they are not passwords, digital signatures, and tokens [84], [85].
identical. Notably, in certain scholarly works, reputation man- 3) Temporal Evidence: It involves verifying the sequence
agement is considered a subset of the broader domain of trust and timing of actions performed by communication
management [81]. Additionally, the terminology surround- entities using timestamps and event sequences.
ing trust management and trust establishment is often used 4) Location Evidence: This provides evidence of the
interchangeably, as highlighted in several studies [78]. This location of communication entities using data from

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2557

Fig. 3. Trust model based on indirect and direct trust evidence.

the Global Positioning System (GPS), geolocation large-scale networks, obtaining direct experiences between all
information, and IP addresses. nodes may be challenging. Hence, a hybrid approach that
5) Energy Evidence: The trustworthiness of communica- aggregates trust opinions through recommendations (direct
tion entities is evaluated by monitoring their energy and indirect) remains a fundamental building block of suc-
consumption, such as power consumption and battery cessful trust models [89]. Combining direct and indirect
life [86]. evidence allows for a more comprehensive evaluation of the
6) Communication Evidence: This evaluates the trustwor- trustworthiness of communication entities, leading to better
thiness of communication entities by monitoring their decision-making and ensuring the reliability, integrity, and
communication behavior and records, such as call logs, security of communication systems.
data transfer logs, and network traffic. Figure 3 depicts the process of determining the trustworthi-
While direct trust models may work well in small self- ness of a node based on both direct and indirect evidence. A
organized environments with limited applications, such as node’s trust in another node is influenced by its trust in the
military surveillance, they have a short lifespan and may node itself, which encompasses considerations of potentially
not be suitable for large-scale networks where comprehensive malicious behavior, as well as its trust in the behavior exhibited
knowledge of the overall trust level is lacking [87]. Therefore, by the target node. The latter includes various factors such as
indirect trust is introduced in trust-based models. communication behavior, energy consumption, environmental
Indirect Trust is obtained through inference or indirect conditions, and channel quality. The Figure also illustrates the
observation and cannot be directly verified. Common types of establishment of indirect trust between nodes through trust
indirect evidence include: chains or trust referrals, also known as transitive chains.
1) Reputation Evidence: This assesses the trustworthiness In UWSNs, hybrid methods integrating direct and indirect
and reputation of communication entities based on eval- trust aggregation are often employed. Crafting a distributed
uations and feedback from other entities. trust model for UWSNs involves merging the target node’s
2) Behavioral Evidence: It evaluates the trustworthiness direct trust with third-party nodes’ indirect trust, bolstering
of communication entities based on their behavioral network-wide trustworthiness. Existing trust models encom-
patterns and characteristics, such as communication pass diverse trust evidence forms, detailed in Table IV.
frequency, data transfer patterns, and data access pat-
terns [32].
3) Context Evidence: This assesses the trustworthiness D. Characteristics of Trust
of communication entities by considering contextual The concept and application of trust exhibit variations in
information, such as environmental conditions, network different disciplinary fields, resulting in diverse characteristics
topology, and communication relationships. of trust. Trust is a multifaceted notion with various dimen-
4) Security Log Evidence: It analyzes security logs and sions and attributes. Several common characteristics of trust
audit logs to detect abnormal behavior and security observed across different networks include:
events, thereby assessing the trustworthiness of commu- • Multidimensionality: Trust encompasses various aspects,
nication entities [88]. such as the relationship between the trustor and trustee,
Indirect trust aggregation often relies on node interac- the objectives of the trust, and the trust-building
tions to evolve the global trust value of an agent. In process [94].

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2558 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE IV
S UMMARY OF T RUST E VIDENCE

• Dynamism: Trust levels are subject to change over time Additionally, trust often faces skepticism and potential
as interactions and experiences between entities evolve. attacks, such as Sybil attacks, collusion attacks, switch attacks,
• Transitivity: Trust can be transferred or shared among dif- and oscillation attacks, which introduce risks to trust-based
ferent entities in a network, allowing for trust propagation decision-making. In UWSNs, these risks involve data integrity
through interconnected nodes. and reliability, communication link disruptions, node failures,
• Uncertainty: The establishment and maintenance of trust as well as information security and privacy risks. The unique
involve uncertain factors and risks, which can lead to the characteristics of the underwater environment, including prop-
breakdown of trust relationships or fluctuations in trust agation properties, energy constraints, and node mobility,
values [8]. present distinct challenges and uncertainties in establishing
• Mutuality: Trust operates as a mutual relationship, and maintaining trust. The interplay of these risk factors can
wherein mutual trust between two entities relies on each result in the breakdown of trust relationships, compromised
other’s behavior and decisions, rather than unidirectional data integrity, information leakage, and malicious attacks,
trust. thereby impacting the overall operation and reliability of the
• Subjectivity: The formation of trust is influenced by network.
subjective factors, incorporating emotions, beliefs, and
social perceptions, in addition to objective facts and data.
• Context Dependency: Trust is context-dependent, shaped E. Trust-Based Applications
by specific contextual conditions, and is not a static or The trust model has found wide-ranging applications in
universal construct [29]. various systems, giving rise to numerous trust-based applica-
In the context of UWSNs, trust management mechanisms tions such as security routing, identity authentication, intrusion
exhibit some specific features due to the distinctive nature of detection, access control, and key management.
the underwater environment and nodes. These include: 1) Security Routing: Trust management is vital for ensur-
• Long Latency and Unreliable Communication: The ing secure data transmission in networks by establishing
underwater medium, water, presents high signal absorp- trusted paths between nodes [95], [96], [97]. Key security
tion and scattering characteristics, leading to long functions in UWSNs encompass secure routing, malicious
communication latency and unreliable transmission. node detection, and addressing attacks. Trust-based routing
• Hardware and Software Reliability: Underwater devices protocols mitigate malicious node impact, and two strategies
operate in harsh conditions, facing challenges such as exist end-to-end and hop-by-hop. End-to-end routing selects
water pressure and temperature, resulting in the need forwarding nodes per packet, while hop-by-hop relies on
for robust hardware and software solutions to ensure each node’s trust level. Trust relationships identify potential
reliability [46]. threats, aiding secure routing and data protection through the
• Energy Constraints: Underwater devices primarily rely evaluation of indicators like reputation and behavior.
on limited battery power, necessitating energy-efficient 2) Identity Authentication: Trust management is also appli-
communication and computation techniques to prolong cable in identity authentication systems to assess the
battery life. trustworthiness of users or entities. Message authentication
• Diversity and Heterogeneity: UWSNs comprise various algorithms have been proposed to address communication
types of devices and sensors with diverse data for- security issues in underwater wireless acoustic networks,
mats, communication protocols, and security capabilities, which are susceptible to forged messages [98]. Converging
requiring careful consideration of their heterogeneity in nodes utilize belief messages from trusted nodes for authen-
trust mechanisms. tication decisions, effectively detecting forged messages. By

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2559

Fig. 4. Network model.

considering multiple factors comprehensively, a trust evalu- trust level of users or entities, enabling precise access
ation model is established to verify and authorize identities, control.
reducing the risks of identity forgery and unauthorized access. 5) Key Management: In the context of key management,
Identity authentication protocols distinguish legitimate trans- trust management primarily involves key distribution and
missions from eavesdropping in the physical layer security verification processes. Fully self-organized public key manage-
framework of underwater acoustic networks using machine ment systems have been proposed for mobile ad hoc networks,
learning techniques to extract feature statistics and assess the eliminating the need for online access to trusted authorities
authenticity of received data packets. or centralized servers [99]. The trust model and encryption
3) Intrusion Detection: Trust management is instrumental algorithms are utilized to ensure the security and integrity
in intrusion detection to create normal behavior models and of keys, detecting and intercepting potential malicious attacks
identify deviations that indicate abnormal behaviors. Trust- while maintaining communication confidentiality and integrity.
based malicious node detection schemes have been proposed Multi-part trust-based public key management approaches in
for underwater acoustic sensor networks to quantify the mobile ad hoc networks enhance security by establishing trust
impact of the underwater environment on communication relationships between neighboring nodes for verifying and
trust, effectively detecting abnormal activities. Monitoring and authenticating generated keys, without relying on third-party
analyzing network and system activities, and evaluating their trusted authorities.
trustworthiness, allows trust management to identify abnormal
behaviors, vulnerabilities, and malicious activities, leading to
timely response measures to ensure system security. V. C HALLENGES FOR UWSN S
4) Access Control: Trust management is employed in In this section, we present the overall architecture of
access control to implement fine-grained access control poli- UWSNs, the characteristics of underwater communication, and
cies, determining whether a user or entity is authorized to the security challenges of UWSNs.
access resources or perform specific operations. A network
architecture using steganography overlays and an explicit
zero-trust approach is proposed, where relationships between A. Overview of UWSNs
trust levels, roles, and permissions are established to ensure As depicted in Figure 4, an Underwater Wireless Sensor
that only authorized users or entities can access specific Network (UWSN) typically consists of network nodes
resources or perform specific operations. Trust management equipped with acoustic communication capabilities. These
assesses the legitimacy of access requests based on the nodes may be fixedly anchored on the sea floor, suspended, or

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2560 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

floating in the water, leading to random drifts with the under- high-frequency signals, which reduces signal coverage
water currents [100], [101], [102]. Moreover, mobile network and network bandwidth.
nodes, such as Autonomous Underwater Vehicles (AUVs), • Limited Frequency Selection: The available frequency
might be employed to dynamically collect data from sensors range for acoustic signals is narrow, typically between
deployed in various underwater regions [103], [104], [105]. 10Hz and 100kHz. This limited bandwidth poses con-
Acoustic links serve as the primary communication mode straints on system performance.
between these nodes, although occasional usage of cables • Environmental Sensitivity: Acoustic signal transmission is
might occur, such as linking the cluster head of a UWSN to highly susceptible to environmental factors like temper-
a surface gateway. ature, salinity, and noise in the water, leading to signal
Surface nodes within the UWSN facilitate connections attenuation, time-varying multipath effects, and backscat-
to different gateway types, including nautical gateways tering, all of which affect communication quality [116].
(e.g., a ship), terrestrial gateways (e.g., a base sta- • Time-Varying Phase Changes: Changing relative posi-
tion), and even space gateways (e.g., a high platform tions of sensors and signal sources cause phase changes in
or a satellite) [106], [107], [108], which further link the the received signal, introducing time-varying frequency-
UWSNs to terrestrial backbone networks like the Internet. selective effects that impact system performance.
Communication between surface nodes and gateways utilizes • Severe Multi-Path Effects: Acoustic signals undergo
radio links, while optical fibers are commonly employed reflection and scattering in water, resulting in severe
for connectivity between terrestrial gateways and backbone multi-path effects that cause signal fading, phase rotation,
networks. The distinctive characteristics of UWSNs present and delay spread, ultimately reducing communication
several challenges in ensuring their security, which will be quality [117].
elaborated upon after providing an introduction to UWSNs. In conclusion, while acoustic transmission technology offers
advantages such as low cost and low power consumption,
its distinctive characteristics give rise to significant technical
B. Characteristics of Underwater Communication difficulties and implementation challenges [118]. Particularly,
Acoustic transmission plays a crucial role in underwa- the deployment of nodes in unattended and harsh underwater
ter networks, and its theoretical research and technological environments makes them vulnerable to failure or voluntary
development hold significant potential for further advance- attacks. To achieve efficient, secure, and reliable acoustic
ment [109]. While acoustic communications in underwater communication, technologies like coding, equalization, array
environments pose numerous challenges, they are widely processing, and tracking must be employed to suppress adverse
acknowledged as the most viable medium compared to optical effects and enhance network performance. This calls for
and radio frequency (RF) for enabling long-distance communi- comprehensive consideration of channel characteristics and
cation due to their lesser signal attenuation and ability to travel communication requirements in UWSNs.
further distances [110], [111]. The initial implementation of
UWSNs with RF and optical links highlighted the necessity for
new solutions and approaches tailored to the underwater envi- C. Challenges for UWSNs Security
ronment, which presents distinct challenges and limitations Security in UWSNs is significantly impacted by the dis-
concerning signal propagation, radio wave efficiency, limited tinct characteristics of the underwater network environments.
transmission range, and optical wave scattering [112], [113]. Firstly, the deployment of underwater nodes in unattended and
Anticipated advancements in available bandwidth and tech- hostile surroundings renders them susceptible to compromise
nological progress are expected to significantly enhance and intrusion, as implementing comprehensive physical coun-
the performance of acoustic networks, enabling applications termeasures is challenging. Detecting compromised security
across broader areas [114]. However, achieving long-distance, in underwater networks relies on identifying abnormalities
high-speed acoustic communication brings its correspond- in communication and movement patterns; however, passive
ing set of challenges, necessitating continuous research and threats and well-crafted attacks present difficulties in such
innovation. Acoustic transmission exhibits the following key scenarios. Secondly, the removal of compromised nodes is
characteristics: imperative but incurs substantial costs, necessitating network-
• Low Propagation Speed: Acoustic signals in water have wide rekeying to prevent secret leakage. Moreover, the unique
a propagation speed of about 1500m/s, significantly attributes of underwater environments, such as limited physical
lower than radio frequency signals. This leads to longer protection and open acoustic links, provide adversaries with
signal propagation delays, impacting communication effi- the opportunity to passively intercept and analyze signals or
ciency [115]. actively disrupt network services. In contrast to terrestrial
• Short Propagation Distance: Absorption and scattering environments, underwater acoustic communication lacks a
in the water result in short propagation distances for standardized model, and channel quality is influenced by
acoustic signals. While signals can propagate up to 10km various underwater properties. This complicates the differen-
in seawater, the distance is even shorter in fresh water, tiation between evidence of attacks and instances of abnormal
limiting network coverage and scale [86]. communication in situations where multiple communication
• Severe Signal Attenuation: Acoustic signals experi- systems coexist [119], [120]. Additionally, the absence of
ence rapid attenuation during propagation, especially Global Positioning System (GPS) capabilities in underwater

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2561

environments affects localization and time synchronization and humidity. Ensuring the security of UWSNs requires
methods, which can be exploited by attackers [114], [121]. robust hardware and devices that can withstand these
environmental challenges and remain resilient against
D. Advantages of Trust Over Other Security Methods in potential attacker exploits.
UWSNs • Node Authentication: Node authentication is vital for
ensuring communication security in UWSNs. However,
The security concern in UWSNs has been extensively inves-
attackers may attempt to impersonate legitimate nodes,
tigated by numerous researchers [122], [123]. The deployment
leading to information disclosure and tampering.
of UWSN nodes in hazardous or hostile areas, often in
• Trust Bootstrapping: Existing trust models often rely on
large numbers, necessitates cost-effective solutions, which may
arbitrary initial trust values assigned to newly encoun-
result in lower reliability and increased vulnerability to poten-
tered nodes, which might not accurately reflect the true
tial adversarial attacks. Several network security standards
trustworthiness of these nodes based on their history and
have been established to bolster network security, forming the
behavior. Therefore, it becomes crucial to determine the
foundational framework for wireless network security [124].
actual initial trust value for each node.
However, the traditional methods employed, such as cryptog-
• Update of Trust Values: Effectively managing the lifetime
raphy, authentication, confidentiality, and message integrity,
of trust values for neighbors presents another significant
do not comprehensively address the security challenges. For
challenge in UWSNs. Due to limited storage capacity in
instance, adversaries might gain access to valid cryptographic
sensor nodes, storing trust values for all nodes becomes
keys and exploit them to infiltrate other nodes in the network.
impractical. Therefore, it is essential to implement an
Additionally, the reliability concern remains unaddressed when
efficient mechanism for stored trust values.
sensor nodes are affected by system faults. These two sources
Addressing these trust management challenges is crucial
of problems, system faults, and malicious nodes propagating
to establishing a secure and reliable UWSN infrastructure,
erroneous data or manipulating routing, can lead to a com-
enabling efficient data communication, and safeguarding the
plete breakdown of the network. Cryptography alone proves
integrity and confidentiality of critical information in under-
insufficient in resolving these issues. Consequently, it becomes
water environments.
crucial to assess the trustworthiness of nodes and the shared
information before basing any decisions on such data. This
underscores the importance of establishing a robust trust VI. T RUST M ANAGEMENT IN UWSN S
model, allowing a sensor node to infer the trustworthiness of Trust models play a vital role as security mechanisms
another node in the network. in UWSNs by facilitating the evaluation and establishment
of trust relationships among communication entities. This
E. Challenges of Trust Management in UWSNs section compares trust evaluation methods in UWSNs and
conventional WSNs, highlighting the need for specific mod-
Mitigating security challenges in UWSNs is of utmost
ifications in UWSNs due to their unique environmental and
significance due to their burgeoning prominence and
operational challenges. It then reviews UWSN-specific trust
wide-ranging applications encompassing marine resource
models, focusing on their role in enhancing network reliability
development, environmental preservation, military safeguard-
and countering security threats.
ing, and civilian affairs [125], [126]. Lapses in implementing
robust security protocols or the exploitation of vulnerabilities
A. Comparison of Trust Evaluation Between UWSNs and
by malicious actors can lead to dire ramifications, encompass-
WSNs
ing depletion of marine resources, ecological degradation, and
jeopardizing national security. Furthermore, UWSNs function In the realm of trust evaluation methods, a multitude of
in demanding underwater settings characterized by elevated approaches exist, catering to various network environments.
hydrostatic pressure, frigid temperatures, elevated salinity, and However, as elucidated earlier, the singular characteristics
restricted luminosity, engendering substantial impediments to of the underwater milieu render many of these methods
both system dependability and security. Consequently, the unsuitable for direct application in UWSNs. UWSNs, by their
imperative resolution of these security exigencies is imperative very nature, present a set of operational and environmental
to ensure the sustained advancement and efficient operation of challenges that are markedly distinct from those encountered
UWSNs. Some of the primary trust management challenges in terrestrial WSNs, which encompass IoT, MANETs, and IoV
faced by UWSNs are as follows: applications.
• Data Security: Data security is crucial in UWSNs, espe- The design and implementation of trust evaluation methods
cially as data often traverses through multiple nodes. in UWSNs necessitate a nuanced understanding of these
However, communication between nodes in underwater unique challenges. Factors such as acoustic communication
environments is constrained by signal propagation, lead- medium, higher latency due to signal propagation underwater,
ing to packet loss, retransmissions, and delays, thereby node mobility influenced by aquatic currents, and the height-
increasing the risks of data leakage and tampering. ened energy constraints inherent in underwater environments
• Hardware and Device Security: Nodes and devices critically impact the trust assessment process [8]. These
deployed in the underwater environment are subjected to aspects necessitate a tailored approach to trust evaluation in
extreme conditions, including high pressure, temperature, UWSNs, diverging significantly from conventional practices
in terrestrial networks.
Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2562 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE V
S UMMARY OF T RUST E VALUATION M ETHODS

A detailed comparison and exploration of these varying imposes significant constraints on communication and oper-
factors, and their implications on trust evaluation methods, ational capabilities, directly impacting the effectiveness of
are systematically presented in Table VI. This comparative trust methods traditionally used in terrestrial wireless sensor
analysis underscores the need for specialized trust evaluation networks. This section discusses how trust evaluation methods
strategies in UWSNs, taking into account their distinct opera- must be adapted for UWSNs. These modifications are driven
tional paradigms and environmental exigencies. by the unique operational characteristics and environmental
challenges of UWSNs. Key among these modifications are:
• Adaptation to High Delay and Variable Speed: Unlike
B. Modifications in Conventional Trust Evaluation Methods terrestrial networks, UWSNs experience high latency due
for UWSNs to the slow speed of acoustic signals. Trust evaluation
UWSNs present a unique set of challenges that necessitate methods must account for these delays in communication
specific considerations in the application and modification to avoid misjudging the trustworthiness of nodes based
of trust evaluation methods. The underwater environment on delayed responses [12].

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2563

TABLE VI
C OMPARISON OF T RUST E VALUATION B ETWEEN UWSN S AND WSN S

• Energy Efficiency Considerations: Nodes in UWSNs typi- accurate, efficient, and reliable under the specific conditions
cally have limited energy resources, and energy efficiency of UWSNs.
is a crucial factor. Trust evaluation methods need to be
energy-aware, minimizing the energy consumption for
trust computations and communications. C. Trust Management
• Mobility and Dynamic Topology: UWSNs often have The trust model plays a significant role in characterizing the
mobile nodes with dynamic topologies due to water lifecycle of node trust states [149]. Trust management aims to
currents. Trust evaluation methods should be adaptable to establish and maintain trust relationships within the network,
changing network topologies and capable of reassessing ensuring system security and reliability. These trust manage-
trust relationships as nodes move and new interactions ment approaches also enable the detection and response to
occur. potential attacks. In this subsection, we categorize trust models
• Robustness to Harsh Conditions: Underwater envi- according to a developed taxonomy and subsequently delve
ronments are prone to noise, signal attenuation, and into the details of these methodologies.
multi-path fading. Trust evaluation methods must be 1) Classification for Trust Management Approaches: Trust
robust against these harsh conditions and capable of management methods in UWSNs are diverse and multifaceted,
distinguishing between malicious behavior and environ- encompassing a range of approaches to assess and manage
mental interference. Moreover, UWSNs often have sparse trustworthiness within the network. As summarized in Fig. 5,
node deployments. Trust evaluation methods should be these methods can be broadly categorized into the following
effective even with limited interaction data and capable categories:
of extrapolating trust from sparse interactions. • Machine Learning-Based Methods: These approaches
• Integration of Environmental Data: Incorporating envi- utilize algorithms like Support Vector Machines, Random
ronmental data, such as water quality and pressure Forest, LSTM, Attention-based LSTM, Markov models,
changes, can enhance trust evaluations by providing con- and Q-learning. They adaptively update trust evaluations
text to the node behavior. For instance, sudden changes based on dynamic evidence and specific learning tech-
in sensor readings might be attributed to environmental niques.
factors rather than node malfunctions or malicious activ- • Weighted Sum-Based Methods: Employing weighted sum
ities. algorithms, these methods calculate trust values based on
Beyond these specific adaptations, trust evaluation in interaction thresholds and the priority of trust metrics.
UWSNs must also prioritize privacy preservation, particu- They update trust considering both cooperative and non-
larly in sensitive applications like military or environmental cooperative interactions.
monitoring [140]. The methods must be scalable and • Cloud Theory-Based Methods: Focusing on conceptualiz-
flexible to accommodate large network sizes and diverse ing trust through cloud-based approaches, these methods
operational scenarios, ranging from resource exploration to use cloud theory and models. Details on trust updates in
surveillance. Hybrid trust models, which combine direct these methods are often not extensively mentioned in the
and indirect trust assessments, are especially beneficial in literature.
UWSNs where direct node interactions might be limited. • Logic-Based Methods: Incorporating forms of logic such
Finally, tailoring trust evaluation methods to specific appli- as subjective logic, fuzzy logic, decision trees, and
cations within UWSNs is essential to ensure both the evolutionary game theory, these methods apply logical
accuracy and contextual relevance of trust assessments, meet- frameworks for trust assessment. Some, like decision tree
ing the unique requirements and challenges of each UWSN methods, update trust based on time-window evaluations.
deployment. • Probabilistic and Statistic-Based Methods: This category
In summary, adapting traditional trust evaluation methods includes Bayesian-based approaches, Beta distribution,
for UWSNs involves addressing the unique physical and oper- Gaussian distribution, and entropy methods. They often
ational characteristics of the underwater environment. These update trust considering factors like cooperative interac-
adaptations are essential to ensure that the trust evaluations are tions and statistical distributions.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2564 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

Fig. 5. Classification of Trust management methods.

The trust evaluation methodologies delineated in these The TMCA mechanism innovatively incorporates the concepts
categories provides a comprehensive toolkit for assessing of node reliability and link reliability to resolve these conflicts,
trustworthiness in UWSNs. The methodologies outlined in thereby streamlining the trust decision-making process. Khan
Tab. V encapsulate the diverse and dynamic nature of trust and Singh [130] introduced an effective weight-based trust
management in UWSNs. These approaches, which will be management approach that leverages communication (direct
subsequently introduced, offers a unique perspective and and indirect) trust and data trust, in combination with compres-
addresses different aspects of trust evaluation. The selection sive sensing techniques, to mitigate various internal attacks,
of a particular trust evaluation method is contingent upon such as badmouthing, black-hole, and grey-hole attacks. The
the nuanced requirements and distinctive characteristics of the indirect trust PRx ,y (Δt) of node ny , maintained at node nx
UWSN under consideration. Each method, with its unique at time Δt can be computed by the strategy employed for
operational mechanism and theoretical underpinnings, con- assigning weights to the participated Sensor Nodes (SNs).
tributes significantly to the overarching goal of maintaining Moreover, in [151], the authors utilize a threshold-based fil-
a secure and reliable network environment. This comparative tering technique to preprocess the reports received from nodes
analysis of existing trust models underscores the diversity in the network, ensuring that nodes with low reputation values
and sophistication of approaches available for addressing the have minimal impact on the final decision. This approach
complexities of trust management in UWSNs. allows such nodes an opportunity to rebuild their reputation as
2) Weighted Sum Trust Evaluation Method: This approach their reported results are still compared with the final collective
is commonly employed in WSNs, where the sensing results decision. Nodes reporting the same results as the final decision
from various nodes in the network are aggregated at a base are rewarded with increased reputation values.
station or fusion center to form an assessment of the monitored Although the Weighted Trust Evaluation method is simple
environment’s status [130], [132], [150]. In decision-making, and interpretable, it does have limitations in terms of accuracy
the weighted sum model serves as a simple method for and adaptability when compared to more sophisticated statis-
evaluating criteria among nodes, akin to the subjective logic tical and machine learning models. To enhance the robustness
trust evaluation. Agent trust towards another is determined of weighted evaluation, careful design considerations must
by combining direct experiences, indirect trusts, recommenda- be taken into account, including weight selection, metric
tions, ratings, and scores from other devices. pre-processing, model validation, and contextual threshold
To handle conflicts in third-party recommendations, a novel calibration. By addressing these factors, the accuracy and
trust management mechanism named Trust Management based effectiveness of the weighted trust evaluation can be improved.
on Conflict Adjudication (TMCA) is proposed for UWSNs. 3) Logic-Based Trust Management: Logic-based trust
The weighted sum method is used to calculate the trust value. management models primarily rely on logical rules and

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2565

reasoning mechanisms to infer trust relationships between

entities. Within this category, there are various trust models
based on logic.
Subjective logic (SL): The SL-based trust model constitutes
an innovative approach to trust modeling that is deeply rooted
in subjective judgments and individual beliefs, rendering it
particularly efficacious for trust management within intricate
networks such as WSNs and UWSNs [152]. Its distinctive
attributes encompass simplicity, efficient application, and the
quantification of belief, disbelief, and uncertainty. This model Fig. 6. Membership function of fuzzy logic-base trust model.
encompasses an intrinsic recognition of individual subjective
assessments and employs subjective weighting, thereby diverg-
ing from exclusive reliance on objective behavioral records weights deduced through the medium of fuzzy inference,
or evidentiary support. Opinions within the framework of culminates in a cohesive aggregation conducive to the pursuit
SL, denoted as (b, d, u), serve as meaningful indicators of of trust evaluation.
the degree of belief in the veracity of a given statement. In order to identify unreliable recommendations and dis-
Consequently, the calibration of trust levels through such opin- honest nodes in the network and mitigate potential risks,
ions becomes a viable proposition, effectively encapsulating the authors in [10] propose a recommendation management
scenarios of both diminished trust (e.g., T = {0.02, 0.9, 0.08}) trust mechanism based on collaborative filtering and variable
and heightened trust (e.g., T = {0.9, 0.0, 0.1}). This innovative weight fuzzy algorithm (CFFTM). They treat trust evaluation
paradigm thereby empowers higher-tier nodes to discern the as a multiattribute decision-making problem using the fuzzy
trustworthiness of entities by generating opinions that are comprehensive evaluation algorithm based on variable weight.
grounded in the assimilation of received data. The trust levels are represented by six fuzzy subsets U =
Researchers have applied SL to trust management. For {u1 , u2 , u3 , u4 , u5 , u6 }, where ui (i = 1, 2, 3, 4, 5, 6) denotes
instance, authors in [133] presents a trust model for underwater the ith fuzzy subset, representing complete distrust, very
networks, crucial for robust operations in military and public distrust, a little distrust, a little trust, very trust, and full trust,
safety scenarios. The model focuses on identifying attackers respectively. The trapezoidal affiliation function is used for the
through a reputation system, addressing a wide range of fuzzy subsets and is formulated as follows:
attacks. It monitors node behaviors, leveraging subjective logic ⎧
⎪
⎪ 0, x < a, x > b
to assign trust metrics to neighboring nodes. Uniquely, the ⎪
⎨ x −a , a ≤ x ≤ b
model also incorporates the varying conditions of underwater E (x , a, b, c, d ) = b−a (1)
acoustic channels, distinguishing between genuine packet loss ⎪
⎪ 1, b<x <c
⎪
⎩ d−x
d−c , c ≤ x ≤ d
and malicious activities. Another related work is the SL-
based trust network analysis (TNA-SL) approach proposed
in [153], which provides a symbolic representation and compu- Furthermore, the weight assigned to each trust evidence
tational technique to simplify and compute trust relationships changes dynamically to emphasize nodes with low trust
in networks. However, this model may be influenced by factors. In a proposed cluster head and cluster member node
subjective factors, which could lead to a lack of objectivity selection scheme [135], the trust value of nodes is calculated
and quantifiability. To address this issue, the triple SL model using fuzzy logic rules and several fuzzy parameters based
(3VSL) was proposed in [134] to study trust among vehicles. on node distance, energy, and relative mobility. However, this
This approach allows for both objective and subjective trust approach may suffer from information loss and computational
evaluations of vehicles in a distributed manner, thereby avoid- complexity issues. To address these concerns, an Anchor-
ing potential issues with pure subjective judgments. based No-Distance Measurement Collaborative Multi-Trust
Fuzzy logic: Fuzzy logic trust evaluation entails a pro- Management System (ARCMT) was introduced in [136].
cess of reasoning that anticipates approximate trust values This system employs a multi-trust-based scheme to protect
within the interactions among nodes. In this context, nodes sensor nodes from untrusted or malicious nodes and utilizes a
engage in submitting trust opinions pertaining to specific node lightweight Mamdani fuzzy model to reduce circuit complex-
services [95], employing fuzzy linguistic constructs such as ity and memory consumption.
‘Very bad,’ ‘Bad,’ ‘Normal,’ ‘Good,’ and ‘Very good.’ Within Game theory: Game theory offers a formal framework
this schema, each of these linguistic trust opinion expressions for analyzing interactions among rational agents making
is ascribed an associated weight, a determination achieved strategic decisions. Trust models employing evolutionary
through the intricate orchestration of the fuzzy inference game theory treat entities as strategic decision-makers whose
mechanism. The embodiment of the decision factors within the behavior is influenced by evolutionary games [154]. For
Fuzzy Inference System (FIS) is characterized by membership instance, the Evolutionary Game-based Security Clustering
functions, as exemplified in Figure 6, where the membership Protocol with Fuzzy Trust Evaluation and Outlier Detection
functions of the input linguistic variables align with a trape- (EGSCFO) [137] was proposed. This protocol employs a fuzzy
zoidal distribution. Subsequently, the amalgamation of the trust evaluation method to address the uncertainty of trust
linguistically conveyed feedback, alongside their concomitant and uses evolutionary game theory for a security clustering

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2566 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

TABLE VII
PAYOFF TABLE Decision Tree: Decision tree-based models offer an intuitive
and comprehensible approach for trust modeling. A study [64]
proposed a trust evaluation and update mechanism (TEUC)
based on the C4.5 decision tree algorithm for underwater
wireless sensor networks. The TEUC combined classification
algorithms with fuzzy logic theory to assess and update the
trust level of sensor nodes. The C4.5 decision tree algorithm
was utilized to classify nodes as trust or untrust, avoiding
protocol that selects secure cluster heads while isolating the complexities and security concerns of traditional weighted
suspicious nodes to conserve energy. trust computing. The key to decision tree learning lies in
The proposed approach defines a game as CEG = {N, S, selecting optimal partition attributes. The information gain
U}, where N denotes the set of game players, S = {Si |i ∈ N } ratio in C4.5 algorithm, calculated as follows, was used to
represents the strategy combination of all players, and U = prioritize attributes with higher information gain and avoid
{Ui |i ∈ N } stands for the set of utility functions. The payoff overfitting.
matrix for a simple game with two players is illustrated in In the TEUC approach, the adoption of the C4.5 decision
Table VII. In this matrix, a player who does not declare to be tree algorithm enables the classification of node trust without
the cluster head while no other player becomes the head risks the need for comprehensive trust weight calculation present
a malicious node becoming the head and launching attacks, in traditional algorithms. However, decision tree-based models
resulting in a loss z for the entire population. However, if may suffer from overfitting and difficulty in handling intricate
at least one other player declares to be the cluster head, it relationships.
can gain v by enjoying secure data transmission. On the other Isolation Forest: Likewise, models based on isolation
hand, if a player declares to be the cluster head, its payoff is forests can effectively handle high-dimensional data and out-
(v−c) due to an additional cost c associated with serving other liers. A study [71] proposed an Isolation Forest-based Trust
nodes. model (ITrust) to achieve precise calculation of node trust-
The authors of this research employed game theory to worthiness while considering resilience against anomalies and
construct a cluster head election model that achieves a balance attacks. Nodes with trust scores below a predefined threshold
between energy and security considerations. However, models are identified as defective. Nevertheless, these models may
based on evolutionary game theory may face challenges in encounter challenges in terms of computational complexity
terms of strategy selection and game model establishment. To and interpretability. The credibility of node ni based on
address this, a study [138] modeled the evolutionary process of the degree of anomaly or uncertainty (s) in the nodes is
attack strategies by malicious users, considering the dynamics represented by Trusti in the model.
and diversity of attack strategies. They implemented a multi- E(h (ti ))
−
utility reputation management scheme to better simulate trust Trust i = 1 − s = 1 − 2 c(ψ) , (2)
⎧
evolution and cooperative behavior. ⎪
⎨2H (ψ − 1) − 2(ψ − 1)/n, ψ>2
Blockchain: In trust models based on blockchain, the c(ψ) = 1, ψ = 2. (3)
establishment and maintenance of trust rely on consen- ⎪
⎩
sus and smart contract mechanisms within the blockchain 0, otherwise
network [155], [156]. In [139], a blockchain-based trust In the context of the proposed trust model, the notation
management model is proposed to identify Sybil nodes in E(h(ti )) represents the expectation for a set of h(x) outcomes
UWSNs. This model utilizes the HMM, enabling sensor nodes generated from m iTrees, and c(ψ) denotes the average path
to analyze and learn from each other’s behavioral patterns. length of the iTrees. Subsequently, using the calculated trust
These behavioral insights are then securely stored within the values, the sink node proceeds to assess the trustworthiness of
blockchain’s blocks. This approach allows any sensor node sensor nodes.
within the network to ascertain the trustworthiness of other Dempster-Shafer Theory: The trust model based on the
nodes by consulting the blockchain, thereby eliminating the Dempster-Shafer evidence theory treats trust evaluation as
need for repetitive trust value computations. Moreover, the an inference process, involving the collection, combination,
integration of blockchain technology enhances the system’s and analysis of evidence from various sources to quantify
capability to detect Sybil attacks, where an intruder node entity trustworthiness. In a study by [51], a redeemable SVM-
attempts to compromise the network by replicating the identity Dempster-Shafer (SVM-DS) fusion-based trust management
of a legitimate node. The blockchain framework effectively mechanism (SDFTM) is proposed for UASNs. The goal is to
identifies such malicious attempts, ensuring the integrity and achieve accurate trust evaluation of nodes while minimizing
security of the UWSN. Despite their advances, blockchain- misjudgments of normal nodes. The DS evidence theory is
based models may encounter scalability and performance utilized to synthesize the output results from each SVM
challenges. To address these issues in ad hoc networks, a model and effectively utilize uncertainty information among
study [157] proposed a three-dimensional link duration-based different pieces of evidence, thereby enhancing the precision
maximum minimum stable routing algorithm and a multipoint and robustness of the evaluation. The Belief-Plausibility-
relay-driven Byzantine fault-tolerant consensus mechanism to Ambiguity (BPA) function of each proposition is defined based
establish a trustworthy network framework. on the SVM classification of each individual trust evidence.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2567

Additionally, the DS decision rule for multi-feature decision The node trust values obtained from the Beta-based trust
results is applied to any trust classification. By combining the model are utilized to guide the selection of relay nodes,
trust classification results from the three types of trust evidence thereby mitigating the risks of internal attacks [144]. Node
using the Dempster-Shafer evidence theory, the overall trust behavior is classified into positive and negative categories
classification of each node is determined. In the study [158], a using the Beta distribution. While the Beta-based trust model
chance-based routing protocol based on the Dempster-Shafer effectively models the probability of binary events and updates
evidence theory was proposed, achieving prolonged network based on prior and posterior probabilities, it may have limita-
lifetimes and improved packet delivery rates through uniform tions in modeling multivariate events.
residual energy allocation. Models employing the Dempster- Gaussian Distribution: The Gaussian distribution-based
Shafer evidence theory can effectively handle uncertainty trust model is well-suited for modeling and analyzing contin-
and conflicts but require accurate determination of evidence uous random variables [145] and multivariate variables [146],
weights and credibility functions. The authors in [65] intro- distinguishing it from other approaches that primarily address
duced an enhanced Dempster-Shafer evidence theory and communication and binary perspectives. In [146], the proposed
fusion rule to ascertain the weights of routing channel param- algorithm introduces a novel trust computation scheme for a
eters for evaluating the trust value of each node, resulting in multivariate Gaussian reputation-based network. The construc-
a clustering tree-enhanced Dempster-Shafer evidence theory- tion of a robust trust-based network against misinformation
based bidirectional butterfly optimization algorithm. caused by malicious sensor nodes is achieved through mul-
4) Probability and Statistic-Based Trust Management: tivariate Gaussian reputation and trust modeling. However,
Probabilistic statistical-based trust models typically assess the Gaussian distribution-based model may exhibit bias when
and manage trust relationships by quantifying event proba- dealing with non-normally distributed data and may per-
bilities, analyzing data distributions, and making decisions form poorly in handling outliers. Additionally, the Gaussian
based on statistical inference. Some common probabilistic distribution-based trust model may not be ideal for handling
statistical-based trust models include Bayesian, Beta distribu- trust values in a state of confusion.
tion, Gaussian distribution, and entropy. Entropy theory: Entropy serves as an effective measure of
Bayesian-Based Approach: The Bayesian-based approach uncertainty and information content, enabling the assessment
integrates logical reasoning with probabilistic concepts to of the level of uncertainty and richness of information in trust
quantify uncertainty, allowing for inference based on both relationships [159], [160], [161]. In [147], authors introduced
logical principles and trust calculation utilizing probabilistic a trust model for underwater acoustic MANETs, addressing
statistical methods. This enables Bayesian-based methods security challenges stemming from inherent vulnerabilities and
to comprehensively address reasoning and uncertainty in marine characteristics. Entropy theory is utilized to assess the
trust management, resulting in more accurate and reliable uncertainty in trust values of sensor nodes. Utilizing machine
trust metrics and inference outcomes. While Bayesian-based learning and a cloud model, the model differentiates natural
trust models excel in accurate inference through the updat- disturbances from intentional misbehavior, enhancing trust
ing of prior beliefs, they require precise estimation of management in challenging underwater conditions. In another
prior probabilities. Therefore, in [141], a trust strategy- study [162], the authors explore the utilization of an entropy-
based dynamic Bayesian game (TSDBG) model addresses based method to evaluate node trustworthiness in a network.
cooperation and acoustic channel unreliability challenges in The entropy of a random variable x is calculated as follows:
underwater acoustic sensor networks. TSDBG uses trust
H (x ) = −p(x ) log2 p(x ) − (1 − p(x )) log2 (1 − p(x )). (4)
and payoff calculations to identify misbehaving nodes, with
continuous monitoring and trust value updates via Bayes’
rule. In [142], a method for identifying internal attacks In this context, p(x) represents the probability of occurrence
in healthcare insurance software-defined networks (SDNs) of a random variable x. The authors utilize entropy as a
assumes probability-equivalent prior information with equal metric to assess the trustworthiness of a node, leveraging
probabilities to follow a uniform prior distribution, achieving the degree of consistency in its behavior pattern to make a
unbiasedness and simplicity. The Bayesian approach is flexible trust decision. However, it is important to note that entropy
in handling uncertainty and updating prior knowledge but may solely captures the consistency of a node’s behavior pattern.
entail high computational complexity in dealing with high- As a consequence, nodes that consistently behave well and
dimensional data. nodes that consistently exhibit misbehavior may both yield
Beta Distribution: The Beta distribution is a commonly low entropy values. To differentiate between these cases and
used probability distribution for modeling uncertainty in determine the trustworthiness value, the authors propose the
random variables [70]. In [143], the author proposed a Beta- following mapping from the entropy value:

based Trust and Reputation Evaluation System (BTRES), H (p)
T (H (p)) = 2 , 0 ≤ p < 0.5
(5)
utilizing the Beta distribution to describe the distribution of H (p)
node trustworthiness based on node behavior monitoring and 1 − 2 , 0.5 < p ≤ 1
guiding the selection of relay nodes to mitigate internal attack where p represents the probability of positive interactions.
risks. In BTRES, the direct trust evaluation, representing the However, they do not address the issue of smoothing the
statistical expectation of the reputation function when node ni evidence, which leads to challenges associated with sparse
interacts with node nj , is calculated by the Beta distribution. evidence and its adverse effects on trustworthiness assessment.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2568 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

Hidden Markov Model: Hidden Markov Models (HMMs)

are particularly used for analyzing temporal or sequential
data. HMMs are extensively employed in fields such as signal
processing, pattern recognition, computational finance, and
in areas of artificial intelligence like machine learning and
computer vision. For instance, in [50], the authors introduce a
trust management model using Hidden Markov Model (HMM)
to assess sensor node trustworthiness in harsh underwater
environments, addressing internal threats. It also proposes
ChaCha20, a lightweight stream cipher, for energy-efficient
Fig. 7. Machine learning-Based method. protection against external attacks in UWSNs.
Random Forest (RF): RF is an ensemble learning method
known for its ability to handle high-dimensional data, robust-
ness against noise, and feature importance assessment. In the
5) Cloud Theory-Based Trust Management: The trust RFtrust model proposed by the authors in [127], random
model based on cloud theory integrates the advantages of sub- forests and SL are utilized to detect and remove black hole
jective logic-based and fuzzy logic-based trust models [163]. nodes, enhancing trusted routing in the IoT environment to
It leverages subjective assessment and uncertainty reasoning improve network performance. Nonetheless, Random Forest
to handle complex, uncertain, and fuzzy trust issues [90]. may exhibit subpar performance when dealing with high-
Additionally, the trust model based on cloud theory inherits dimensional sparse data and may require parameter tuning and
the advantages of probability statistics, enabling the quantifi- increasing the number of trees to address overfitting concerns.
cation and management of trust relationships between entities Deep Learning (DL): Deep Learning (DL) is a significant
and supporting decision-makers in making reasonable trust subfield of machine learning, known for its ability to handle
decisions in uncertain environments. However, the subjective complex and high-dimensional data representations. In a study
assessment in the trust model based on cloud theory may by Deng et al. [166], a novel trust-based approach for rec-
be influenced by the decision-maker’s subjective cognition ommendation in social networks was developed. The authors
and evaluation, potentially introducing subjective biases [91]. employed deep learning techniques to enhance trust-aware
In [148], the authors present an Energy-balanced Trust Cloud social recommendations. Specifically, a deep autoencoder was
Migration scheme (ETCM) for UASNs to improve energy used to pre-train the rating matrix and learn the initial values
efficiency in trust management. ETCM uses a Destination of the latent features of users and items. A two-phase rec-
Node Determination process for trust cloud migration, opti- ommendation process was proposed to leverage deep learning
mizing energy distribution via simulated annealing algorithms. for initialization and to combine users’ interests with those
It employs Cluster Ability and Node Ability indicators for of their trusted friends, along with the impact of community
selecting suitable clusters and nodes, focusing on residual effects, to provide personalized recommendations.
energy and connectivity. This approach aims to balance energy Long Short-Term Memory (LSTM): LSTM is a type of
consumption across UASNs, enhancing the network’s overall recurrent neural network well-suited for processing sequential
lifetime and efficiency. data, capturing temporal relationships, and handling long-term
6) Machine Learning-Based Trust Management: Machine dependencies [96], [167]. Both Ltrust and Atrust [92], [128]
learning-based trust models have gained popularity due to their utilize LSTM algorithms to establish trust models and predict
excellent performance in trust measurement and evaluation trust values by collecting multiple direct trust evidence.
across various domains. The Machine learning-Based methods In [92], the LTrust model, based on an LSTM recurrent
are classified in Figure 7 [164]. neural network, is designed to calculate trust metrics. The
Support Vector Machines (SVM): SVM is a powerful LSTM architecture involves output ht and external input xt
technique for handling classification and regression tasks in at time t, along with cell state Ct and activation function
high-dimensional feature spaces. It demonstrates good general- σ as the sigmoid function. The forget gate ft , input gate it ,
ization capabilities, especially for small-scale datasets. In [72], and output gate ot are integral components in the LSTM that
the authors propose a collaborative trust model called Support play significant roles in the prediction process within a given
Vector Machines-based Trust Management System (STMS). interval t. Their respective mapping functions are as follows:
The model utilizes SVM techniques to train a trust prediction  
model, enabling an accurate evaluation of trust. Similarly, ft = σ Wif xt + bif + Whf ht−1 + bhf ,
in another study [165], the authors introduce a quantifiable it = σ(Wii xt + bii + Whi ht−1 + bhi ),
trust assessment model using SVM. They employ machine ot = σ(Wio xt + bio + Who ht−1 + bho ). (6)
learning principles to classify trust features and generate a
final trust value for decision-making. However, SVM may With the integration of direct trust and filtered recommen-
face challenges with computational complexity when dealing dation trust as input, LTrust constructs an adaptive trust model
with large-scale datasets, sensitivity to noise and outliers, and based on an LSTM network. Using this model, sink nodes
the requirement of kernel functions for handling nonlinear evaluate the trust value of entities and identify anomalous
problems. nodes. However, LSTM, despite its advantages, exhibits high

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2569

Fig. 8. Network structure of LSTM.

computational complexity and may suffer from overfitting Khalid et al. [171] propose an Adaptive Trust-based Routing
when dealing with short sequence data or sequences with Protocol (ATRP), which incorporates direct trust, indirect trust,
simple patterns, necessitating appropriate parameter tuning and and witness trust while considering multiple factors, such
data preprocessing. as resources and security, in its trustworthiness evaluation
Reinforcement Learning: Trust models utilizing reinforce- through pairwise comparison. The main objective of ATRP
ment learning offer superior evaluation performance compared is to provide an efficient trust-based routing protocol for
to traditional approaches. These models, like reinforcement selecting relay nodes, taking into account multiple trust factors
learning agents, learn optimal strategies through interactions and utilizing the Q-learning technique for multilevel trust
and feedback from the environment. In trust evaluation, this evaluations. Q-learning proves to be a suitable technique
involves gathering trust evidence from entities and making for rewarding every hop during data transmission from the
decisions based on this information. Additionally, entities in source node to the destination, facilitating the determination
trust models are rewarded or penalized based on their behavior, of trust reputation for individual sensor nodes and aiding in
akin to reinforcement learning, where agents adjust actions the identification of malicious nodes. However, it is essential
according to reward and punishment signals [55], [73]. Two to acknowledge that Q-learning-based trust models may have
main categories of reinforcement learning-based trust models longer learning times and complex model training processes.
are Markov-based and Q-learning-based models. Markov mod- 7) Comparison of Trust Evaluation Methods: The weighted
els are proficient in modeling state transitions and are easily sum method is simple but has limitations in handling
interpretable, helping to understand the dynamic shifts in complexity and uncertainty. Logic-based approaches, such as
entities’ behavior and trustworthiness [75]. However, they typ- subjective logic, fuzzy logic, and Dempster-Shafer theory,
ically consider only the immediate previous state, potentially model uncertainty through reasoning but may lack objectivity,
overlooking longer-term historical data, and might struggle cause information loss, and require accurate input weights.
with computational efficiency in large or continuous state Game theory and blockchain offer trust evolution modeling
spaces. and tamper resistance but face challenges related to complexity
Markov models offer a valuable probabilistic framework and scalability. Statistical models, such as Bayesian infer-
for Q-learning in trust models. Q-learning-based trust models ence, distributions, and entropy, provide mathematical trust
demonstrate adaptability to uncertain environments and do not quantification but rely on assumptions about data normality
require prior knowledge [168], [169], [170]. In [129], a trust- and may have high computational needs. Machine learning
based routing protocol leveraging Q-learning for learning trust techniques, including SVMs, random forests, and LSTMs, can
elements to detect attacks and ensure network performance was learn trust patterns but may encounter issues such as over-
proposed. The protocol introduces a distributed transmission fitting, noise sensitivity, and interpretability. Reinforcement
technology prioritizing the trustworthiness of mission-critical learning, involving Markov chains and Q-learning, enables
data through Q-learning, taking into account trustworthiness, online trust adaptation but requires long training times and
Quality of Service (QoS), and energy factors. Reinforcement provides limited historical context. Cloud theory combines
learning is formulated using the Markov decision process subjective, fuzzy, and probability models, inheriting their
(MDP), represented by a quintuple (S, A, E, T, R), signifying advantages and limitations, including subjective bias. While
states of the system, actions, external events, the transition traditional probabilistic statistical methods focus on using
function, and rewards. The trust value is updated using the explicit probability distributions and statistical models to
Q-learning technique, where the agent learns an action-value address uncertainty, Cloud Theory processes fuzziness and
function, denoted as Q(state, action), which describes the value randomness simultaneously through the generation of cloud
of performing an action in a given state s. models, providing a more flexible approach to trust evaluation.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2570 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

In conclusion, each trust model is applicable to specific of the network. By considering data from various layers, such
scenarios and environments. Therefore, it is crucial to consider as the physical signal characteristics, MAC layer coopera-
the requirements carefully. Furthermore, a single trust model tion, and network routing behaviors, this approach provides
may have certain disadvantages compared to the evaluation a holistic understanding of node activities [6]. It enables
performance achieved by integrating multiple trust models. more accurate and context-aware trust evaluations, crucial for
Hence, it is necessary to understand the advantages and dynamic underwater environments where network conditions
limitations of different trust models, as they can comple- can vary significantly. Effective cross-layer trust management
ment and combine with each other to provide more robust can lead to improved network performance, adaptability, and
and comprehensive trust management solutions. By selecting resilience against various types of attacks and failures.
appropriate model combinations and integration methods, we 4) Security Against Sophisticated Attacks: Security against
can leverage the strengths of each model to meet the needs of sophisticated attacks in UWSNs is a critical aspect that ensures
complex trust management problems. the integrity and functionality of the network. Advanced
security measures, such as specialized intrusion detection
VII. O PEN I SSUES AND D ISCUSSIONS systems and robust encryption protocols, are essential to pro-
tect against threats like Sybil, wormhole, or denial-of-service
Despite the considerable amount of research conducted
attacks [9], [172]. Continual adaptation and updating of these
on trust management in UWSNs, several research directions
security measures are vital to counteract evolving attack
require further exploration in UWSNs. In this section, we
techniques. By implementing multi-layered security strategies
highlight some open issues, propose potential future research
that combine various defensive mechanisms, UWSNs can
trends related to trust management, and discuss challenges on
maintain high levels of security and reliability even in the face
trust computing in UWSNs.
of complex and sophisticated cyber threats.
5) Privacy-Preserving Trust Management: Privacy-
A. Future Research Directions preserving trust management in UWSNs addresses the need to
1) Resilience to Uncertainty of Environmental Factors: safeguard sensitive data and node identities while maintaining
In UWSNs, trust management systems face the dual chal- effective trust assessments. The challenge lies in adapting
lenge of adapting to the dynamic aquatic environment and terrestrial privacy models, often too complex and resource-
effectively handling the inherent uncertainty and ambiguity. intensive for UWSNs, to the unique UWSNs environment.
Environmental factors like water currents, pressure varia- This necessitates the development of streamlined, lightweight
tions, and acoustic channel fluctuations significantly influence privacy models tailored for UWSNs. Techniques such as
network performance, necessitating the development of adap- k-anonymity, l-diversity, and differential privacy should be
tive trust management methods [172]. These methods must modified to fit the constraints of underwater networks.
adjust to changes in node mobility or signal attenuation These refined methods will enable trust evaluations while
and integrate environmental sensing for real-time network ensuring data confidentiality and node anonymity. Moreover,
adjustments. Furthermore, techniques such as probabilis- compliance with data protection regulations and adherence
tic modeling, fuzzy logic, and Bayesian inference can be to ethical standards are fundamental to these privacy-
employed to manage uncertainty and derive meaningful trust preserving frameworks, guaranteeing that UWSNs operate
assessments. These approaches allow for flexible and nuanced within established legal and moral parameters, thus upholding
trust evaluations, accommodating the variability and ambiguity network trustworthiness.
inherent in underwater sensor data. Additionally, the use 6) Communication Overhead Reduction: In UWSNs,
of robust hardware and error-correction techniques ensures reducing communication overhead is crucial for efficient
continuous operation and reliable data transmission even under network operation, given the limited bandwidth and high
adverse environmental conditions. propagation delays typical in underwater environments.
2) Energy-Efficient Trust Management: In the context of Participating in a trust-based system incurs additional costs
UWSNs, energy-efficient trust management is paramount due for nodes, including increased computation and storage
to the limited energy resources available to underwater sensors. burdens [8]. For distributed communication networks, the
Developing trust management protocols that minimize energy dissemination of trust evidence and updating node reputations
consumption without compromising the accuracy of trust require additional communication overhead. Existing research
assessments is crucial [87]. Techniques such as optimizing the has not adequately addressed these issues. Trust models
trust update intervals, reducing redundant data transmissions, often use flooding mechanisms to disseminate trust evidence
and employing energy-efficient cryptographic methods can widely, but this can adversely impact network performance.
significantly enhance the energy efficiency of these networks. Studies are needed to determine whether propagating trust
The implementation of energy-saving strategies, including evidence to all nodes is necessary. Strategies to minimize
intelligent sleep/wake cycling for sensor nodes, further con- overhead include optimizing data packet sizes, employing
tributes to the sustainability and longevity of UWSNs. efficient routing protocols, and utilizing data aggregation
3) Cross-Layer Trust Management Approaches: Cross- techniques. By reducing the frequency of data transmissions
layer trust management in UWSNs involves integrating trust and adopting protocols that avoid unnecessary handshakes
assessment mechanisms across different layers of the network or acknowledgments, UWSNs can conserve valuable energy
protocol stack, enhancing the overall reliability and efficiency resources and enhance network lifetime. Implementing

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2571

intelligent data compression algorithms and prioritizing B. Discussions on Trust Computation

critical data transmissions also contribute to minimizing The process of trust computation encompasses several
communication overhead, ensuring the delivery of essential phases, each with its unique challenges. While significant
information while reducing the overall network load. advancements have been made in this domain, there remain
7) Development of Autonomous Energy Replenishment unresolved complexities that are inherently associated with the
Systems for UWSNs: Considering the insights from Section V, unique attributes of UWSNs. Several challenges need to be
a crucial aspect of UWSNs is the management of power further discussed, as outlined below.
resources. The nodes in UWSNs typically operate with 1) Trust Bootstrapping in UWSNs: The determination of
finite energy reserves, and recharging within the underwater initial trust values is crucial, especially in dynamic environ-
environment presents significant logistical challenges. This ments like UWSNs where nodes frequently encounter new
limitation not only restricts battery longevity but also impacts entities. The challenge lies in accurately assessing a node’s
the overall network lifespan, as the failure of even a single trustworthiness without prior interaction history. Hybrid mech-
node can disrupt network operations. In addressing these anisms incorporating historical data, and cooperative behaviors
power constraints, recent studies like Chin et al. [173] have might offer a solution [177]. However, this approach needs
explored the potential of specialized Lithium-based batteries further research to address the specific operational dynamics
for underwater applications, demonstrating successful test- of UWSNs.
ing of these power systems. Additionally, Cho et al. [174]
2) Trust Composition in UWSNs: The selection and com-
proposed a pioneering concept of a self-sufficient marine buoy
position of trust metrics in UWSNs are critical. Considering
system, utilizing seawater batteries (SWBs) that offer efficient
the harsh underwater conditions and varying acoustic signal
power management and predictive capabilities for degrada-
qualities, trust composition must factor in data perception
tion and faults. Building on these advancements, there is a
trust [178], particularly the integrity and quality of sensed data.
pressing need for the development of underwater-compatible
This involves not only assessing the data during collection
power solutions, such as an autonomous underwater bat-
and pre-processing but also considering sensor reliability and
tery system or an automatic recharging module for
environmental impacts on sensor performance.
UWSNs.
8) Enhancing Real-World Evaluations in UWSNs: The 3) Trust Propagation in UWSNs: In UWSNs, trust prop-
intricacies and limitations inherent to underwater environ- agation faces the challenge of efficiently disseminating trust
ments necessitate tailored technological solutions. Notably, information in a dynamic and often disconnected environment.
current research predominantly revolves around theoretical Both distributed and centralized trust propagation approaches
frameworks, with only a select few proposals, such as those have their limitations [179], such as the need for efficient
detailed in [103] and [51], undergoing real-world testing. The data filtering and the challenges of establishing infrastructure
complexity and dynamism of underwater acoustic networking for centralized systems. Tailoring these approaches to suit the
environments pose significant challenges in creating accurate mobility and connectivity patterns in UWSNs is a significant
simulations. Consequently, field testing emerges as a critical challenge.
method for authentic validation of trust management systems 4) Trust Aggregation in UWSNs: Aggregating trust evi-
in UWSNs. To bridge the gap between theory and practice, dence from various sources is complicated by issues like
there is an urgent need to conduct comprehensive sea trials value imbalance and the need for categorization [180]. The
that rigorously evaluate the practical efficacy and performance aggregation algorithms must be robust against manipulations
of these trust management solutions. and capable of categorizing trust scores based on different
9) Application of Multi-Medium Communication: criteria. This is particularly challenging in UWSNs where
Employing a hybrid communication model in UWSNs that nodes may have varying levels of energy and connectivity.
integrates acoustic, optical, RF, and Magnetic Induction (MI) 5) Trust Update in UWSNs: Updating trust values in a
technologies is essential to optimize performance. In the timely and efficient manner is critical. Due to the mobility of
literature, there are some hybrid-mode routing protocols nodes and the vast size of the network, storing all past trust
for UWSNs [175], [176]. Each medium presents unique values is impractical [64]. Developing an efficient policy for
advantages and challenges: optical systems offer high data the lifetime and decay of trust values, considering the unique
rates in clear waters but are range-limited, RF excels in aspects of UWSNs like node mobility and energy constraints,
penetration and longer distances at the cost of lower data rates, is essential.
and MI is suited for turbid waters with short-range efficacy 6) Trust Formation in UWSNs: Forming a composite trust
yet complex implementation. Nonetheless, the integration of score from multiple metrics involves balancing trust met-
trust management within such hybrid communication models rics with QoS. In UWSNs, this balance is crucial as both
remains underexplored. A multi-medium approach addresses interactive relationships and QoS are affected by the unique
these individual limitations, enhancing UWSNs by augmenting underwater conditions. Addressing this challenge involves
transmission speeds, extending battery life, and ensuring creating trust formation models that can adapt to the changing
reliable data transfer. This integrated communication strategy underwater environment and node behaviors.
is vital for tackling the challenges in UWSN environments, In summary, trust computation in UWSNs is a multi-
as outlined in Section V-B, through a synergistic utilization faceted challenge that requires innovative approaches tailored
of diverse communication mediums. to the specific characteristics of underwater environments.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2572 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

Each phase of trust computation needs to be adapted to address [14] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro,
the unique challenges posed by UWSNs. “Geographic and opportunistic routing for underwater sensor
networks,” IEEE Trans. Comput., vol. 65, no. 2, pp. 548–561,
Feb. 2016.
VIII. C ONCLUSION [15] M. Saeed, M. Aftab, R. Amin, and D. Koundal, “Trust management
model in IoT: A comprehensive survey,” in Proc. Int. Conf. Innovat.
In conclusion, trust management in UWSNs presents unique Bio-Inspir. Comput. Appl., 2021, pp. 675–684.
challenges due to the complex underwater communication [16] A. K. Junejo, N. Komninos, M. Sathiyanarayanan, and B. S. Chowdhry,
media, logistical constraints, and environmental factors. The “Trustee: A trust management system for fog-enabled cyber phys-
ical systems,” IEEE Trans. Emerg. Topics Comput., vol. 9, no. 4,
dynamic nature of underwater communication, coupled with pp. 2030–2041, Oct.–Dec. 2021.
the deployment and maintenance challenges of underwa- [17] N. U. R. Junejo et al., “A survey on physical layer techniques and
ter nodes, requires tailored trust management mechanisms. challenges in underwater communication systems,” J. Mar. Sci. Eng.,
vol. 11, no. 4, p. 885, 2023.
Integrating various trust models, including weighted sum [18] H. Li, Y. He, X. Cheng, H. Zhu, and L. Sun, “Security and privacy
methods, logic-based techniques, probability and statistics in localization for underwater sensor networks,” IEEE Commun. Mag.,
models, machine learning paradigms, and cloud theory, can vol. 53, no. 11, pp. 56–62, Nov. 2015.
[19] M. C. Domingo, “Securing underwater wireless communication
lead to more robust solutions for UWSNs. Effective trust networks,” IEEE Wireless Commun., vol. 18, no. 1, pp. 22–28,
management is crucial to ensure reliable, secure, and efficient Feb. 2011.
data transmission, enabling the successful operation of UWSN [20] J. Jiang, G. Han, and C. Lin, “A survey on opportunistic routing
protocols in the Internet of Underwater Things,” Comput. Netw.,
applications. Future research should focus on refining existing vol. 225, Apr. 2023, Art. no. 109658.
trust models, exploring new approaches, and testing proposed [21] Y. Dong and P. Liu, “Security considerations of underwater acoustic
solutions in real-world UWSN deployments. With improved networks,” in Proc. Int. Congr. Acoust. (ICA), 2010, pp. 23–27.
[22] G. Ferri et al., “Cooperative robotic networks for underwater surveil-
trust management, UWSNs can achieve reliable and secure lance: An overview,” IET Radar, Sonar Navig., vol. 11, no. 12,
communications, enabling them to fulfill their potential in pp. 1740–1761, 2017.
diverse underwater applications. [23] W. Kong, X. Li, L. Hou, and Y. Li, “An efficient and credible
multi-source trust fusion mechanism based on time decay for edge
computing,” Electronics, vol. 9, no. 3, p. 502, 2020.
R EFERENCES [24] A. Hbaieb, S. Ayed, and L. Chaari, “A survey of trust management
in the Internet of Vehicles,” Comput. Netw., vol. 203, Feb. 2022,
[1] X. Wei, H. Guo, X. Wang, X. Wang, and M. Qiu, “Reliable data col- Art. no. 108558.
lection techniques in underwater wireless sensor networks: A survey,”
[25] J. Wang, Z. Yan, H. Wang, T. Li, and W. Pedrycz, “A survey on
IEEE Commun. Surveys Tuts., vol. 24, no. 1, pp. 404–431, 2nd Quart.,
trust models in heterogeneous networks,” IEEE Commun. Surveys Tuts.,
2022.
vol. 24, no. 4, pp. 2127–2162, 4th Quart., 2022.
[2] S. Khisa and S. Moh, “Survey on recent advancements in energy-
[26] R. Hussain, J. Lee, and S. Zeadally, “Trust in VANET: A survey of
efficient routing protocols for underwater wireless sensor networks,”
current solutions and future research opportunities,” IEEE Trans. Intell.
IEEE Access, vol. 9, pp. 55045–55062, 2021.
Transp. Syst., vol. 22, no. 5, pp. 2553–2571, May 2021.
[3] T. Qiu, Z. Zhao, T. Zhang, C. Chen, and C. L. P. Chen, “Underwater
Internet of Things in smart ocean: System architecture and open issues,” [27] Q.-u.-A. Arshad, W. Z. Khan, F. Azam, M. K. Khan, H. Yu, and
IEEE Trans. Ind. Informat., vol. 16, no. 7, pp. 4297–4307, Jul. 2020. Y. B. Zikria, “Blockchain-based decentralized trust management in IoT:
[4] M. Jahanbakht, W. Xiang, L. Hanzo, and M. R. Azghadi, “Internet of Systems, requirements and challenges,” Complex Intell. Syst., vol. 9,
Underwater Things and big marine data analytics—A comprehensive pp. 1–22, Apr. 2023.
survey,” IEEE Commun. Surveys Tuts., vol. 23, no. 2, pp. 904–956, 2nd [28] I. U. Din, M. Guizani, B.-S. Kim, S. Hassan, and M. K. Khan, “Trust
Quart., 2021. management techniques for the Internet of Things: A survey,” IEEE
[5] Y. Yang, Y. Xiao, and T. Li, “A survey of autonomous underwater Access, vol. 7, pp. 29763–29787, 2018.
vehicle formation: Performance, formation control, and communication [29] A. I. A. Ahmed et al., “Trust and reputation for Internet of Things:
capability,” IEEE Commun. Surveys Tuts., vol. 23, no. 2, pp. 815–841, Fundamentals, taxonomy, and open research challenges,” J. Netw.
2nd Quart., 2021. Comput. Appl., vol. 145, Nov. 2019, Art. no. 102409.
[6] A. Boukerche and P. Sun, “Design of algorithms and protocols for [30] O. Khalid et al., “Comparative study of trust and reputation systems
underwater acoustic wireless sensor networks,” ACM Comput. Surv., for wireless sensor networks,” Secur. Commun. Netw., vol. 6, no. 6,
vol. 53, no. 6, pp. 1–34, 2020. pp. 669–688, 2013.
[7] Y. Noh, U. Lee, P. Wang, B. S. C. Choi, and M. Gerla, “VAPR: Void- [31] G. Han, J. Jiang, L. Shu, J. Niu, and H.-C. Chao, “Management and
aware pressure routing for underwater sensor networks,” IEEE Trans. applications of trust in wireless sensor networks: A survey,” J. Comput.
Mobile Comput., vol. 12, no. 5, pp. 895–908, May 2013. Syst. Sci., vol. 80, no. 3, pp. 602–617, 2014.
[8] S. Jiang, “On securing underwater acoustic networks: A survey,” IEEE [32] F. Ishmanov and Y. Bin Zikria, “Trust mechanisms to secure routing
Commun. Surveys Tuts., vol. 21, no. 1, pp. 729–752, 1st Quart., 2019. in wireless sensor networks: Current state of the research and open
[9] H. Mousa, S. B. Mokhtar, O. Hasan, O. Younes, M. Hadhoud, and research issues,” J. Sensors, vol. 2017, Feb. 2017, Art. no. 4724852.
L. Brunie, “Trust management and reputation systems in mobile [33] S. El-Rabaie, D. Nabil, R. Mahmoud, and M. A. Alsharqawy,
participatory sensing applications: A survey,” Comput. Netw., vol. 90, “Underwater wireless sensor networks (UWSN), architecture, rout-
pp. 49–73, Oct. 2015. ing protocols, simulation and modeling tools, localization, security
[10] M. Zhang, R. Feng, H. Zhang, and Y. Su, “A recommendation issues and some novel trends,” Netw. Commun. Eng, vol. 7, no. 8,
management defense mechanism based on trust model in underwater pp. 335–354, 2015.
acoustic sensor networks,” Future Gener. Comput. Syst., vol. 145, [34] M. M. Ogonji, G. Okeyo, and J. M. Wafula, “A survey on privacy and
pp. 466–477, Aug. 2023. security of Internet of Things,” Comput. Sci. Rev., vol. 38, Nov. 2020,
[11] H. Yu, Z. Shen, C. Miao, C. Leung, and D. Niyato, “A survey of Art. no. 100312.
trust and reputation management systems in wireless communications,” [35] M. Anugraha and S. Krishnaveni, “Recent survey on efficient trust
Proc. IEEE, vol. 98, no. 10, pp. 1755–1772, Oct. 2010. management in mobile ad hoc networks,” in Proc. Int. Conf. Circuit,
[12] G. Han, J. Jiang, N. Sun, and L. Shu, “Secure communication for Power Comput. Technol. (ICCPCT), 2016, pp. 1–4.
underwater acoustic sensor networks,” IEEE Commun. Mag., vol. 53, [36] A. B. Usman and J. Gutierrez, “Toward trust based protocols in a
no. 8, pp. 54–60, Aug. 2015. pervasive and mobile computing environment: A survey,” Ad Hoc
[13] R. Zhu, Q. Jiang, X. Huang, D. Li, and Q. Yang, “A reinforcement- Netw., vol. 81, pp. 143–159, Dec. 2018.
learning-based opportunistic routing protocol for energy-efficient [37] K. Govindan and P. Mohapatra, “Trust computations and trust dynamics
and void-avoided UASNs,” IEEE Sensors J., vol. 22, no. 13, in mobile adhoc networks: A survey,” IEEE Commun. Surveys Tuts.,
pp. 13589–13601, Jul. 2022. vol. 14, no. 2, pp. 279–298, 2nd Quart., 2012.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2573

[38] Y. Yu, K. Li, W. Zhou, and P. Li, “Trust mechanisms in wireless sensor [62] I. Ahmad et al., “Analysis of security attacks and taxonomy in under-
networks: Attack analysis and countermeasures,” J. Netw. Comput. water wireless sensor networks,” Wireless Commun. Mobile Comput.,
Appl., vol. 35, no. 3, pp. 867–880, 2012. vol. 2021, pp. 1–15, Dec. 2021.
[39] D. R. K. Mary, E. Ko, S.-G. Kim, S.-H. Yum, S.-Y. Shin, and [63] T. Kalidoss, L. Rajasekaran, K. Kanagasabai, G. Sannasi, and
S.-H. Park, “A systematic review on recent trends, challenges, privacy A. Kannan, “QoS aware trust based routing algorithm for wireless
and security issues of underwater Internet of Things,” Sensors, vol. 21, sensor networks,” Wireless Pers. Commun., vol. 110, pp. 1637–1658,
no. 24, p. 8262, 2021. Feb. 2020.
[40] G. Yang, L. Dai, and Z. Wei, “Challenges, threats, security issues and [64] J. Jiang, X. Zhu, G. Han, M. Guizani, and L. Shu, “A dynamic
new trends of underwater wireless sensor networks,” Sensors, vol. 18, trust evaluation and update mechanism based on C4.5 decision tree
no. 11, p. 3907, 2018. in underwater wireless sensor networks,” IEEE Trans. Veh. Technol.,
[41] A. G. Yisa, T. Dargahi, S. Belguith, and M. Hammoudeh, “Security vol. 69, no. 8, pp. 9031–9040, Aug. 2020.
challenges of Internet of Underwater Things: A systematic literature [65] S. N. Mahapatra, B. K. Singh, and V. Kumar, “Secure energy aware
review,” Trans. Emerg. Telecommun. Technol., vol. 32, no. 3, 2021, routing protocol for trust management using enhanced dempster Shafer
Art. no. e4203. evidence model in multi-hop UWAN,” Wireless Netw., vol. 28, no. 7,
[42] M. Majid et al., “Applications of wireless sensor networks and Internet pp. 3059–3076, 2022.
of Things frameworks in the industry revolution 4.0: A systematic [66] T. Zahariadis et al., “Design and implementation of a trust-aware
literature review,” Sensors, vol. 22, no. 6, p. 2087, 2022. routing protocol for large WSNs,” Int. J. Netw. Secur. Appl. (IJNSA),
[43] G. Yang, L. Dai, G. Si, S. Wang, and S. Wang, “Challenges and security vol. 2, no. 3, pp. 52–68, 2010.
issues in underwater wireless sensor networks,” Procedia Comput. [67] X.-Y. Xiao, W.-C. Peng, C.-C. Hung, and W.-C. Lee, “Using sensor-
Sci., vol. 147, no. 147, pp. 210–216, Feb. 2019. ranks for in-network detection of faulty readings in wireless sensor
[44] Y. Cong, G. Yang, Z. Wei, and W. Zhou, “Security in underwater networks,” in Proc. 6th ACM Int. Workshop Data Eng. Wireless Mobile
sensor network,” in Proc. Int. Conf. Commun. Mobile Comput., 2010, Access, 2007, pp. 1–8.
pp. 162–168. [68] J. Hur, Y. Lee, H. Yoon, D. Choi, and S. Jin, “Trust evaluation model
[45] C. Lal, R. Petroccia, K. Pelekanakis, M. Conti, and J. Alves, “Toward for wireless sensor networks,” in Proc. 7th Int. Conf. Adv. Commun.
the development of secure underwater acoustic networks,” IEEE J. Technol., 2005, pp. 491–496.
Ocean. Eng., vol. 42, no. 4, pp. 1075–1087, 2017. [69] G. Han, J. Jiang, L. Shu, and M. Guizani, “An attack-resistant trust
[46] R. W. Anwar, M. Bakhtiari, A. Zainal, A. H. Abdullah, and model based on multidimensional trust metrics in underwater acoustic
K. N. Qureshi, “Security issues and attacks in wireless sensor network,” sensor network,” IEEE Trans. Mobile Comput., vol. 14, no. 12,
World Appl. Sci. J., vol. 30, no. 10, pp. 1224–1227, 2014. pp. 2447–2459, Dec. 2015.
[47] A. P. Das and S. M. Thampi, “Secure communication in mobile [70] S. Ganeriwal, L. K. Balzano, and M. B. Srivastava, “Reputation-based
underwater wireless sensor networks,” in Proc. Int. Conf. Adv. Comput., framework for high integrity sensor networks,” ACM Trans. Sensor
Commun. Inform. (ICACCI), 2015, pp. 2164–2173. Netw., vol. 4, no. 3, pp. 1–37, 2008.
[48] R. Zhu, A. Boukerche, and Q. Yang, “An efficient secure and adap- [71] J. Du, G. Han, C. Lin, and M. Martinez-Garcia, “ITrust: An anomaly-
tive routing protocol based on GMM-HMM-LSTM for Internet of resilient trust model based on isolation forest for underwater acoustic
Underwater Things,” IEEE Internet Things J., early access, Jan. 16, sensor networks,” IEEE Trans. Mobile Comput., vol. 21, no. 5,
2024, doi: 10.1109/JIOT.2024.3354820. pp. 1684–1696, May 2022.
[49] R. Jadhav and V. Vatsala, “Security issues and solutions in wireless [72] G. Han, Y. He, J. Jiang, N. Wang, M. Guizani, and J. A. Ansere,
sensor networks,” Int. J. Comput. Appl., vol. 162, no. 2, pp. 14–19, “A synergetic trust model based on SVM in underwater acoustic
2017. sensor networks,” IEEE Trans. Veh. Technol., vol. 68, no. 11,
[50] M. M. Arifeen, D. Bhakta, S. R. H. Remu, M. M. Islam, M. Mahmud, pp. 11239–11247, Nov. 2019.
and M. S. Kaiser, “Hidden Markov model based trust management [73] Y. He, G. Han, J. Jiang, H. Wang, and M. Martinez-Garcia, “A trust
model for underwater wireless sensor networks,” in Proc. Int. Conf. update mechanism based on reinforcement learning in underwater
Comput. Adv., 2020, pp. 1–5. acoustic sensor networks,” IEEE Trans. Mobile Comput., vol. 21, no. 3,
[51] Y. Su, S. Ma, H. Zhang, Z. Jin, and X. Fu, “A redeemable SVM- pp. 811–821, Mar. 2022.
DS fusion-based trust management mechanism for underwater acoustic [74] A. Boukerche and X. Li, “An agent-based trust and reputation man-
sensor networks,” IEEE sensors J., vol. 21, no. 22, pp. 26161–26174, agement scheme for wireless sensor networks,” in Proc. IEEE Glob.
Nov. 2021. Telecommun. Conf., 2005, p. 5.
[52] A. Almutairi, Y. He, and S. Furnell, “A multi-level trust framework for [75] A. Signori, F. Campagnaro, I. Nissen, and M. Zorzi, “Channel-based
the Internet of Underwater Things,” in Proc. IEEE Int. Conf. Cyber trust model for security in underwater acoustic networks,” IEEE
Secur. Resil. (CSR), 2022, pp. 370–375. Internet Things J., vol. 9, no. 20, pp. 20479–20491, Oct. 2022.
[53] D. Tronchin et al., “A secure cross-layer communication stack for [76] Y. Ren and A. Boukerche, “Modeling and managing the trust for
underwater acoustic networks,” in Proc. OCEANS, 2021, pp. 1–8. wireless and mobile Ad Hoc networks,” in Proc. IEEE Int. Conf.
[54] A. P. Nasution, V. Suryani, and A. A. Wardana, “IoT object security Commun., 2008, pp. 2129–2133.
towards on-off attack using trustworthiness management,” in Proc. 8th [77] A. Boukerche and Y. Ren, “A trust-based security system for ubiquitous
Int. Conf. Inf. Commun. Technol. (ICoICT), 2020, pp. 1–6. and pervasive computing environments,” Comput. Commun., vol. 31,
[55] G. Han, Y. He, J. Jiang, H. Wang, Y. Peng, and K. Fan, “Fault-tolerant no. 18, pp. 4343–4351, 2008.
trust model for hybrid attack mode in underwater acoustic sensor [78] J.-H. Cho, A. Swami, and R. Chen, “A survey on trust management
networks,” IEEE Netw., vol. 34, no. 5, pp. 330–336, Sep./Oct. 2020. for mobile Ad Hoc networks,” IEEE Commun. Surveys Tuts., vol. 13,
[56] A. Boukerche, A. Martirosyan, and R. Pazzi, “An inter-cluster com- no. 4, pp. 562–583, 4th Quart., 2011.
munication based energy aware and fault tolerant protocol for wireless [79] H. Son, N. Kang, B. Gwak, and D. Lee, “An adaptive IoT trust
sensor networks,” Mobile Netw. Appl., vol. 13, pp. 614–626, Jul. 2008. estimation scheme combining interaction history and stereotypical
[57] Z. Wang et al., “Cyber-physical systems for water sustainability: reputation,” in Proc. 14th IEEE Annu. Consum. Commun. Netw. Conf.
Challenges and opportunities,” IEEE Commun. Mag., vol. 53, no. 5, (CCNC), 2017, pp. 349–352.
pp. 216–222, May 2015. [80] P. Michiardi and R. Molva, “Core: A collaborative reputation mech-
[58] R. Zhu, A. Boukerche, X. Huang, and Q. Yang, “GHL-SAR: Secure anism to enforce node cooperation in mobile Ad Hoc networks,” in
and adaptive routing based on GMM-HMM-LSTM for UASNs,” in Advanced Communications and Multimedia Security. Boston, MA,
Proc. IEEE Glob. Commun. Conf., 2023, pp. 4540–4545. USA: Springer, 2002, pp. 107–121.
[59] G. De La Torre, P. Rad, and K.-K. R. Choo, “Driverless vehicle [81] J. Granatyr, V. Botelho, O. R. Lessing, E. E. Scalabrin, J.-P. Barthès,
security: Challenges and future research opportunities,” Future Gener. and F. Enembreck, “Trust and reputation models for multiagent
Comput. Syst., vol. 108, pp. 1092–1111, Jul. 2020. systems,” ACM Comput. Surv., vol. 48, no. 2, pp. 1–42, 2015.
[60] T. Dargahi, H. H. Javadi, and H. Shafiei, “Securing underwater sensor [82] J. Ryu and S. Kim, “Reputation-based opportunistic routing protocol
networks against routing attacks,” Wireless Personal. Commun., vol. 96, using Q-learning for MANET attacked by malicious nodes,” IEEE
pp. 2585–2602, May 2017. Access, vol. 11, pp. 47701–47711, 2023.
[61] C. Lal, R. Petroccia, M. Conti, and J. Alves, “Secure underwater [83] F. Hendrikx, K. Bubendorfer, and R. Chard, “Reputation systems:
acoustic networks: Current and future research directions,” in Proc. A survey and taxonomy,” J. Parallel Distrib. Comput., vol. 75,
IEEE 3rd Underwater Commun. Netw. Conf. (UComms), 2016, pp. 1–5. pp. 184–197, 2015.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2574 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

[84] S. Misra, S. Dash, M. Khatua, A. V. Vasilakos, and M. S. Obaidat, [107] S. Yang, X. Liu, and Y. Su, “A traffic-aware fair MAC protocol for
“Jamming in underwater sensor networks: Detection and mitigation,” layered data collection oriented underwater acoustic sensor networks,”
IET Commun., vol. 6, no. 14, pp. 2178–2188, 2012. Remote. Sens., vol. 15, no. 6, p. 1501, 2023. [Online]. Available:
[85] A. Boukerche, Algorithms and Protocols for Wireless Sensor Networks. https://doi.org/10.3390/rs15061501
Hoboken, NJ, USA: Wiley, 2009. [108] R. Zhu, X. Huang, X. Huang, D. Li, and Q. Yang, “An on-site-
[86] A. Gkikopouli, G. Nikolakopoulos, and S. Manesis, “A survey on based opportunistic routing protocol for scalable and energy-efficient
underwater wireless sensor networks and applications,” in Proc. 20th underwater acoustic sensor networks,” Appl. Sci., vol. 12, no. 23, 2022,
Mediterr. Conf. control Autom. (MED), 2012, pp. 1147–1154. Art. no. 12482.
[87] A. Boukerche, R. Zhu, and Q. Yang, “T-SAPR: An efficient Q-learning [109] M. Zhou et al., “Underwater acoustic channel modeling under differ-
trust-based secure routing protocol for underwater acoustic sensor ent shallow seabed topography and sediment environment,” in Proc.
networks,” in Proc. 19th ACM Int. Symp. QoS Secur. Wireless Mobile OCEANS, 2022, pp. 1–7.
Netw., 2023, pp. 1–6. [110] R. W. Coutinho, A. Boukerche, and A. A. Loureiro, “Modeling power
[88] A. Boukerche and Y. Ren, “A security management scheme using a control and anypath routing in underwater wireless sensor networks,”
novel computational reputation model for wireless and mobile ad hoc in Proc. IEEE wireless Commun. Netw. Conf. (WCNC), 2018, pp. 1–6.
networks,” in Proc. 5th ACM Symp. Perform. Eval. wireless Ad Hoc, [111] S. Climent, A. Sanchez, J. V. Capella, N. Meratnia, and J. J. Serrano,
Sensor, Ubiquitous Netw., 2008, pp. 88–95. “Underwater acoustic wireless sensor networks: Advances and future
[89] F. Moyano, C. Fernandez-Gago, and J. Lopez, “A model-driven trends in physical, MAC and routing layers,” Sensors, vol. 14, no. 1,
approach for engineering trust and reputation into software services,” pp. 795–833, 2014.
J. Netw. Comput. Appl., vol. 69, pp. 134–151, Jul. 2016. [112] T. Rahman, I. Ahmad, A. Zeb, I. Khan, G. Ali, and M. ElAffendi,
[90] J. Jiang, G. Han, L. Shu, S. Chan, and K. Wang, “A trust model based “Performance evaluation of routing protocols for underwater wireless
on cloud theory in underwater acoustic sensor networks,” IEEE Trans. sensor networks,” J. Mar. Sci. Eng., vol. 11, no. 1, p. 38, 2022.
Ind. Informat., vol. 13, no. 1, pp. 342–350, Feb. 2017. [113] M. C. Domingo, “Overview of channel models for underwater wireless
[91] J. Jiang, G. Han, C. Zhu, S. Chan, and J. J. P. C. Rodrigues, “A trust communication networks,” Phys. Commun., vol. 1, no. 3, pp. 163–182,
cloud model for underwater wireless sensor networks,” IEEE Commun. 2008.
Mag., vol. 55, no. 3, pp. 110–116, Mar. 2017. [114] R. W. Coutinho, A. Boukerche, L. F. Vieira, and A. A. Loureiro, “A
[92] J. Du, G. Han, C. Lin, and M. Martínez-García, “LTrust: An adaptive novel void node recovery paradigm for long-term underwater sensor
trust model based on LSTM for underwater acoustic sensor networks,” networks,” Ad Hoc Netw., vol. 34, pp. 144–156, Nov. 2015.
IEEE Trans. Wireless Commun., vol. 21, no. 9, pp. 7314–7328, [115] I. Ullah, J. Chen, X. Su, C. Esposito, and C. Choi, “Localization and
Sep. 2022. detection of targets in underwater wireless sensor using distance and
[93] R. Diamant, P. Casari, and S. Tomasin, “Cooperative authentication in angle based algorithms,” IEEE Access, vol. 7, pp. 45693–45704, 2019.
underwater acoustic sensor networks,” IEEE Trans. Wireless Commun., [116] X. Su, I. Ullah, X. Liu, D. Choi et al., “A review of underwater local-
vol. 18, no. 2, pp. 954–968, Feb. 2019. ization techniques, algorithms, and challenges,” J. Sensors, vol. 2020,
[94] A. Boukerch, L. Xu, and K. El-Khatib, “Trust-based security for Jan. 2020, Art. no. 6403161.
wireless Ad Hoc and sensor networks,” Comput. Commun., vol. 30, [117] G. Han, L. Liu, J. Jiang, L. Shu, and J. J. Rodrigues, “A collaborative
nos. 11–12, pp. 2413–2427, 2007. secure localization algorithm based on trust model in underwater
[95] R. Zhu, A. Boukerche, X. Huang, and Q. Yang, “DESLR: Energy- wireless sensor networks,” Sensors, vol. 16, no. 2, p. 229, 2016.
efficient and secure layered routing based on channel-aware trust model
[118] A. Fotouhi et al., “Survey on UAV cellular communications: Practical
for UASNs,” Comput. Netw., vol. 234, Oct. 2023, Art. no. 109939.
aspects, standardization advancements, regulation, and security chal-
[96] R. Zhu, A. Boukerche, L. Feng, and Q. Yang, “A trust management-
lenges,” IEEE Commun. Surveys Tuts., vol. 21, no. 4, pp. 3417–3442,
based secure routing protocol with AUV-aided path repairing for
4th Quart., 2019.
underwater acoustic sensor networks,” Ad Hoc Netw., vol. 149,
[119] A. Caruso, F. Paparella, L. F. M. Vieira, M. Erol, and M. Gerla,
Oct. 2023, Art. no. 103212.
“The meandering current mobility model and its impact on underwater
[97] A. Boukerche and S. Nikoletseas, “Protocols for data propagation in
mobile sensor networks,” in Proc. 27th Conf. Comput. Commun., 2008,
wireless sensor networks,” in Wireless Communications Systems and
pp. 221–225.
Networks. Boston, MA, USA: Springer, 2004, pp. 23–51.
[98] J. Luo, Y. Chen, M. Wu, and Y. Yang, “A survey of routing protocols for [120] L. M. Brekhovskikh, Y. P. Lysanov, and J. P. Lysanov, Fundamentals
underwater wireless sensor networks,” IEEE Commun. Surveys Tuts., of Ocean Acoustics. New York, NY, USA: Springer, 2003.
vol. 23, no. 1, pp. 137–160, 1st Quart., 2021. [121] A. Boukerche and S. E. Nikoletseas, “Energy-efficient algorithms in
[99] K. Rehan and G. Qiao, “A survey of underwater acoustic communi- wireless sensor networks,” in Algorithms and Protocols for Wireless
cation and networking techniques,” Res. J. Appl. Sci. Eng. Technol, Sensor Networks. New York, NY, USA: Wiley, 2008.
vol. 5, no. 3, pp. 778–789, 2013. [122] B. Zhao, Y. Liu, X. Li, J. Li, and J. Zou, “TrustBlock: An adaptive trust
[100] T. Wang, D. Zhao, S. Cai, W. Jia, and A. Liu, “Bidirectional evaluation of SDN network nodes based on double-layer blockchain,”
prediction-based underwater data collection protocol for end-edge- PLoS One, vol. 15, no. 3, 2020, Art. no. e0228844.
cloud orchestrated system,” IEEE Trans. Ind. Informat., vol. 16, no. 7, [123] J. Wang, X. Jing, Z. Yan, Y. Fu, W. Pedrycz, and L. T. Yang, “A survey
pp. 4791–4799, Jul. 2020. on trust evaluation based on machine learning,” ACM Comput. Surv.,
[101] Y. Sun, M. Zheng, X. Han, S. Li, and J. Yin, “Adaptive clustering vol. 53, no. 5, pp. 1–36, 2020.
routing protocol for underwater sensor networks,” Ad Hoc Netw., [124] M. Momani and S. Challa, “Survey of trust models in different network
vol. 136, Nov. 2022, Art. no. 102953. domains,” 2010, arXiv:1010.0168.
[102] S. Fattah, A. Gani, I. Ahmedy, M. Y. I. Idris, and I. A. Targio Hashem, [125] Q. Wu, P. Sun, and A. Boukerche, “Unmanned aerial vehicle-assisted
“A survey on underwater wireless sensor networks: Requirements, energy-efficient data collection scheme for sustainable wireless sensor
taxonomy, recent advances, and open research challenges,” Sensors, networks,” Comput. Netw., vol. 165, Dec. 2019, Art. no. 106927.
vol. 20, no. 18, p. 5393, 2020. [126] A. Martirosyan and A. Boukerche, “Performance evaluation of an
[103] Y. Su, Y. Liu, R. Fan, L. Li, M. Han, and H. Zhang, “A secure energy-aware clustering protocol for wireless sensor networks,” in
relay selection scheme based on cooperative jamming for underwater Proc. Int. Conf. Parallel Process. Workshops, 2008, pp. 67–72.
acoustic sensor networks,” Comput. Netw., vol. 217, Nov. 2022, [127] K. Prathapchandran and T. Janani, “A trust aware security mecha-
Art. no. 109307. [Online]. Available: https://doi.org/10.1016/j.comnet. nism to detect sinkhole attack in RPL-based IoT environment using
2022.109307 random forest–RFTRUST,” Comput. Netw., vol. 198, Oct. 2021,
[104] X. Zhang, J.-H. Cui, S. Das, M. Gerla, and M. Chitre, “Underwater Art. no. 108413.
wireless communications and networks: Theory and application,” IEEE [128] J. Du, G. Han, and C. Lin, “An edge-computing-enabled trust mecha-
Commun. Mag., vol. 54, no. 2, pp. 30–31, Feb. 2016. nism for underwater acoustic sensor networks,” IEEE Commun. Stand.
[105] R. Zhu, L. Liu, P. Li, N. Chen, L. Feng, and Q. Yang, “DC-MAC: A Mag., vol. 6, no. 1, pp. 44–51, Mar. 2022.
delay-aware and collision-free MAC protocol based on game theory [129] D. Keum and Y.-B. Ko, “Trust-based intelligent routing protocol with
for underwater wireless sensor networks,” IEEE Sensors J., vol. 24, Q-learning for mission-critical wireless sensor networks,” Sensors,
no. 5, pp. 6930–6941, Mar. 2024. vol. 22, no. 11, p. 3975, 2022.
[106] K.-Y. Tsao, T. Girdler, and V. G. Vassilakis, “A survey of cyber [130] T. Khan and K. Singh, “RTM: Realistic weight-based reliable trust
security threats and solutions for UAV communications and flying ad- model for large scale WSNs,” Wireless Personal Commun., vol. 129,
hoc networks,” Ad Hoc Netw., vol. 133, Aug. 2022, Art. no. 102894. no. 2, pp. 953–991, 2023.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 ZHU et al.: DESIGN GUIDELINES ON TRUST MANAGEMENT FOR UWSNs 2575

[131] S. Hua, J. Jiang, and G. Han, “A lightweight trust management [153] A. Josang, R. Hayward, and S. Pope, “Trust network analysis with
mechanism based on conflict adjudication in underwater acoustic subjective logic,” in Proc. Conf. 29th Australas. Comput. Sci. Conf.
sensor networks,” in Proc. Comput., Commun. IoT Appl. (ComComAp), (ACSW), 2006, pp. 85–94.
2021, pp. 258–262. [154] S.-j. Ji et al., “A pre-evolutionary advisor list generation strategy for
[132] M. Singh, A. R. Sardar, K. Majumder, and S. K. Sarkar, “A lightweight robust defensing reputation attacks,” Knowl.-Based Syst., vol. 103,
trust mechanism and overhead analysis for clustered WSN,” IETE J. pp. 1–18, Jul. 2016.
Res., vol. 63, no. 3, pp. 297–308, 2017. [155] S. Bhattacharya et al., “Blockchain for Internet of Underwater
[133] A. Signori, E. Coccolo, F. Campagnaro, I. Nissen, and M. Zorzi, Things: State-of-the-art, applications, challenges, and future direc-
“Trustworthiness in the GUWMANET protocol for underwater acoustic tions,” Sustainability, vol. 14, no. 23, p. 15659, 2022.
mobile ad-hoc networks,” in Proc. 15th Int. Conf. Underwater Netw. [156] Z. Yang, K. Yang, L. Lei, K. Zheng, and V. C. Leung, “Blockchain-
Syst., 2021, pp. 1–7. based decentralized trust management in vehicular networks,” IEEE
[134] T. Cheng, G. Liu, Q. Yang, and J. Sun, “Trust assessment in vehicular Internet Things J., vol. 6, no. 2, pp. 1495–1505, 2018.
social network based on three-valued subjective logic,” IEEE Trans. [157] X. Jian, P. Leng, Y. Wang, M. Alrashoud, and M. S. Hossain,
Multimedia, vol. 21, no. 3, pp. 652–663, 2019. “Blockchain-empowered trusted networking for unmanned aerial vehi-
[135] V. Krishnaswamy and S. S. Manvi, “Trusted node selection in clusters cles in the B5G era,” IEEE Netw., vol. 35, no. 1, pp. 72–77,
for underwater wireless acoustic sensor networks using fuzzy logic,” Jan./Feb. 2021.
Phys. Commun., vol. 47, Aug. 2021, Art. no. 101388. [158] Z. Jin, Z. Ji, and Y. Su, “An evidence theory based opportunistic routing
[136] S. Saha and R. Arya, “ARCMT: Anchor node-based range free coop- protocol for underwater acoustic sensor networks,” IEEE Access, vol. 6,
erative multi trusted secure underwater localization using fuzzifier,” pp. 71038–71047, 2018.
Comput. Commun., vol. 193, pp. 246–265, Sep. 2022. [159] Y. L. Sun, W. Yu, Z. Han, and K. R. Liu, “Information theoretic
[137] L. Yang, Y. Lu, S. X. Yang, Y. Zhong, T. Guo, and Z. Liang, “An framework of trust modeling and evaluation for ad hoc networks,” IEEE
evolutionary game-based secure clustering protocol with fuzzy trust J. Sel. Areas Commun., vol. 24, no. 2, pp. 305–317, Feb. 2006.
evaluation and outlier detection for wireless sensor networks,” IEEE [160] X. Yin and S. Li, “Trust evaluation model with entropy-based weight
Sensors J., vol. 21, no. 12, pp. 13935–13947, Jun. 2021. assignment for malicious node’s detection in wireless sensor networks,”
[138] Z. Tian, X. Gao, S. Su, J. Qiu, X. Du, and M. Guizani, “Evaluating EURASIP J. Wireless Commun. Netw., vol. 2019, pp. 1–10, Aug. 2019.
reputation management schemes of Internet of Vehicles based on [161] J. Zhao, J. Huang, and N. Xiong, “An effective exponential-based trust
evolutionary game theory,” IEEE Trans. Veh. Technol., vol. 68, no. 6, and reputation evaluation system in wireless sensor networks,” IEEE
pp. 5971–5980, Jun. 2019. Access, vol. 7, pp. 33859–33869, 2019.
[139] M. M. Arifeen, A. Al Mamun, T. Ahmed, M. S. Kaiser, and [162] D. Hongjun, J. Zhiping, and D. Xiaona, “An entropy-based trust
M. Mahmud, “A blockchain-based scheme for sybil attack detection modeling and evaluation for wireless sensor networks,” in Proc. Int.
in underwater wireless sensor networks,” in Proc. Int. Conf. Trends Conf. Embed. Softw. Syst., 2008, pp. 27–34.
Comput. Cogn. Eng., 2021, pp. 467–476. [163] B. Ma, “Cross-layer trust model and algorithm of node selection in
[140] S. Abbas, H. Nasir, A. Almogren, A. Altameem, and N. Javaid, wireless sensor networks,” in Proc. Int. Conf. Commun. Softw. Netw.,
“Blockchain based privacy preserving authentication and malicious 2009, pp. 812–815.
node detection in Internet of Underwater Things (IoUT) networks,” [164] M. Waqas, S. Tu, Z. Halim, S. U. Rehman, G. Abbas, and Z. H. Abbas,
IEEE Access, vol. 10, pp. 113945–113955, 2022. “The role of artificial intelligence and machine learning in wireless
[141] R. Muthukkumar and D. Manimegalai, “Secured transmission using networks security: Principle, practice and challenges,” Artif. Intell. Rev.,
trust strategy-based dynamic Bayesian game in underwater acoustic vol. 55, no. 7, pp. 5215–5261, 2022.
sensor networks,” J. Ambient Intell. Human. Comput., vol. 12, [165] U. Jayasinghe, G. M. Lee, T.-W. Um, and Q. Shi, “Machine learning
pp. 2585–2600, Feb. 2021. based trust computational model for IoT services,” IEEE Trans. Sustain.
[142] W. Meng, K.-K. R. Choo, S. Furnell, A. V. Vasilakos, and C. W. Probst, Comput., vol. 4, no. 1, pp. 39–52, Jan.–Mar. 2019.
“Towards Bayesian-based trust management for insider attacks in [166] S. Deng, L. Huang, G. Xu, X. Wu, and Z. Wu, “On deep learning for
healthcare software-defined networks,” IEEE Trans. Netw. Service trust-aware recommendations in social networks,” IEEE Trans. Neural
Manag., vol. 15, no. 2, pp. 761–773, Jun. 2018. Netw. Learn. Syst., vol. 28, no. 5, pp. 1164–1177, May 2017.
[143] W. Fang, C. Zhang, Z. Shi, Q. Zhao, and L. Shan, “BTRES: Beta-based [167] M. Martínez-García, Y. Zhang, K. Suzuki, and Y.-D. Zhang, “Deep
trust and reputation evaluation system for wireless sensor networks,” recurrent entropy adaptive model for system reliability monitoring,”
J. Netw. Comput. Appl., vol. 59, pp. 88–94, Jan. 2016. IEEE Trans. Ind. Informat., vol. 17, no. 2, pp. 839–848, Feb. 2021.
[144] Y. Su, S. Mal, Z. Jin, X. Fu, Y. Li, and X. Liu, “A trust model [168] O. Mistry, A. Gürsel, and S. Sen, “Comparing trust mechanisms for
for underwater acoustic sensor networks based on fast link quality monitoring aggregator nodes in sensor networks,” in Proc. 8th Conf.
assessment,” in Proc. Glob. Oceans, 2020, pp. 1–6. AAMAS, 2009, pp. 985–992.
[145] M. Momani and S. Challa, “GTRSSN: Gaussian trust and reputation [169] D. Kim and S. Park, “Reinforcement learning-based dynamic adapta-
system for sensor networks,” in Proc. Adv. Comput. Inf. Sci. Eng., 2008, tion planning method for architecture-based self-managed software,”
pp. 343–347. in Proc. ICSE Workshop Softw. Eng. Adapt. Self-Manag. Syst., 2009,
[146] R. K. Sinha and A. K. Jagannatham, “Gaussian trust and reputation pp. 76–85.
for fading MIMO wireless sensor networks,” in Proc. IEEE Int. Conf. [170] A. Vijaya Kumarm and A. Jeyapal, “Self-adaptive trust based ABR
Electron., Comput. Commun. Technol. (CONECCT), 2014, pp. 1–6. protocol for MANETs using Q-learning,” Sci. World J., vol. 2014,
[147] M. P. Lowney, H. Liu, and E. Chabot, “Trust management in under- Aug. 2014, Art. no. 452362.
water acoustic MANETs based on cloud theory using multi-parameter [171] N. A. Khalid, Q. Bai, and A. Al-Anbuky, “Adaptive trust-based
metrics,” in Proc. Int. Carnahan Conf. Secur. Technol. (ICCST), 2018, routing protocol for large scale WSNs,” IEEE Access, vol. 7,
pp. 1–5. pp. 143539–143549, 2019.
[148] G. Han, J. Du, C. Lin, H. Wu, and M. Guizani, “An energy-balanced [172] T. Ning, Y. Liu, Z. Yang, and H. Wu, “Incentive mechanisms for data
trust cloud migration scheme for underwater acoustic sensor networks,” dissemination in autonomous mobile social networks,” IEEE Trans.
IEEE Trans. Wireless Commun., vol. 19, no. 3, pp. 1636–1649, Mobile Comput., vol. 16, no. 11, pp. 3084–3099, Nov. 2017.
Mar. 2020. [173] C. S. Chin et al., “System design of underwater battery power system
[149] M. Ylianttila et al., “6G white paper: Research challenges for trust, for marine and offshore industry,” J. Energy Stor., vol. 21, pp. 724–740,
security and privacy,” 2020, arXiv:2004.11665. Feb. 2019.
[150] S. Dahmane, C. A. Kerrache, N. Lagraa, and P. Lorenz, “WeiSTARS: [174] J. Cho et al., “Seawater battery-based wireless marine buoy system
A weighted trust-aware relay selection scheme for VANET,” in Proc. with battery degradation prediction and multiple power optimization
IEEE Int. Conf. Commun. (ICC), 2017, pp. 1–6. capabilities,” IEEE Access, vol. 9, pp. 104104–104114, 2021.
[151] T. Qin, H. Yu, C. Leung, Z. Shen, and C. Miao, “Towards a trust [175] Z. Shen, H. Yin, L. Jing, X. Ji, Y. Liang, and J. Wang,
aware cognitive radio architecture,” ACM SIGMOBILE Mobile Comput. “A power control-aided Q-learning-based routing protocol for
Commun. Rev., vol. 13, no. 2, pp. 86–95, 2009. optical-acoustic hybrid underwater sensor networks,” IEEE
[152] F. Firoozi, V. I. Zadorozhny, and F. Y. Li, “Subjective logic-based Trans. Green Commun. Netw., vol. 7, no. 4, pp. 2117–2129,
in-network data processing for trust management in collocated and Dec. 20233.
distributed wireless sensor networks,” IEEE Sensors J., vol. 18, no. 15, [176] Z. Shen, H. Yin, L. Jing, Y. Liang, and J. Wang, “A cooperative
pp. 6446–6460, Aug. 2018. routing protocol based on Q-learning for underwater optical-acoustic

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.
 2576 IEEE COMMUNICATIONS SURVEYS & TUTORIALS, VOL. 26, NO. 4, FOURTH QUARTER 2024

hybrid wireless sensor networks,” IEEE Sensors J., vol. 22, no. 1, Libo Long received the undergraduate degree in
pp. 1041–1050, Jan. 2022. computer science from Carleton University, Ottawa,
[177] N. Schear, P. T. Cable, T. M. Moyer, B. Richard, and R. Rudd, Canada. He is currently pursuing the Ph.D. degree
“Bootstrapping and maintaining trust in the cloud,” in Proc. 32nd Annu. in computer science with the University of Ottawa,
Conf. Comput. Secur. Appl., 2016, pp. 65–77. Canada. His research interests include network
[178] B. Hurl, R. Cohen, K. Czarnecki, and S. Waslander, “TruPercept: security and video frame interpolation.
Trust modelling for autonomous vehicle cooperative perception from
synthetic data,” in Proc. IEEE Intell. Veh. Symp., 2020, pp. 341–347.
[179] R. Urena, G. Kou, Y. Dong, F. Chiclana, and E. Herrera-Viedma,
“A review on trust propagation and opinion dynamics in social
networks and group decision making frameworks,” Inf. Sci., vol. 478,
pp. 461–475, Apr. 2019.
[180] X. Fan, L. Liu, R. Zhang, Q. Jing, and J. Bi, “Decentralized trust
management: Risk analysis and trust aggregation,” ACM Comput. Surv.,
vol. 53, no. 1, pp. 1–33, 2020.

Rongxin Zhu (Member, IEEE) received the M.E.

degree in software engineering from Nanjing
University, Nanjing, China, in 2017. He is currently
pursuing the Ph.D. degree in cyberspace security
with Hainan University. He was a Visiting Scholar
with the University of Ottawa for one year. His
research interests include the security of underwater
sensor networks, protocol design, and intelligent
algorithms.

Azzedine Boukerche (Fellow, IEEE) is a

Distinguished University Professor and holds a
Canada Research Chair Tier-1 Position with the
University of Ottawa. He is the Founding Director of
the PARADISE Research Laboratory and the DIVA
Strategic Research Center, and NSERC-CREATE
TRANSIT with the University of Ottawa. His
current research interests include sustainable sensor
networks, autonomous and connected vehicles,
wireless networking and mobile computing, wireless
multimedia, QoS service provisioning, performance
evaluation and modeling of large-scale distributed and mobile systems, and
large scale distributed and parallel discrete event simulation. He has published
extensively in these areas and received several best research paper awards
Qiuling Yang received the B.E. degree from
for his work. He has received the C. Gotlieb Computer Medal Award, the
Shenyang Aerospace University, Shenyang, China,
Ontario Distinguished Researcher Award, the Premier of Ontario Research
in 2003, the M.E. degree from Guangxi University,
Excellence Award, the G. S. Glinski Award for Excellence in Research,
Nanning, China, in 2010, and the Ph.D. degree
the IEEE Computer Society Golden Core Award, the IEEE CS-Meritorious
from Tianjin University, Tianjin, China, in 2016. She
Award, the IEEE TCPP Technical Achievement and Leaderships Award, the
is currently a Professor and a Doctoral Supervisor
IEEE ComSoc ComSoft and IEEE ComSoc Technical Achievement Award,
with Hainan University, Haikou, China. Her research
and the University of Ottawa Award for Excellence in Research. He serves as
interests include protocol design and security of
an Editor-in-Chief for ACM ICPS and an associate editor for several IEEE
underwater acoustics sensor networks.
transactions and ACM journals, and is also a Steering Committee Chair for
several IEEE and ACM international conferences. He is a Fellow of the
Engineering Institute of Canada, the Canadian Academy of Engineering, and
the American Association for the Advancement of Science.

Authorized licensed use limited to: NORTHWESTERN POLYTECHNICAL UNIVERSITY. Downloaded on March 19,2025 at 04:26:50 UTC from IEEE Xplore. Restrictions apply.