A comprehensive review of significant learning for anomalous transaction detection using a machine learning method in a decentralized blockchain network

Abstract

Blockchain is a distributed ledger technology (DLT) that enables decentralized applications (DApps) such as
Hyperledger, Bitcoin, Ethereum, decentralized finance (DeFi), and non-fungible token (NFT) to operate on it. Bitcoin is
a popular application that leverages blockchain technology (BT) to provide secure transactions in a decentralized
environment. Among the main reasons for BT being utilized is to eliminate the dependence of the process on third parties
such as lawyers, insurance firms, and banks to execute approvals, verifications, and signatures. Due to the immutability
and transparency of this technology, it has been deployed in several industries, including agri-food, supply chain,
automotive, pharma, healthcare, insurance, land registration, and higher education, to increase verification, security, and
traceability. Modern research indicates that blockchain distributed ledger may still be susceptible to various privacy,
security, and dependability challenges, despite their impressive characteristics. To resolve these challenges, it is essential
to notice abnormal behavior in a timely manner. Consequently, machine learning (ML) approaches with anomaly
detection play a comprehensive role in preventing fraud. This paper explores the technological integration of anomaly
detection models into BT and compares supervised and unsupervised ML techniques for detecting rogue and legitimate
transactions. The studies for the review come from three well-known databases: Web of Science (WoS), Google Scholar
and Scopus. Using sophisticated search tactics, specialized keywords such as Bitcoin, Ethereum, blockchain, fraud
detection, anomaly detection, data mining, and ML are employed. These findings are based on three main points (1) the
background of BT (2) Blockchain integration with ML, and (3) the ML approach for supervised learning and
unsupervised learning. According to the findings of this study, supervised learning is the most prevalent method for
studying anomalous detection in blockchain networks. This study also equips researchers with the knowledge necessary to
perform research on the detection of blockchain network anomalies using ML techniques.

Full text

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
ISSN (Print): 2394-5443 ISSN (Online): 2394-7454
http://dx.doi.org/10.19101/IJATEE.2021.876322
1366
A comprehensive review of significant learning for anomalous transaction
detection using a machine learning method in a decentralized blockchain
network
Sabri Hisham*, Mokhairi Makhtar and Azwa Abdul Aziz
Faculty of Informatic and Computing, Universiti Sultan Zainal Abidin,22000 Besut,Terengganu, Malaysia
Received: 04-July-2022; Revised: 20-October-2022; Accepted: 21-October-2022
©2022 Sabri Hisham et al. This is an open access article distributed under the Creative Commons Attribution (CC BY) License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.Introduction
Blockchain technology has changed the global
trading of assets. A blockchain can be viewed as a
connected ledger managed by a distributed peer-to-
peer (P2P) network. Blockchain offers distinctive
characteristics such as transactional privacy, the
immutability of data, transparency and cryptographic,
among others. These features paved the door for
blockchain to develop numerous technology solution,
including voting applications [1,2], internet of things
(IoT) [3,4], and supply chain management (SCM)
[5,6], among others. The increasing desire for
technological advancements stimulated the
development of BT.
*Author for correspondence
A blockchain logs the transfer of assets during
transactions and arranges them into blocks.
Cryptographic algorithms connect blocks to their
predecessors, establishing a blockchain. Satoshi
Nakamoto popularized the blockchain in 2008 as the
public ledger of Bitcoin transactions [7].
Bitcoin is the first popular blockchain technology
(BT) and the first real implementation of a
cryptocurrency ecosystem. Ethereum is the principal
distributed blockchain network that now supports
digital smart contracts and the second-largest
cryptocurrency, known as Ether. Despite these
advantages, BT is susceptible to certain assaults and
problems. Security concerns and weaknesses, such as
majority assaults [8], weak smart contracts [9],
eclipse attacks [10], and phishing schemes [11], have
Review Article
Abstract
Blockchain is a distributed ledger technology (DLT) that enables decentralized applications (DApps) such as
Hyperledger, Bitcoin, Ethereum, decentralized finance (DeFi), and non-fungible token (NFT) to operate on it. Bitcoin is
a popular application that leverages blockchain technology (BT) to provide secure transactions in a decentralized
environment. Among the main reasons for BT being utilized is to eliminate the dependence of the process on third parties
such as lawyers, insurance firms, and banks to execute approvals, verifications, and signatures. Due to the immutability
and transparency of this technology, it has been deployed in several industries, including agri-food, supply chain,
automotive, pharma, healthcare, insurance, land registration, and higher education, to increase verification, security, and
traceability. Modern research indicates that blockchain distributed ledger may still be susceptible to various privacy,
security, and dependability challenges, despite their impressive characteristics. To resolve these challenges, it is essential
to notice abnormal behavior in a timely manner. Consequently, machine learning (ML) approaches with anomaly
detection play a comprehensive role in preventing fraud. This paper explores the technological integration of anomaly
detection models into BT and compares supervised and unsupervised ML techniques for detecting rogue and legitimate
transactions. The studies for the review come from three well-known databases: Web of Science (WoS), Google Scholar
and Scopus. Using sophisticated search tactics, specialized keywords such as Bitcoin, Ethereum, blockchain, fraud
detection, anomaly detection, data mining, and ML are employed. These findings are based on three main points (1) the
background of BT (2) Blockchain integration with ML, and (3) the ML approach for supervised learning and
unsupervised learning. According to the findings of this study, supervised learning is the most prevalent method for
studying anomalous detection in blockchain networks. This study also equips researchers with the knowledge necessary to
perform research on the detection of blockchain network anomalies using ML techniques.
Keywords
Blockchain, Ethereum, Bitcoin, Machine learning, Anomaly detection, Artificial intelligence, Fraud detection.

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1367
been significant obstacles to BT. The Ethereum
blockchain has become an important platform despite
security issues such as phishing scams, which
account for about half of all Ethereum cybercrime
[12]. In addition, the distributed ledger made publicly
accessible by the Ethereum blockchain network
would undoubtedly be classified as big data
ecosystem. Furthermore, manually searching through
all of these transactions to discover any transactions
suspected of exhibiting abnormality characteristics
would be impractical or time-consuming. In theory,
machine learning approach would help distinguish
between transactions that exhibit normal and
abnormal behavior among user address by learning
the attributes that correspond to either normal or
abnormal conduct. Blockchains are susceptible to
numerous types of harmful assaults and fraudulent
behavior, which may be detectable by analyzing
transaction patterns. Consequently, BT can profit
from using machine learning (ML) algorithms via
their capacity to evaluate, learn, and improve with
large amounts of data. Detecting abnormalities has
been studied extensively for ages. Numerous
independent ML approaches have been created and
utilized for anomaly detection in various applications.
The process of identifying abnormal patterns in a
transaction is relatively difficult to detect [13].
Therefore, anomaly identification is used in various
applications. For example, anomalous detection is
accomplished by integrating an ML approach with
blockchain in a study on intruder detection in drone
technology of unmanned aerial vehicle (UAV) [14].
Other extensively used applications for identifying
cybercrime anomalies include crypto-jacking [15],
phishing [16], frauds [17], money laundering [18],
crypto-ransomware [19], and Ponzi schemes [20].
Another example is the detection of abnormalities in
blockchain-IoT-based application for the construction
of a secure framework [21], decentralized IoT data in
smart cities [21], and monitoring wastewater reuse
and electricity usage in smart grids [22].
Overall, it is crucial to note that anomaly detection is
one of the critical topics for securing future
blockchain networks and that a large amount of work
is being conducted on this subject from multiple
viewpoints, which will be discussed in this study. A
detailed evaluation of these studies includes (1) a
basic understanding of blockchain architecture, (2) a
review of blockchain and ML integration, and (3)
identifying the principal anomaly detection methods
using ML approaches (such as supervised,
unsupervised learning), datasets, metric
measurement, tools, and the types of data they
exploit.
This review is structured as follows: The second part
explains how the article selection process is carried
out. The third section provides an overview of
blockchains such as smart contracts, versions,
consensus, properties, and classifications. Section 4
and 5 gives an overview of the combination of
blockchain-based anomaly detection techniques. In
section 6, the study's findings are discussed and
section 7 shows the conclusions and directions of
future studies.
2.Method
2.1Articles selection strategy
The selection of articles related to anomaly detection
in the blockchain network involves a systematic
article selection process (see Figure 1). Thus, three
online databases, namely Scopus, Google Scholar,
and web of science, are used for searching the
articles. This process is carried out based on the steps
outlined in the Reporting Items for Systematic
Reviews and Meta-Analysis (PRISMA) for the
selection of documents (identification, screening, and
eligibility).
In general, the article search technique begins with an
extraction of relevant articles from major journal
databases (WoS, Scopus, and Google Scholar). This
search encompasses a variety of primary subject
categories, including computer science, security,
information systems, artificial intelligence (AI),
engineering, decision science, and mathematics.
However, this database has millions of articles
published in journals from over the world. Since the
primary study focuses on the detection of
abnormalities in the block chain network, the search
for papers is conducted using a search string related
to the study's title. "Anomaly detection,"
"blockchain," "Bitcoin," "Ethereum," "ML,"
"abnormal," "suspicious," "malicious," and "fraud"
are among the most popular search terms. The search
string used to locate items in the database yielded 919
results. This complete study is comprised of excerpts
from papers published between 2017 and 2022.
The extracted articles will then be recognised and
analysed, while the irrelevant articles removed.
Initially, the article search was limited to journal
articles and did not include the categories of books,
trade journals, or conference proceedings. As far as
the language of the article is concerned, the selection
includes only English-language articles. Therefore,

Sabri Hisham et al.
1368
the article version written in a language other than
English is excluded. After undergoing this filtering
procedure, 476 articles were retained while 443 were
excluded.
Finally, the paper is finalised through the eligibility
procedure. This procedure involves a final screening
that eliminates duplicate items. Out of a total of 476
items, only 133 remained after this process.
Therefore, based on the study, 343 articles were
excluded since they did not emphasise blockchain,
machine learning, or detecting whether something is
incorrect.
Figure 1 The study's data flow diagram
2.2 Workflow review analysis
This study is described in depth in accordance with
the analysis of the work flow that refers to the articles
chosen according to the subtheme (see Figure 2).
First, a comprehension of blockchain technology will
be offered. It examines blockchain architecture, the
contrast between blockchain and conventional
databases, blockchain versions, consensus processes,
blockchain classification, blockchain characteristics,
blockchain platforms, and smart contracts. The
integration of BT and ML was subsequently
analysed. This section describes how both of these
technologies can be advantageous to both parties. In
the section on anomaly detection analysis, it
describes the use of ML models to detect anomalous
blockchain transactions. This analysis includes a
review of earlier studies as well as the use of
supervised learning and unsupervised learning
techniques for anomaly identification. In spite of this,
ML model construction (training, testing)
necessitates the usage of suitable data sets, as these
determine the performance of the final model created.
Therefore, the datasets and tools section describe the
analysis of prior studies, covering the sorts of
datasets (live datasets, social media, open datasets,
and tool datasets) and the application of tools

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1369
(application programming interface (API),
development tools, software, and smart devices).
Lastly, it involves analysing the sorts of performance
indicators utilised to quantify experimental outcomes
in past investigations.
Figure 2 Workflow review analysis diagram
3.Decentralized blockchain network
3.1Blockchain architecture
Blockchain was first discussed after Satoshi
Nakamoto introduced a decentralized cryptocurrency.
The main function of blockchain is to keep updated
news entries and authorization processes for each
network participant. All the transactions are
maintained in a particular order with the collection of
blocks. The blockchain architecture consists of
private, public, and consortium [23]. In addition, the
architecture of blockchain consists mostly of the six
layers depicted in Figure 3, comprising hardware,
data, the network, the consensus, and the application.
The majority of blockchain applications are hosted on
a server in an on-premise or cloud data center and
operate on a P2P basis. Conceptually, P2P networks
operate on a large scale of connected nodes
(computers) for data sharing, validation, verification,
and storing all the transactions in the ledger [23]. On
the P2P basis, all the programs run as a client-server
concept, in which the request from client browsers is
returned with the result.
The data structure in the blockchain is stored and
structured in a list of blocks at the data layers. It
comprises of two fundamental components, namely
pointers and linked lists. Clearly, linked blocks refer
to a series of linked lists containing data pointing to
the prior block. Because each blockchain block might
include thousands of transaction records, the Merkle
(binary tree of hashes) algorithm and Merkle tree root
[24] were employed to produce the final hash value.
In general, hash values are utilized as input and
output identifiers. In the complete block, each
transaction output is only valid once used as an input
for the entire blockchain [25]. The blockchain
operates on consensus features, cryptography, and
Merkle trees. As seen in Figure 4, each block
contains the Merkle tree's root hash, as well as
critical information such as difficulty, nonce, data,
block number, timestamp and the preceding block's
hash. Thus, a Merkle tree provides security,
reliability, and incontestability. In addition, the
security and integrity of the data stored in the
distributed blockchain are assured. To reach this
point, each transaction must be digitally signed with a
secure key(private key) and validated by a signer.
A P2P network is a data transmission mechanism and
the primary network-level communication
architecture. It preserves network integrity by
allowing nodes to interact and synchronize with one
another. In a blockchain setting, these nodes consist
of numerous computers that are interconnected via a
blockchain network in order to conduct transactions.
The consensus layer is one of the crucial layers in
blockchain architecture, including Ethereum,
Hyperledger, and Bitcoin. Apart from that, this layer
is comprised of consensus algorithms mechanism
such as proof of authority (PoA), proof of work
(PoW), proof of stake (PoS) and practical byzantine
fault tolerance (PBFT). These algorithms run the
ranking process, block validation, and confirm that
all nodes reach consensus. The application layer of a
blockchain architecture consists of layers that interact
with end-users via the programming logic embedded
in a contract (smart contract in Ethereum, chain code
in Hyperledger) based on business cases, rules, and
algorithms. This application type is known as a
DApps. Smart contracts include procedures, models,
assets, rules, and business logic that function without
needing a third party with blockchain-based trust.
Typically, the application layer will communicate
with end-users through the execution layer's
application interface. These two layers interact
through the API, application framework, and smart
contracts. Concurrently, the validation process and
consensus among nodes are conducted on the
blockchain.

Sabri Hisham et al.
1370
Figure 3 Blockchain architecture
Figure 4 Block structure [14]
3.2 Blockchain and traditional database
Conceptually, blockchain and distributed ledger
technology are distributed and decentralized systems
that store structured data in public ledgers and blocks.
Pertaining to trust, blockchain eliminates using
trusted third parties on whom databases rely,
improving the content's veracity and reliability [26].
There is a variation in data structure between
blockchain and conventional databases. In contrast to
databases, blockchains store data in blocks. Over
time, databases embraced and utilized a relational
model that enabled increasingly complicated data
collection methods by linking information from
numerous databases. An administrator can modify,

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1371
manage, update, and control a database. In addition,
administrators can undertake database administration
tasks such as speed improvement, tuning, monitoring
and database size reduction. Generally, a huge
database slows down the performance index. Thus,
administrators employ optimization techniques to
increase the database's performance. A recursive
database allows you to alter or delete a record if you
have permission to do so. The contrast between
blockchains and databases is summarized in Table 1.
A blockchain, unlike a database, maintains data in the
chain of blocks, including hash code information
from previous blocks. This blockchain is extremely
safe and hack-proof due to the cryptographic
techniques that protect it. The SHA-256 algorithm is
the most widely used secure hashing algorithm and
function. Nevertheless, SHA-256 is predominantly
used for one-way encryption [27].
Table 1 Compares blockchain and databases
Criteria
Database
Blockchain
Architecture
Centralized
Decentralized/Distributed
Performance
Connected Centralized Server:
slow
Many blockchain Nodes: faster
Security
Append data to the database
without tracing
Traceable Block Transaction
Downtime for system update
When updating the server, the
system goes down.
No system downtime. Other blockchain nodes can be
updated at any moment.
Access Control
Admin has complete access to
the database.
Smart Contract: Permission or access is protected and
encrypted. No Admin
System Backup
Must do backup database:
Weekly/Monthly
Original Copied: Each blockchain node
DDOS
The entire system hangs and
shuts down.
One node down. Other nodes are running
Application Development and
Deployment
It at least requires web server
hosting services. There is no
direct programming in the
database. Only SQL
programming for data analysis.
All programming and data are in one place via smart
contract programming. No need for hosting services.
Everything is executed by blockchain nodes.
3.3 Blockchain version
Blockchain Version 1.0 was introduced by Hall
Finley in 2005, who implemented DLT, marking its
first cryptocurrency-based application. In 2008,
Satoshi Nakamoto launched Bitcoin, which leverages
BT 1.0 [28].
After the problems of digital currency being spent
twice and digital transactions being performed
without a trusted third party were resolved, Bitcoin
became a popular method of conducting financial
transactions on the blockchain network [29]. Thus,
any participant can conduct valid Bitcoin transactions
with this version, and this type is predominantly used
for currencies or payments.
In Version 1.0, Bitcoin mining was inefficient, and
the network lacked scalability. Hence, the latest
version of blockchain Version 2.0 addresses these
issues. In this iteration, smart contracts will be added
to the blockchain in improvement to cash
management. This smart contract is an executable
software that automatically verifies the previously
established requirements, such as facilitation,
verification, or enforcement, and minimizes
transaction cost efficiency. The adoption of Ethereum
after Bitcoin has resulted in an increase in
transactions on public networks in a short time. These
scenarios have changed most domains from focusing
on cryptocurrency to DApps. This led to blockchain
Version 3.0, which focused on developing DApps in
health, industry, digital government, and other areas.
Figure 5 depicts the chronology of blockchain
network versions.
Figure 5 The history of the blockchain
3.4 Blockchain consensus
Authors in [30] explains consensus algorithms that
enable a safe updating of the shared ledger. A
common method for error detection in the
decentralized ledger is distributing the shared data
over multiple network replicas. Consequentially, the
method utilized in such operations is the consensus

Sabri Hisham et al.
1372
algorithm, and it is often important for processes to
agree on the desired data value during calculation.
The consensus algorithms consist of various types,
such as PoW, PoS, PBFT, PoA and Delegated Proof
of Stake (DPoS). Table 2 compares the PoW and the
PoS.
Table 2 PoW and PoS comparison [31]
Property
PoW
PoS
Efficiency of Energy
No
Yes
Tolerated Power
Less than 25 percent computing power
Less than 51 percent stake
Hardware
Very essential
No Needed
Forking
When two nodes find an appropriate
nonce
Very difficult
Attack for double-spending
Yes
Difficult
Speed of Block Creating
Low according to the variant
Fast
Pool Mining
Yes. It is preventable
Extremely difficult to avoid
Example
Bitcoin
Nextcoin
Previous research related to consensus algorithms
was conducted by [32]. This work introduced the first
variant of the consensus algorithm named PoW. For a
related understanding, the PoW consensus can be
seen in the mining process. In general, the mining
process increases the turnover of transactions in one
block compared to another block. Each public node
in the network is eligible to be a miner for new block
validation before being added to the network.
However, the PoW consensus method consumes too
much electricity. Nonetheless, the phenomena of the
affluent getting richer may manifest in the PoS
consensus procedure. The alternate solutions are
produced in the network by the initial miner. Each
block varies in terms of estimation computation
difficulty and quantity. The data will be processed to
be stored in the blocks and converted to a unique
hash value according to the cryptographic method.
Therefore, one-way hashes are hash values that are
encrypted on one side and cannot be reversed to the
original data. Further, due to its high energy
consumption, PoW is environmentally unfavorable.
As a result of this consensus adoption of Ethereum,
PoW is now commonly employed to incorporate BT
in applications.
PoS is an algorithm comparable to PoW. In terms of
consensus mechanisms, the PoS has no competition
compared to the PoW [33]. If the given validator
cannot confirm the transaction, the network will
select the next validator and continue doing so until
another node can confirm the transaction. Miners in
PoS must demonstrate possession of the required
amount of cash. It is expected that wealthy
individuals would be less inclined to assault the
network. It implements the CASPER protocol when
in PoS. However, PoS will be implemented on
Ethereum in the near future. PoS is more efficient
and conserves more energy than PoW. As the mining
cost is almost zero, it is possible that attacks will
occur. Many blockchains begin with PoW and
transition to PoS over time.
PBFT is an algorithm that was developed in the 90s
to solve Byzantine tolerance errors. It is widely used
in BT and distributed computing. The major BT
Hyperledger employs the PBFT consensus [34]. Note
that DPoS was introduced based on the evolution of
PoS. Basically, this algorithm uses the concept of
voting to select a representative (called a witness) for
the purpose of validating the next block. These
delegates are selected by collecting tokens into
betting groups and linking them to specific delegates.
PoA operates through nodes that have been granted
authorization only to verify transactions. Therefore, it
needs to be approved by a majority of the authorities.
3.5 Blockchain classification
Current BT systems generally fall into three types:
Public BT, Private BT, and Consortium BT.In the
Public BT, anyone connected to a public network and
able to access all records is allowed to participate in
the process by consensus. These participants will be
rewarded if they have reached an agreement.
Consequently, the authors [35] explains that anyone
can join a Public BT network and engage without
authorization because it is completely open. To
encourage additional users to join the network, there
is typically an incentive structure in place. The
formation and operation of Public BT are secure.
Thus, transactions in the blockchain go through an
expensive consensus mechanism process even though
each member is free to join any node in the network.
The hash value is a cryptographic mechanism that
makes it impossible and difficult to change the
information(data) on the blocks. In addition, each
node in a blockchain network can be anonymous,

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1373
which is one of the properties that protect the privacy
of its members [36]. One of the most popular and
widely used Public BT networks is Bitcoin. Since
Bitcoin operates publicly, it has become a major
drawback because of the high processing power
required for the mining process (placing data in a
ledger). This process is carried out in PoW, which
requires consent at each node by implementing a
complex cryptographic mechanism. This concept of
openness has an impact on the absence of privacy.
For Private BT or permissioned networks, only nodes
from a single company would be permitted to join the
consensus mechanism process and offer consumers
the ultimate privacy they desire. Thus, each member
requires an invitation from the initiator of the
network or from someone who has been authorized
for access permission to the network [37]. In
addition, this is a measure to limit and control access
to members who want to join the network node to
implement the business process that has been
developed. In this case, they need to get an invitation
and permission first before accessing the network
node.
The Consortium BT is also known as the federated
blockchain, which consists of several organizations
that are administered on one platform and are
managed by one entity. The Consortium BT
established by multiple companies is somewhat
decentralized, as only a subset of nodes would be
chosen to ascertain consensus. This type's consensus
process is slower than Private BT but faster than
Public BT. There are fewer known players in a
Consortium BT. As a voting-based system, it ensures
low latency and good performance. Every node can
read or write transactions, but none may add a block.
To confirm this block, every node (or a
supermajority) must do so. If this rule is not met, the
block cannot be inserted. As for Consortium BT, it is
applicable to numerous business applications. The
blockchain framework used to develop business
applications using the Consortium BT network is
Hyperledger. Furthermore, due to enhanced security
measures and the participation of various companies,
this version of the blockchain is also more resistant to
hacking.
3.6 Blockchain features
BT has existed for a considerable amount of time and
continues to be in the public eye. Bitcoin, a
prominent cryptocurrency, initially brought the
technology to public attention. In addition,
additionally, blockchain manages data through
distributed ledgers. Any node connected to the
network will get a copy of the ledger, and it is very
different from the centralized database approach. The
failure of a centralized database could result in data
loss, and blockchain solves this issue. Transparency
of the transactions is another significant benefit. The
following properties have been identified for the BT,
as indicated in Figure 6.
Figure 6 Blockchain feature
Decentralized Technology means the network is
decentralized, meaning it is not rights by a central
authority or managed by a single individual. A node
that connects a network and then renders it
decentralized is the crucial feature of the blockchain.
As the system does not require any regulatory
authority, we can store our assets on it immediately
over the web. Thus, it may store anything, including
cryptocurrencies, vital papers, contracts, and other
digital items of value. With the aid of blockchain, we
will have direct control over them using the private
key. Thus, the decentralized system restores the
common people's authority and property rights.
Multiple computers, often known as nodes, house the
blockchain ledger. In BT, central authority is not
required to verify information on nodes in a P2P
manner[38]. Consensus protocols are a set of rules
and methods used to validate transactions and
maintain information consistency and integrity.
Among the fascinating characteristics of BT is
immutability. Also known as un-tamperability [39],
immutability refers to something that cannot be
changed or altered. The unchanging and permanent
nature of the network is an important feature in the
BT environment. The agreement needs to be made by
1.Blockchain
Features
Decentralized
Technology
Enhanced
Security
Capacity
Anonymity
Consensus-
based
Immutability

Sabri Hisham et al.
1374
the majority to change the data to make it tamper-
proof. Apart from that, the data block stores the
timestamp in the hash value to ensure the data is
permanent [40]. However, once a transaction is
approved into blocks, the data cannot be altered,
deleted, or restored to the original[41].
In terms of security enhancement, misappropriation
by certain parties can be avoided because the
blockchain eliminates the role of the central authority
to ensure that there is no modification in the network.
In addition, the protection aspect is enhanced through
the cryptography mechanism for the encryption
process. The method of using private keys (disclosed
to the owner) and public keys (disclosed to anyone) is
the practice of cryptography to ensure the level of
security of transactions is guaranteed. This key is
used as access for secure transactions with real
ownership and immutability. Each block in the
distributed ledger has its own unique hash and
includes the hash of the previous block. Changing or
attempting to alter the data would necessitate
changing all the transaction hash (Tx Hash) which is
nearly impossible. Therefore, public keys and private
keys are required to access data to perform
transactions. Additionally, the decentralization of
data structures in BT uses P2P consensus to avoid the
occurrence of failures in the data. It differs from a
centralized design that is prone to disruption and
failure.
The consensus algorithm is what makes BT effective.
It is an important component of every blockchain and
a defining feature. A simple explanation is that
consensus is a mechanism that helps make decisions
within network nodes after the agreement process is
accepted. The P2P model is practised within the
blockchain to generate democratic decision
agreements by each node. Technically, the ledger will
be copied and distributed to other node blocks by
consensus after a new or existing block addition
transaction takes place [42].
Anonymity conceals the identity of users and
maintains their identities confidential. Therefore, this
mechanism allows transaction verification to be
carried out without knowing the identity and
disclosing personal information. This is the aspect of
trust that uses algorithms for data transactions
between nodes. The information transfer is done
anonymously, and the information on the node does
not need to be disclosed.
Lastly, BT is its ability to enhance the capacity of a
whole network. Thousands of interconnected
computers can be more powerful than a few
centralized servers. The smart contract system is
another amusing fact. This can expedite the
settlement of any type of contract. This is one of the
greatest advantages and benefits of BT to date.
Digital smart contracts on the blockchain may
automatically generate transactions, make decisions,
and store data. Using particular consensus protocols
[38], all system nodes can automatically exchange
and verify data.
3.7 Blockchain platform difference
Emerging blockchain platforms are essentially
indistinguishable from core BT at this point. In a
blockchain, data structures are stored in blocks that
contain timestamp information and previous blocks.
In addition, it contains cryptographically signed
digital records that are shared through a distributed
ledger. Digital assets consist of tangible and
intangible objects such as patient records, security,
currency, etc. There are three main blockchain
platforms: Bitcoin, Ethereum, and Hyperledger.
Table 3 displays a comparison of this blockchain
platform.
Table 3 Blockchain platform
Scope
Bitcoin[7]
Ethereum
[43]
Hyperledger fabric[44]
Consensus
PoW
PoW, PoS
PBFT
Currency
BTC
Ether
None
Smart contract
Yes
Yes
None
Miner Participation
Public
Public, Private, Hybrid
Private
Operation Mode
Permission less
Permission less
Permissioned
Governance
Bitcoin Developer
Ethereum Developer
Linux Foundation
Application
Cryptocurrency
Yes
Yes
Data Access
Public Network
Public and Private Network
Authorize (Private Network)
Bitcoin is a decentralized digital currency that is
transferred between two parties without the need for intermediaries such as regulators,banking,insurance
or other financial institutions. With Bitcoin, digital or

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1375
virtual currencies known as cryptocurrencies gained
prominence. Hyperledger [44] is a multi-project open
source collaborative initiative formed by The Linux
Foundation in December 2015 to improve various
industry BT. Hyperledger Fabric is a modular
enterprise architecture platform for creating private
and permissioned blockchains. Members of the
blockchain network must register with a Membership
Service Provider (MSP). In the Hyperledger, several
channels can be created, each of which consists of a
separate ledger to be accessed by several groups of
authorized users. Ethereum is a distributed
blockchain network platform that enables a P2P
network for securely executing and verifying smart
contracts (program code). In Ethereum, a smart
contract is a programme code developed to allow
parties to transact without middlemen. Thus, all
records cannot be modified and securely distributed
across decentralized network nodes, allowing
participants to gain full ownership of data
transactions. In general, Ethereum accounts are run
on the basis of transactions between the sender and
the recipient that have been signed and spent digital
money (Ether) for each transaction. Technically,
Ethereum uses a machine state approach to manage
transactions. The transfer of transactions from state to
final state begins with the genesis state [45] that
holds the various data.
3.8 Smart contract
Szabo introduced smart contracts in 1994 [46].
Initially, smart contracts were not fully utilized until
BT became popular. Smart contracts have begun to
gain attention after the rapid development of BT in
various sectors. The adaptation of smart contract
technology has led to the norm for the preparation of
contracts in business procedures. On the other hand,
conventional contracts run transactions through a
middleman, as opposed to smart contracts that are
enforced automatically without the intervention of a
middleman. The implementation of smart contracts
has seen improved management quality, operating
cost savings, and time as well as risk reduction [47].
A smart contract consists of programme code
developed to run and simulate business functions in
the real world. The programme will be run once it
meets the set criteria or conditions. This allows the
agreement document to be executed automatically
after activation occurs in the smart contract when all
conditions are met. A smart contract is synonymous
with contracts developed in the Ethereum
environment [48]. Program code is developed in
bytecode format and execute in the environment of
the Ethereum virtual machine (EVM). The main
language(programming) for smart contract
development in the Ethereum environment is Solidity
[49]. Meanwhile, chain code is a smart contract
developed in the Hyperledger environment [21].
Among the main languages for chain code
development are Java, Go, and Node.js.
4.Integration of blockchain and ML
Blockchain is a decentralized ledger that stores data
in blocks and links them using cryptographic
techniques. Based on write-only database
capabilities, blockchain activities are implemented in
a decentralized distributed P2P environment. Such
cryptocurrencies are commonly used applications in
the blockchain realm, such as Bitcoin and Ethereum.
Nowadays, BT is increasingly widely used in various
sectors, including health, industry, education,
pharmacy, finance, and so on. As a result, it has
opened up space for irresponsible parties to hack and
misuse this technology for personal gain. Eventually,
consumers fall victim to fraud, data leakage, loss of
digital money, and so on. Despite these benefits, BT
is not hundred percent safe, and it is still prone to
attacks and vulnerabilities [50]. For example, a large
number of Ponzi schemes have been devised to steal
money from honest users, and a large volume of
fraudulent accounts are being created routinely to
carry out money laundering. Therefore, an effective
method is to use ML technology to make early
detection of transaction patterns, whether normal or
abnormal. Researchers are concerned about the
combination of BT and ML nowadays. Among them
is research to create new ecosystems that decentralize
infrastructure, data storage management,
administration and ML-based applications.
ML using traditional centralized databases is
changeable and unreliable. The database
administrator has complete control over the database
and has full access to it. Therefore, the combination
of ML technology in the environment can ensure that
the data does not change and remains. The role of
smart contracts also makes the process done without
third parties and automated using ML. There are
many benefits when these two technologies are
combined. This situation has inspired researchers to
propose and build blockchain-ML-based solutions to
enhance the effectiveness of electronic health record
management and SCM and empower security aspects
within the IoT and networks [51, 52]. It is more
interesting when the adaptation of the ML algorithm
in the smart contract is impossible to do [53]. Figure
7 displays a representation of the architecture for ML
adaption in a BT-based application.

Sabri Hisham et al.
1376
Figure 7 Blockchain ML adoption
Blockchain applications in the construction of ML
models provide many advantages compared to
traditional database ecosystems that manage data
centrally. Figure 8 demonstrates the benefit of ML in
the blockchain network. ML algorithms have
incredible possibilities for learning.
Figure 8 The Advantages of Using ML in blockchain
These skills can be applied to the blockchain to make
the chain smarter than previously. Furthermore, we
can create much better ML models leveraging the
decentralized data architecture feature of BT. This
integration can be useful in the improvement of the
privacy, security and transparency of the distributed
ledger network. For example, in improving the
management of sensitive data in the health sector,
Medrec features secure accountability,
confidentiality, and authentication[54].
Computational features found in ML can reduce
processing time and improve aspects of data sharing.
This goal has been achieved through a combination
of cryptographic mechanisms, distributed storage
management, and consensus operations by BT. The
advantages found in blockchain have resulted in
better data value and avoiding data silos [55].
In the analytical aspect, the ML model is developed
using data stored in a blockchain network to make
predictions. This can be seen in past research using a
combination of data sources from blockchain, smart
device tools, sensors, and IoT devices. Integrated
integration between ML and blockchain provides
real-time data analysis. This can be seen in terms of
improving data quality, such as avoiding data
duplication and cleaning up the data. In addition, the
integration of these two technologies has provided
space for researchers to study aspects of data
structures. The study focuses on data security issues
for personal data protection and data analysis issues
to produce predictions for people's habits [56].
Integration of ML models can assist assure the
sustainability of terms and conditions that were
agreed upon before. In addition, the ML model is
updated according to the nodes(chain) environment
of the distributed blockchain network. The
traceability feature in the blockchain allows data to
be tracked in detail, starting from the genesis initiator
block. This has increased the value of transparency
and traceability of the data recorded on the
distributed node network. Pertaining to the
traceability of the BT in IoT, we can also analyze the
hardware of different machines so that ML models
will not stray from the learning path for which they
are allocated in the environment.
5.Anomaly detection method
Blockchains can be exposed to several forms of
harmful assaults and fraudulent activities [57], which
possibly can be discovered by studying the
transaction patterns. Thus, unusual behaviour in data
Security and
privacy
Utilize
Machine
Learning
Model
Data
analytics
Traceability
Data Sharing

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1377
transaction patterns is termed abnormal detection
[58]. Authors [59] explains it as a process which is
used to detect typical patterns in data which are
different from the normal behaviour of the complete
dataset. The ML algorithms would help distinguish
between transactions behavior among user accounts
by learning the associated attributes that pertain to
either anomalous or normal behavior. The method of
identifying normal and abnormal transaction patterns
is also called outlier detection. Therefore, security on
blockchain networks can be protected using an
abnormal detection approach. This is done by making
a prediction about how often abnormal transactions
will happen based on data that has been looked at to
find signs of fraud or attack in blockchain
transactions.
5.1 ML classification model
The ML method is a prominent model used for
anomaly detection in the distributed blockchain
network. In general, the ML model is capable of
generating predictive data (as output) in the future
based on the experience of past data (as input). Thus,
the following section shows the ML experiments with
anomaly detection analysis in the blockchain
network. The workflow process implemented in the
ML model is illustrated in Figure 9.
Figure 9 Typical workflow of ML [56]
This process is initiated in the training process
through pre-processing of the raw(original) data
input. Then, feature extraction is performed, and
models are trained using this data. Feature extraction
should also be performed in the training phase to
produce a final model using test data, and finally, the
analysis process is performed. ML provides several
types of algorithms on the appropriate model based
on the data set provided to produce intelligent results.
Basically, the ML model being developed is based on
training data, where each part of this data has inputs
and outputs. In the training phase, analysis of the
estimated distance between inputs and outputs is used
to help generate the model. Subsequently, further
analysis of normal or abnormal behaviour in
transactions is identified and estimated. Four primary
methods of ML are reviewed such as supervised,
unsupervised, reinforcement learning (RL), neural
network (NN) and deep learning (DL). Nevertheless,
in this review, we focused on supervised and
unsupervised approach. The difference between the
two approaches is shown in Table 4.
Correspondingly, Figure 10 depicts the ML
taxonomy.
5.1.1Supervised learning
In ML and AI, supervised learning is one of the most
widely used methods to make predictions. This
method trains new algorithms to make more accurate
predictions by feeding them sets of data that have
been labelled [60]. In the validation process, the
adjustment of weights based on data input is made to
train the model to reach a good level of suitability.
Next, a learning supervision process is implemented
to produce the desired model as an output using
training data. Clearly, the inputs and outputs in the
training data set will help model learning. The
accuracy of the model was determined using
measurement methods and modified to reach an
error-free level. For example, the author [61] used
supervised ML approaches to detect fraudulent
activity. In general, the classification of models in
supervised learning is based on two problem
scenarios: classification and regression.
Classification refers to how algorithms are used to
classify data based on specific categories. The
classification process involves the identification of
data entities by recognizing data for labelling
purposes. There are several models commonly used
in this classification technique, namely Random
Forest (RF), Ensemble Methods, K-Nearest
Neighbour (KNN), Decision Trees (DT) and Support
Vector Machines (SVM). Besides, there are also
studies to improve the accuracy and performance of
the classification model in making predictions
through the ensemble learning approach.
In general, the ensemble method combines
predictions generated from several individual models,

Sabri Hisham et al.
1378
also called weak models, into new and strong models.
Typically, individual models are biased because they
consist of many variants. Thus, the main intention of
the ensemble learning method is to reduce the value
of variants and bias by mixing them to become a
strong and high-performing new model [62]. In
addition, an idea-based ensemble method to help
make decisions based on several views [63].
Therefore, the final decision is based on a predictive
ensemble that goes through a voting and averaging
procedure. The ensemble technique is made up of
two main steps: building the classification of the
ensemble and combining or integrating the ensemble
[64]. However, there are researchers who employ a
three-phase strategy by incorporating an ensemble
pruning step between ensemble construction and
ensemble combination [64]. In the real environment,
ensemble techniques are also widely used in DL
approaches. Therefore, a study conducted using
medical datasets has proved that the accuracy of
predictions is high using the method of combination
of classifiers compared to individual classifiers in the
DL approach [65].
An important component of the regression technique
is the link between the dependent and independent
variables. This method is often used to produce
forecasts in this scenario, such as sales forecasts,
pricing, markets, stocks, etc. Polynomial regression,
logistical regression, and linear regression are three
popular regression approaches [66].
Table 4 Comparison of supervised and unsupervised
Item/Learning Method
Supervised
Unsupervised
Type of Data
Labeled data is used to train algorithms.
When dealing with unlabeled data,
algorithms are used.
Level of Complexity
Easier method
Difficult to calculate
Level of Accuracy
Method that is extremely exact and reliable
Method that is less accurate and reliable
Figure 10 Illustrates the ML taxonomy
5.1.2 Unsupervised learning
Essentially, unlabeled datasets used for the purpose
of model training are called unsupervised learning. In
unsupervised learning, it is important to look at
hidden data and know which data can be split up into
different sets. In terms of supervision, users do not
need to monitor the model as unsupervised learning
is one of the ML techniques. Thus, the process of
identifying transaction patterns is determined by a
model that functions alone. Among the advantages of
unsupervised learning is that it allows users to
complete complex processing tasks as opposed to
supervised learning. However, it is becoming more

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1379
unpredictable in unsupervised learning compared to
other models. Examples of unsupervised learning
models include NN, anomaly detection, and
clustering. Further, within the Bitcoin network, fraud
is detected using unsupervised learning techniques
[67]. Therefore, among the things that cause
unsupervised learning techniques to be chosen are
knowing unknown patterns in data transactions,
finding features for categorization, analyzing and
labelling in real-time, and the fact that unlabeled data
is more easily obtained than labelled data, i.e.,
requiring manual intervention. Generally, the
classification of unsupervised learning based on
problem challenges is divided into clustering and
association. Among the important ones is clustering
in unsupervised learning. It entails the detection of
patterns as well as data structures with uncategorized
data sets. Among the common models for clustering
are K-Means, K-NN, and Principal Component
Analysis (PCA) [68]. The data elements in the
database will be built into associations according to
the rules of the association. As a whole, correlations
between variables in the database are known through
unsupervised learning techniques.
5.2 Analysis of anomaly detection model
The goal of this review is to collect research articles
about how supervised and unsupervised learning can
be used for anomaly detection. In this review, a total
of 71 previous research published papers related to
the detection of abnormalities in ML are analyzed
from 2017 to 2022. For details, past research related
to supervised learning is shown in Table 5, while
Table 6 lists unsupervised learning. The data from
this obtained study article is arranged on a graph by
classification ML type to aid in the review's analysis
(see Figure 11). The analysis of this figure has shown
that a total of 59 research articles were conducted
using the supervised learning approach, which makes
this approach the most dominant of the entire
research articles. This analysis also shows that
unsupervised learning began to be used by
researchers in 2017 and continued to be used until
2022. Overall, from 2017 to 2022, supervised
learning has been very widely used in abnormal
detection research.
Table 5 The selected research article using supervised learning
Year
Reference
Type
Model
Blockchain
Application
Conclusions/findings
2017
[69]
Conference
RF + XGBoost
Bitcoin
HYIP (High
Yield Investment
Program)
With a true positive count of less than 4.4
percent, a total of 83 HYIP cases were
successfully detected.
2017
[70]
Conference
Ensemble
Blockchain-based
(Ripple)
Anomaly
detection
Concluded that the detection of anomalies
in human behavior is very important in
traditional or modern payment systems
2017
[71]
Journal
RF
Bitcoin
Fraud detection
RF (achieved 99.99 percent), GLM
logistic, and boosted regression models all
outperform each other by more than 90
percent.
2018
[17]
Journal
RF
Cryptocurrency
pump and dump
scams
The average AUC score across all coins is
0.74.
2018
[60]
Journal
RF
Bitcoin
Ponzi Scheme
Overall, the RF classifier is marginally
more effective, according to the studies.
2018
[60]
Journal
RF
Bitcoin
Ponzi Scheme
The results of the study showed a true
positive value of 1 percent, and the top
classifiers managed to detect 31 ponzi
schemes.
2018
[72]
Journal
Gradient
Boosting
algorithm
(ensemble)
Bitcoin
Anomaly
Detection
Obtain a 77 percent accuracy rate, and an
F1-score of 0.75. The best performance
comes via gradient boosting.
2018
[73]
Journal
XGBoost
Ethereum
Ponzi Scheme
The results demonstrate that 45 of the 54
contracts (83 percent) are clever Ponzi
schemes.
2019
[2]
Journal
secureSVM
Blockchain-based
IoT
Proved secure SVM’s efficiency and
security
2019
[19]
Journal
Ensemble DT
Bitcoin
crypto-
ransomware
The LA proposed model produced the
lowest FPR value (1.56 percent) by
analyzing classifying hostage software
compared to RF, Naive Bayes (NB), and
Ensemble (NB with RF).
2019
[20]
Journal
J48
Ethereum
Ponzi Scheme
Among the three classifier models, J48
has the best recall of 0.872.

Sabri Hisham et al.
1380
Year
Reference
Type
Model
Blockchain
Application
Conclusions/findings
2019
[61]
Journal
RF
Ethereum
Fraudulent
Accounts
Three alternative classifiers were
investigated, and RF yielded the best
results in recall and false-positive rate.
2019
[74]
Journal
RF
Bitcoin
High Yield
Investment
Programs
(HYIP)
A study result of 93.75 indicates the
accuracy of fraud detection in Bitcoin
2019
[75]
Journal
Ensemble +
Cascading ML
(CML)
Bitcoin
Ponzi Scheme
The results of the study showed the
precision value was 0.95, the F-score was
0.79, and the recall value was 0.69
2019
[76]
Journal
RF
Bitcoin
Address
Identification
Voting-based methods have shown better
performance than non-voting-based
methods (simulated data of more than
200K Bitcoin addresses) calculated based
on F1 score values, recall, and precision.
2019
[76]
Conference
RF + node2vec
Ethereum
Network Traffic
We measured the nodes' features and their
connections' properties by thoroughly
evaluating the dataset.
2019
[77]
Journal
LightGBM
Bitcoin
Address
Identification
Macro-F1 performance was highest with
87 percent and 86 percent performance
using LightGBM
2019
[78]
Conference
RF
Ethereum
Vulnerability
detection
The model is able to discover
vulnerabilities efficiently and fast. The
results of our model's evaluation of 49502
real-world smart contracts confirm its
usefulness and efficiency.
2019
[79]
Journal
SVN +
Decision Tree +
RF
Ethereum
Security
Analysis
Our model correctly identified a critical
software flaw (accuracy of 95 percent).
2019
[80]
Journal
Naïve Bayes
Cryptocurrency
crypto jacking
detection
Accuracy 0.973
Average F-Score: 0.973
2019
[81]
Conference
XGBoost
Cryptocurrency
pump and dump
scams
After applying the model to the entire
time series of 172 coins, the model
identified 612 pump-like occurrences.
2019
[82]
Journal
RF
Cryptocurrency
anomalous
transactions
Employed a high-precision RF technique
to identify suspicious wallets
2020
[18]
Conference
Ensemble
Bitcoin
Anti-Money
Laundering
(AML)
The result is accuracy (98.13 percent),
Precision (99.11 percent), Recall F1
(71.93 percent) and score false 83.36
percent
2020
[22]
Journal
KNN
Blockchain-based
Electricity
Network
The rates of incidence of anomalies we
used were 1, 2, 3, and 4 percent,
respectively. Over the course of 50 runs,
we evaluated the rate of effective anomaly
identification.
2020
[48]
Journal
LightGBM
Ethereum
Honeypot
The honeypot on a smart contract was
successfully detected, with the study
results being an F1 value (0.93) and AUC
value (0.99).
2020
[82]
Journal
RF
cryptocurrency
Fraudulent
Transactions
The result is Precision (0.96 percent),
Recall (0.96 percent) and F1 Score (0.96
percent)
2020
[83]
Journal
RF
Cryptocurrency
Pump and
Dumps
n/a
2020
[84]
Journal
Ensemble +
XGBoost
Ethereum
Honeypots
Even when all contracts relating to one
honeypot technique were removed from
training, the ML models demonstrated to
generalize effectively.
2020
[85]
Journal
KNN
Bitcoin
Abnormal
transaction
The results of the analysis show that KNN
successfully detects suspicious
transactions at the nodes.
2020
[86]
Journal
Isolation forest
Blockchain-based
Battery Health
In comparison to the well-known anomaly
detection system, experiment results show

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1381
Year
Reference
Type
Model
Blockchain
Application
Conclusions/findings
that the method enhances the F-score
values by up to 25.65 percent.
2020
[87]
Journal
SVM
Ethereum
IoT
The result is Evaluation Metrics (0.99),
Accuracy (0.9998), Recall (1) and F1-
Score (0.9998)
2020
[88]
Journal
Ordered
Boosting
Ethereum
Ponzi Scheme
On a real-world dataset, the new model
obtains a 98 percent F-score, greatly
outperforming existing techniques.
2020
[89]
Journal
RF
Cryptocurrency
Cryptojacking
On our dataset, the BRENNTDROID tool
can detect miners with 95 percent
accuracy.
2020
[90]
Journal
SVN + KNN
Blockchain
Malicious Users
KNN and SVM are better choices because
they require a third of the resources of
CNN algorithm and have accuracy values
higher than 0.9, which is 0.9 percent
lower than CNN.
2020
[91]
Conference
XGBoost
Ethereum
Malicious
Account
That assessment is 96.21 percent accurate,
with only 3 percent false positives.
2020
[92]
Conference
GCN (Graph
Convolutional
Network)
Bitcoin
Money
Laundering
The result is F1-Score (0.773), Recall
(0.678), Accuracy (0.974) and Precision
(0.899).
2020
[93]
Journal
RF
Ethereum
fraudulent
behaviour
The findings of this study showed the
detection of fraudulent behaviour
produces good results using RF
2020
[94]
Journal
RF
Bitcoin
money
laundering
The results of the study showed an
inability to detect illegal (abnormal)
activities using unsupervised learning
techniques.
2020
[95]
Journal
Ensemble
(RF, Stacking
Classifier, and
AdaBoost)
Ethereum
Malicious
Transaction
The ensemble approaches perform well
(F1 score of 0.996).
2020
[96]
Journal
XGBoost
Ethereum
illegal activity
XGBoost classification mode swiftly and
successfully detects illicit behaviour on
the Ethereum network.
2020
[97]
Journal
Ensemble DT
Bitcoin
illicit entities
According to the study's findings, 66
percent of users were correctly classified
using the proposed model
2020
[98]
Journal
Multi-layer
Perceptron
(MLP)
Bitcoin
Scam
We found 6,395 addresses explicitly
offered by scam cases by actively hunting
for them and using ML to identify the
findings.
2021
[9]
Journal
XGBoost
Ethereum
Vulnerability
Detection
The Ethereum Micro-F1 and Macro-F1
give a Turing-complete Ethereum Virtual
ContractWard that is over 96 percent
accurate.
2021
[14]
Journal
KNN algorithm
(Stacking
ensemble)
Blockchain-based
intrusion
detection
Studies show that ensemble stacking
techniques increase the level of
effectiveness by 95 percent in forecasting
analysis.
2021
[99]
Journal
RF + Adaboost
+ SVM
Ethereum
Fraudulent
Transactions
The result of accuracy is 0.97 percent
2021
[100]
Journal
AdaBoost
Ethereum
Phishing
The result analysis for data label
(Phishing) is Precision-0.83, F1 score-
0.74 and Recall-0.66.
).
The result analysis for the data label (No
Phishing) is Precision-0.92,
F1 score-0.94 and Recall-0.97.
2021
[101]
Journal
XGBoost
Bitcoin
Blockchain
simulator
A product called BlockEval formulated
the first simulator to use real Bitcoin data
for simulation purposes.
2021
[102]
Conference
RF
Ethereum
Credit Card
Fraud detection
This study shows that RF produces high
true positive values compared to other
models.
2021
[103]
Journal
Ensemble
Ethereum
Under-priced
The DT is the best strategy of the other

Sabri Hisham et al.
1382
Year
Reference
Type
Model
Blockchain
Application
Conclusions/findings
(Decision Tree,
RF, KNN, SVM
and Naïve
Bayes)
DoS attack
models by giving good results with F and
AUC-ROC score values.
2021
[104]
Journal
RF
Ethereum
Fraud Detection
This study showed the adjustment of the
data set, and the characteristics gave good
results (99%) in terms of accuracy of all
the classifiers tested.
2021
[105]
Journal
Naïve Bayes
Bitcoin
untrusted users
of
cryptocurrency
transaction
services
For both datasets, the accuracy finds that
the Nave Bayes algorithm performs better
than other classification algorithms.
2021
[106]
Journal
Ensemble (DT)
Ethereum
malicious
accounts
Studies using the ensemble technique
(ExtraTreesClassifier) have successfully
detected suspicious accounts with a
balanced accuracy with a range of 87.2
and 88.7.
2021
[107]
Journal
Isolation forest
blockchain-based
IoT
The research results indicate malicious
attempts were successfully detected.
2021
[108]
Journal
RF
Bitcoin
Fraudulent
Transactions
The results of the analysis show that RF
has achieved the highest results compared
to other models, that is, the value of F1
(95.9%).
2021
[109]
Journal
Logistic
Regression
Ethereum
fraud detection
We were able to forecast fraud
transactions with 94 percent confidence
using both approaches, which is
promising.
2021
[110]
Journal
RF
Ethereum
fraud detection
The results demonstrate that by employing
the three algorithms, time measurements
improve significantly, and the RF
approach improves the F measure.
2022
[111]
Journal
Isolation forest
Blockchain
Social Media
This study uses tree algorithms for
abnormal detection and gives good
results.
2022
[112]
Journal
Adaboost
Blockchain-based
IoT
The results showed that Adaboost
produced good results with precision
(97.9%), recall (95%), and F-score
(96.3%).
2022
[113]
Journal
Ensemble
Boosting
Cryptocurrency
anomaly
detection
Studies show the Ensemble Boosting
technique produces good performance
compared to other models.
Table 6 Selected research paper using unsupervised learning
Year
Reference
Type
Model
Blockchain
Application
Conclusions
2018
[15]
Journal
LSTM + RNN
cryptocurrency
crypto jacking
The results of the research yielded the
accuracy of the BMDetector prototype
(93.04%)
2019
[114]
Journal
DBSCAN
Hyperledger fabric
Water Network
The result is accuracy (0.94155),
PR (0.9433), Recall (0.9969) and
F1 score (0.96937)
2019
[115]
Journal
OCVM+ K-
Means
Bitcoin
Anomaly
Detection
The study successfully detected 6
DDOS attacks and 5 multiple spending
attacks.
2019
[116]
Journal
Gaussian
Mixture
Model
Bitcoin
Anomaly
Detection
The results showed that abnormal users
were successfully detected among the
entire user population in this study.
2020
[117]
Journal
LSTM
Blockchain-based
Electricity
Network
The results show the effectiveness of the
proposed framework for identifying
abnormal patterns in transactions. FPR
(0.01 percent), UNSW-NB15: DR
(99.80 percent), FPR (2.93), and Power
System: DR (96.27 percent) were the
readings.
2020
[118]
Journal
LSTM
cryptocurrency
Cryptocurrency
For the analysis of crypto investment

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1383
Year
Reference
Type
Model
Blockchain
Application
Conclusions
Deception
transactions, LSTM achieves 98.99
percent accuracy.
2020
[119]
Journal
K-means
Ethereum
behavioural traits
in transactions
This study looked at behaviours on
transactions using supervised learning
and unsupervised learning.
2021
[120]
Journal
OCSVM
Ethereum
Performance
Testing
This study has successfully detected
Ponzi schemes (1,621 cases) with an F-
score (96 percent).
2021
[121]
Journal
LSTM + RNN
Blockchain-based
Privacy
Protection
This study uses Matthews Correlation
Coefficient (MCC) measurement to
make a prediction, and the result is a
range value between (-1 to 1).
2021
[122]
Journal
LSTM
Ethereum
Anomaly
Detection
This study focuses on the evaluation of
anomaly detection in smart contracts,
covering the type of contract and
identifying malicious contracts.
2021
[122]
Journal
Deep
Autoencoder
NN
Ethereum
Anomaly
Detection
The result is Precision (0.988), AUC
(0.9891), Recall (0.988) and F1-Score
(0.988).
2022
[16]
Journal
OCSVM
Ethereum
Phishing
The result is Precision (0.927), Recall
(0.893) and F-Score (0.908)
Figure 11 Classification of anomaly detection per year
Now we will go over the analysis of the ML model in
additional depth, as given in Table 7 and displayed in
graft as represented in Figure 12. According to the
data in Figure 11, RF is utilized in 22 research
publications and is the most popular model utilized
from 2017 to 2022. In addition, researchers have
applied the ensemble approach in 10 papers, with an
increasing tendency from 2017 to 2022.On the other
hand, 6 research papers employed the XGBoost
model for anomaly detection categorization.
Table 7 ML and its classifiers used in research
ML
Classifier model
Reference
Supervised
Learning
Ensemble Method
[14, 18, 19, 70, 72, 75, 84, 97, 95, 103, 106, 113]
AdaBoost
[99, 100, 112]
GCN (Graph
Convolutional
Network)
[92]
Isolation Forest
[86, 107, 111]
J48
[20]
3
5
14
20
14
3
0 1
3 3 4
1
0
5
10
15
20
25
2017 2018 2019 2020 2021 2022
Supervised learning Unsupervised Learning

Sabri Hisham et al.
1384
KNN
[22, 85, 90]
LightGBM
[48, 77]
Logistic Regression
[109]
MLP (Multi-layer
Perceptron)
[98]
Naïve Bayes
[80,105]
RF
[4, 17, 50, 60, 61, 71, 74, 76, 79,78,82,89,83,93,94,99,102,104,108,110]
secureSVM
[2, 87]
SVM
[79,86,90,99]
XGBoost
[9,17,73,81,96,91,101]
Decision Tree
[79]
Unsupervised
Learning
OCVM (one-class
support vector
machine)
[16,115,114,120,123]
DBSCAN
[21]
Deep Autoencoder
NN
[122]
Gaussian Mixture
Model
[116]
K-means
[114,115,119]
LSTM
[15, 117,118,121,122]
Figure 12 Anomaly detection model per year
5.2.1 Dataset and tools
Referring to the authors [13], anomaly detection is a
change in the operating system that will occur if
suspicious transactions are identified during data
processing. Therefore, the main objective is to adapt
anomaly detection for early detection of traction that
shows suspicious patterns on the entire data set. The
authors previously mentioned that they use datasets
acquired from a single source to uncover
abnormalities in data transactions using normal and
abnormal labelling in a previous research paper.
Consequently, [124] offered a supervised technique
for detecting fake Ethereum accounts. This research
has randomly extracted data taken from Etherscan
(349999 wallets) and identified a total of 2200
wallets used for criminal purposes. Researchers, on
the other hand, utilized more than one data source to
develop this study to classify typical and abnormal
transactions. For a better analysis, there are two sets
of Ethereum transactions on etherscan.io. The first
data set contains about 420 fraudulent wallets found
in the etherscamdb.info database, whereas the second
data set contains non-fraudulent actions found in the
etherscan.io database [93]. Datasets are an important
0 5 10 15 20 25
EM
GCN
J48
LightGBM
MLP
RF
SVM
DT
DBSCAN
GMM
LSTM
2017 2018 2019 2020 2021 2022

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1385
aspect of the development of ML models. Apart from
that, the method or way the data is obtained is
dependent on the tool to access the data. This is
described in Table 8, which lists the datasets and
tools used in previous research to produce the
respective studies. Referring to Table 8, a total of 32
different data sets were used in most experiments.
Besides, from the information listed in Table 8, we
summarize that the datasets may be grouped as live
datasets, media social, open datasets, and tools
datasets, as illustrated in Figure 13. From the data in
Figure 13, it is evident that the most commonly used
group of datasets in this selected research paper was
open datasets.
The frequency analysis of the datasets used in the
previous study is shown in Figure 14. Datasets
accessed live from Etherscan sources are the most
popular approach used by researchers. In addition, 5
research publications employed open datasets from
Elliptic and 4 research studies adopted open datasets
from the Computational Biology Lab (University of
Illinois) and Bitcoin.
Tools are ways or means for obtaining data for
analysis, integration, development, and study that
come in the form of application software, web
platforms (API, Web Services), libraries, and smart
devices. These technologies are used to access data as
datasets during the data collecting phase of the
creation of ML models. Following the data
processing and ML model construction phase, these
tools are categorized into APIs, Development Tools,
Software, and smart devices, as shown in Figure
15.The utilization of development tools like Python,
Weka, and others is prevalent in research articles
about anomaly detection applications.
Analysis in Figure 16 displays the frequency with
which different tools are employed in research
articles. The most often used method in research for
anomaly detection is access to Etherscan via APIs,
which is described in 10 research articles. In addition,
7 research articles leveraged the Python library as
tools for ML model construction. Previous studies
also employed a range of methodologies or
instruments to generate more accurate models. For
example, the authors [103] utilizes Ganache as a
personal Ethereum blockchain, a database to keep
live Ethereum transactions collected via service APIs
(Etherscan API), and web3.py to request and receive
transaction execution results from the Ethereum in
the face of an under-priced denial of service (DoS)
attack.
Table 8 Previous research articles with the use of different datasets and tools
Tools
Dataset
 Binance API[81]
 Chrome Web Driver[118]
 Ethereum Client[16][122]
 Etherscan API[9][16][20][50][61][82][96][93][100][122]
 Geth Client[96]
 Github[60][104]
 Google Big Query[88][104][119]
 Honeybadger[48]
 Parity Client[122]
 Pycharm[82]
 Python[18](89)
 R[69][85][102]
 RS-232 connection[122]
 Scikit-learn [70][82][108][102]
 Selenium[118]
 Telegram API[17][81][83]
 Text-blob[118]
 Twitter API[17][83]
 anonymity-in-bitcoin.blogspot.com[108]
 Binance[82][81]
 Telegram Data[17][81][83]
 Twitter data[17][83]
 Bitcointalk[60][88][83][108]
 Etherscan[9][16][20][50][61][73][82][96][93][100][119][122]
 https://goo.gl/k5PCOZ[69]
 Computational Biology Lab (University of Illinois)[71]
 Google BigQuery[88][104]
 PonziTect[88]
 Elliptic[18][50][94](89)[118]
 EtherScamDB[16][96][93]
 https://goo.gl/CvdxBp[20]
 gz.blockchair.com/bitcoin/[116]
 Honeybadger[48]
 goo.gl/ToCho7[60]
 github.com/bitcoinponzi[60]
 blockchain.info/tags[60][74]
 Reddit[60][83]
 rissgroup.org/[19]
 blockchain.com[85][115]
 blockchair.com/dumps[85]
 chainalysis.com[72]
 Kaggle[110][109][113]
 Ripple Bank[70]
 WalletExplorer.com[74][98]
 srg.site.uottawa.ca/bgsieeesb2020/[98]
 bitcoincharts.com[98]

Sabri Hisham et al.
1386
Tools
Dataset
 VJTI blockchain Lab[97]
 Xapo.com[74]
 ULB[102]
 Cryptocurrency Market Data[17]
 Water dataset[122]
 Ethereum Client[95]
 Etherscan API[79][78][84][106][103]
 Honeybadger[84]
 Scikit-learn[103]
 BMDetector[15]
 Solidity[79]
 Ganache[103]
 Web3[103]
 Stratum[80]
 WEKA[80]
 VirusTotal[89]
 BRENNTDROID[89]
 ZombieCoin[123]
 Smart Meter[121]
 Etherscan[79][78][95][84][106][103]
 BMDetector[15]
 Block Explorer[79]
 KDD99[14]
 Stratum[80]
 Smart Meter[121]
 VirusTotal[89]
 BRENNTDROID[89]
 ZombieCoin[123]
 Etherscan API[114]
 Google Big Query[114][120]
 Scikit-learn[77][76]
 Management System[114]
 NetFlow[76]
 Etherscan[114]
 bitcointalk.org[77]
 Google BigQuery[114][120]
 blockchain.info/tags[77]
 Media social login data[111]
 NetFlow[76]
 WalletExplorer.com[77][76][75]
 IOT Meterr[21][22]
 Python[87]
 Scikit-learn[107]
 Web3[87]
 MATLAB[87]
 VirusShare[86]
 IOTPOT[86]
 BlockEval[101]
 SimPy[101]
 Smart Power Network[117]
 bitcoind[101]
 blockchain.info/tags[101]
 IoTID20[112]
 KDD99[107]
 IOT Meter[21][22]
 NSL-KDD[87]
 UNSW-NB15[87]
 UCI ML [2]
 Smart Power Network[117]
 VirusShare[86]
 IOTPOT[86]
 NASA Battery[86]

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1387
Figure 13 Illustrates the blockchain dataset from a different source

Sabri Hisham et al.
1388
Figure 14 Utilized datasets in collected research articles
Figure 15 The tool classification for dataset readiness
1 2 3
2 4 12
4
2
1 5
3
2
2
1 2
2
1 2
1 3
1 2
1
1
1
1
1
1
1
1 3
0 2 4 6 8 10 12 14
anonymity-in-bitcoin.blogspot.com
Telegram Data
Bitcointalk
Laboratory for Computational Biology at the University…
PonziTect
EtherScamDB
blockchair.com
blockchain.info
rissgroup
chainalysis.com
Ripple Bank
srg.site.uottawa.ca/bgsieeesb2020
VJTI Blockchain Lab
ULB
Cryptocurrency Market Data
shorten link website
Tools
API
Etherscan
Google Big
Query
Telegram
Twitter
Binance
Development Tools
Python
Pycharm
Text-blob
Scikit-learn
Weka R Ganache Parity
Client Geth Solidity
Web3
Software
Selenium
Stratum
Chrome Web Driver
Honeybadger
BMDetector
VirusTotal
BRENNTDROID
Smart
Devices
Smart
Meter

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1389
Figure 16 Tools utilized in collected research articles
5.2.2 Performance metrics
The evaluation metrics are used in ML to calculate
the accuracy and performance of your trained ML
models. This can assist you in figuring out how much
better your ML model will perform on a dataset it has
never seen before. A performance evaluation metric
in ML is crucial for determining how well the
proposed ML model performs on a dataset it has
never seen before. In real analysis, the ML model's
performance was measured in more than one way to
make sure that the analysis was as accurate as
possible. In total, we found 41 publications in the
selected research articles that explicitly reported the
performance measures of their suggested models.
Figure 17 illustrates that accuracy and F-score were
the most regularly utilized performance metrics (23
papers). Furthermore, the researcher's favoured
approach to determining the performance parameter
is recalled, with 22 studies using recall and 14
publications utilizing accuracy. It calculates the
number of anomalies that have been accurately
classified. Furthermore, we noticed that 3 of the 41
articles utilized only one performance metric, and the
bulk of those publications only employed accuracy,F-
score and Area under the ROC Curve (AUC), which
is insufficient to verify the ML model's performance.
Figure 17 Frequency of performance metrics among the research papers
1 3
2 10
1
1 2
1
1 2 4
3
1
1
1 2
1
0 2 4 6 8 10 12
Binance API
Ethereum Client
Parity Client
Github
Pycharm
Scikit-learn
Selenium
Text-blob
RS-232 connection
4
1
1 3
2 3 23
22 23
1 4 14
2
2 5
1 2 9
7
1
1 4
4
0 5 10 15 20 25
Roc
Macro-F1
FN
Precision
F-score
Specificity
G-Mean
TPR
TNR
AUC
Mean
Others

Sabri Hisham et al.
1390
6.Discussion
In this study, research information was collected
from 2017 to 2022. After the analysis was done, it
was found that 21 ML models were used in previous
research by researchers. Among the very widely used
models is RF. Nevertheless, we can witness a new
developing tendency, and most researchers have
sought to apply the ensemble method. In terms of
datasets, it was found that the entire experiment
conducted in the previous research used 32 different
datasets. From these observations, the live dataset
category is widely used as training data and testing
data in research. The most widely used approach or
instrument in research for anomaly discovery is
access to Etherscan via API, which is documented in
10 research articles. In addition, 7 research articles
leveraged the Python library as tools for ML model
construction
In general, performance measurement is an important
element of looking at the performance of a final
model that has been developed. The previous analysis
found that 3 out of 41 studies used only one
measurement method. As a result, the use of fewer
measurement methods will produce less accurate
experimental results. An analysis of the frequency of
use of anomaly detection types was also performed.
Referring to Figure 10, a total of 59 research studies
adapted the supervised learning approach and made it
the most dominant use in research. Meanwhile,
unsupervised learning was used in 22 research
studies. From the perspective of the model analysis, a
total of 22 research publications uses RF, making it
the most popular used in research from 2017 to
2022.Although the XGBoost, RF, and AdaBoost
models show high usage trends in research, studies to
improve performance need to be done from time to
time. There has been researched on the use of
ensemble strategies to improve learning performance
on algorithms.
A complete list of abbreviations is shown in
Appendix I.
7.Conclusion and future work
Blockchain is a P2P technology that uses a
distributed ledger to store data in blocks. Thus, the
feature makes the ledger on the Blockchain
unchangeable. Although blockchain has many
benefits in terms of technology, it is still prone to
various security issues, operations, data management,
fraud and so on. Therefore, the adaptation of ML
technology to the blockchain has a positive impact
through its ability to learn from past data.
Furthermore, anomaly detection methods are very
popular in ML approaches, which are responsible for
detecting, from the outset, suspicious transactions.
This indirectly helps predict strange and unusual
transactions within the blockchain network. In this
review, some subjects are explained, particularly the
background of BT, an overview of ML technology,
integration of ML with blockchain and
implementation of anomaly detection methods in the
distributed blockchain network. The analysis of the
ML model is carried out through four main
perspectives, namely the type of anomaly detection,
dataset, tools, and measurement methods.Thus, the
study concludes that supervised machine learning is
the most commonly employed technique in previous
research. Over the years, however, the usage of
ensemble approaches has increased. In terms of
models, XGBoost, RF, and AdaBoost are frequently
employed in research. While the data source (live
dataset), the Etherscan blockchain explorer platform
(API), and the Python library are also the most
adapted from prior studies.The selection of feature
adjustments as well as extraction impacts
performance improvements. In addition, research
needs to use more than one measurement metric to
produce more accurate results.
Based on this study's analysis, there are a number of
challenges that researchers must overcome. Among
them is the difficulty of immediately acquiring the
most recent raw data pulled from the blockchain
network, as it needs an excessive amount of storage
space and a high level of technical expertise.
Consequently, the majority of researchers rely on
outdated datasets posted on community websites or in
public repositories. The only way to create models
that perform better and are more accurate is by
utilising current data. This is supported by the authors
[58], who advised academics and scholars to use the
latest databases in their research.
In line with the development of the ML approach in
the world of data science, among the potential future
research by researchers is the use of auto-ML, ML
Operations (MLOps) and ML designers. These are
suitable for data preprocessing techniques, training
data for models, evaluating model performance,
experiments, and model monitoring in cloud-based.
Besides, performance optimization produced by the
ML approach using optimization methods such as ant
colony, metaheuristic algorithms, etc., is a potential
topic of study in the future.

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1391
Acknowledgment
The Universiti Sultan Zainal Abidin's Research
Management and Innovation Centre funded this study.
Conflicts of interest
The authors have no conflicts of interest to declare.
Author’s contribution statement
Sabri Hisham: Choosing an analysis method, searching a
journal database, reading, writing, making changes, writing
the final draft, checking for plagiarism, and proofreading.
Mokhairi Makhtar: Supervision, input on draught
revision, and final revision.Azwa Abdul Aziz:
Supervision, exchange of sample manuscripts, and draught
evaluation comments.
References
[1] Abd RNH. Blockchain technology in e-voting:
comparative study. 2020.
[2] Shen M, Tang X, Zhu L, Du X, Guizani M. Privacy-
preserving support vector machine training over
blockchain-based encrypted IoT data in smart cities.
IEEE Internet of Things Journal. 2019; 6(5):7702-12.
[3] Warraich J, Singh C, Thapa P. Blockchain-based
intelligent monitored security system for detection of
replication attack in the wireless healthcare network.
European Journal of Engineering and Technology
Research. 2021; 6(6):160-70.
[4] Boughaci D, Alkhawaldeh AA. Enhancing the security
of financial transactions in Blockchain by using
machine learning techniques: towards a sophisticated
security tool for banking and finance. In 2020 first
international conference of smart systems and
emerging technologies (SMARTTECH) 2020 (pp.
110-5). IEEE.
[5] Wang Z, Luo N, Zhou P. GuardHealth: Blockchain
empowered secure data management and graph
convolutional network enabled anomaly detection in
smart healthcare. Journal of Parallel and Distributed
Computing. 2020; 142:1-12.
[6] Dhieb N, Ghazzai H, Besbes H, Massoud Y. A secure
ai-driven architecture for automated insurance
systems: fraud detection and risk measurement. IEEE
Access. 2020; 8:58546-58.
[7] Nakamoto S. Bitcoin: a peer-to-peer electronic cash
system. Decentralized Business Review. 2008.
[8] Moubarak J, Filiol E, Chamoun M. On Blockchain
security and relevant attacks. In IEEE middle east and
north Africa communications conference
(MENACOMM) 2018 (pp. 1-6). IEEE.
[9] Wang W, Song J, Xu G, Li Y, Wang H, Su C.
Contractward: automated vulnerability detection
models for smart contracts. IEEE Transactions on
Network Science and Engineering. 2020; 8(2):1133-
44.
[10] Irwin AS, Turner AB. Illicit bitcoin transactions:
challenges in getting to the who, what, when and
where. Journal of Money Laundering Control. 2018.
[11] Chen L, Peng J, Liu Y, Li J, Xie F, Zheng Z. Phishing
scams detection in ethereum transaction network.
ACM Transactions on Internet Technology (TOIT).
2020; 21(1):1-16.
[12] Conti M, Kumar ES, Lal C, Ruj S. A survey on
security and privacy issues of bitcoin. IEEE
Communications Surveys & Tutorials. 2018;
20(4):3416-52.
[13] Chalapathy R, Chawla S. Deep learning for anomaly
detection: a survey. arXiv preprint arXiv:1901.03407.
2019.
[14] Khan AA, Khan MM, Khan KM, Arshad J, Ahmad F.
A blockchain-based decentralized machine learning
framework for collaborative intrusion detection within
UAVs. Computer Networks. 2021.
[15] Liu J, Zhao Z, Cui X, Wang Z, Liu Q. A novel
approach for detecting browser-based silent miner. In
IEEE third international conference on data science in
cyberspace (DSC) 2018 (pp. 490-7). IEEE.
[16] Wu J, Yuan Q, Lin D, You W, Chen W, Chen C, et al.
Who are the phishers? phishing scam detection on
ethereum via network embedding. IEEE Transactions
on Systems, Man, and Cybernetics: Systems. 2020;
52(2):1156-66.
[17] Mirtaheri M, Abu-el-haija S, Morstatter F, Ver SG,
Galstyan A. Identifying and analyzing cryptocurrency
manipulations in social media. IEEE Transactions on
Computational Social Systems. 2021; 8(3):607-17.
[18] Alarab I, Prakoonwit S, Nacer MI. Comparative
analysis using supervised learning methods for anti-
money laundering in bitcoin. In proceedings of the
2020 5th international conference on machine learning
technologies 2020 (pp. 11-7).
[19] Kok SH, Abdullah A, Jhanjhi NZ, Supramaniam M.
Prevention of crypto-ransomware using a pre-
encryption detection algorithm. Computers. 2019;
8(4):1-15.
[20] Jung E, Le TM, Gehani A, Ge Y. Data mining-based
ethereum fraud detection. In 2019 IEEE international
conference on blockchain (Blockchain) 2019 (pp. 266-
73). IEEE.
[21] Iyer S, Thakur S, Dixit M, Katkam R, Agrawal A,
Kazi F. Blockchain and anomaly detection based
monitoring system for enforcing wastewater reuse. In
2019 10th international conference on computing,
communication and networking technologies
(ICCCNT) 2019 (pp. 1-7). IEEE.
[22] Li M, Zhang K, Liu J, Gong H, Zhang Z. Blockchain-
based anomaly detection of electricity consumption in
smart grids. Pattern Recognition Letters. 2020;
138:476-82.
[23] Bhutta MN, Khwaja AA, Nadeem A, Ahmad HF,
Khan MK, Hanif MA, et al. A survey on blockchain
technology: evolution, architecture and security. IEEE
Access. 2021; 9:61048-73.
[24] Merkle RC. Protocols for public key cryptosystems. In
secure communications and asymmetric cryptosystems
2019 (pp. 73-104). Routledge.
[25] Tschorsch F, Scheuermann B. Bitcoin and beyond: a
technical survey on decentralized digital currencies.
IEEE Communications Surveys & Tutorials. 2016;
18(3):2084-123.

Sabri Hisham et al.
1392
[26] Casino F, Dasaklis TK, Patsakis C. A systematic
literature review of blockchain-based applications:
current status, classification and open issues.
Telematics and Informatics. 2019; 36:55-81.
[27] Arya J, Kumar A, Singh AP, Mishra TK, Chong PH.
Blockchain: basics, applications, challenges and
opportunities.
[28] Schollmeier R. A definition of peer-to-peer
networking for the classification of peer-to-peer
architectures and applications. In proceedings first
international conference on peer-to-peer computing
2001 (pp. 101-2). IEEE.
[29] Ozisik AP, Levine BN. An explanation of Nakamoto's
analysis of double-spend attacks. arXiv preprint
arXiv:1701.03977. 2017.
[30] Baliga A. Understanding blockchain consensus
models. Persistent. 2017; 4(1).
[31] Zheng Z, Xie S, Dai H, Chen X, Wang H. An
overview of blockchain technology: architecture,
consensus, and future trends. In international congress
on big data (BigData congress) 2017 (pp. 557-64).
IEEE.
[32] Back A. Hashcash-a denial of service counter-
measure. 2002.
[33] Nguyen CT, Hoang DT, Nguyen DN, Niyato D,
Nguyen HT, Dutkiewicz E. Proof-of-stake consensus
mechanisms for future blockchain networks:
fundamentals, applications and opportunities. IEEE
Access. 2019; 7:85727-45.
[34] Yao W, Ye J, Murimi R, Wang G. A survey on
consortium blockchain consensus mechanisms. arXiv
preprint arXiv:2102.12058. 2021.
[35] Cai W, Wang Z, Ernst JB, Hong Z, Feng C, Leung
VC. Decentralized applications: the blockchain-
empowered software system. IEEE Access. 2018;
6:53019-33.
[36] Ali MS, Vecchio M, Pincheira M, Dolui K, Antonelli
F, Rehmani MH. Applications of blockchains in the
internet of things: a comprehensive survey. IEEE
Communications Surveys & Tutorials. 2018;
21(2):1676-717.
[37] Xie J, Tang H, Huang T, Yu FR, Xie R, Liu J, et al. A
survey of blockchain technology applied to smart
cities: research issues and challenges. IEEE
Communications Surveys & Tutorials. 2019;
21(3):2794-830.
[38] Lu Y. Blockchain: a survey on functions, applications
and open issues. Journal of Industrial Integration and
Management. 2018; 3(4).
[39] Xinyi Y, Yi Z, He Y. Technical characteristics and
model of blockchain. In international conference on
communication software and networks (ICCSN) 2018
(pp. 562-6). IEEE.
[40] Lin IC, Liao TC. A survey of blockchain security
issues and challenges. International Journal of
Network Security. 2017; 19(5):653-9.
[41] Puthal D, Malik N, Mohanty SP, Kougianos E, Yang
C. The blockchain as a decentralized security
framework [future directions]. IEEE Consumer
Electronics Magazine. 2018; 7(2):18-21.
[42] Yang R, Yu FR, Si P, Yang Z, Zhang Y. Integrated
blockchain and edge computing systems: a survey,
some research issues and challenges. IEEE
Communications Surveys & Tutorials. 2019;
21(2):1508-32.
[43] Nakamoto WS. A next generation smart contract &
decentralized application platform. Etherum. 2014.
[44] Androulaki E, Barger A, Bortnikov V, Cachin C,
Christidis K, De CA, et al. Hyperledger fabric: a
distributed operating system for permissioned
blockchains. In proceedings of the thirteenth Eurosys
conference 2018 (pp. 1-15).
[45] https://en.bitcoin.it/wiki/Genesis_block. Accessed 15
May 2022.
[46] Szabo N. Smart contracts: building blocks for digital
markets. EXTROPY: The Journal of Transhumanist
Thought,(16). 1996; 18(2).
[47] Zheng Z, Xie S, Dai HN, Chen W, Chen X, Weng J, et
al. An overview on smart contracts: challenges,
advances and platforms. Future Generation Computer
Systems. 2020; 105:475-91.
[48] Chen W, Guo X, Chen Z, Zheng Z, Lu Y, Li Y.
Honeypot contract risk warning on ethereum smart
contracts. In international conference on joint cloud
computing 2020 (pp. 1-8). IEEE.
[49] Tikhomirov S, Voskresenskaya E, Ivanitskiy I,
Takhaviev R, Marchenko E, Alexandrov Y.
Smartcheck: static analysis of ethereum smart
contracts. In proceedings of the 1st international
workshop on emerging trends in software engineering
for blockchain 2018 (pp. 9-16).
[50] Chen W, Zheng Z, Ngai EC, Zheng P, Zhou Y.
Exploiting blockchain data to detect smart ponzi
schemes on ethereum. IEEE Access. 2019; 7:37575-
86.
[51] Salah K, Rehman MH, Nizamuddin N, Al-fuqaha A.
Blockchain for AI: review and open research
challenges. IEEE Access. 2019; 7:10127-49.
[52] Chen F, Wan H, Cai H, Cheng G. Machine learning
in/for blockchain: future and challenges. Canadian
Journal of Statistics. 2021; 49(4):1364-82.
[53] Wang T. A unified analytical framework for trustable
machine learning and automation running with
blockchain. In 2018 IEEE international conference on
big data (Big Data) 2018 (pp. 4974-83). IEEE.
[54] Azaria A, Ekblaw A, Vieira T, Lippman A. Medrec:
using blockchain for medical data access and
permission management. In 2016 2nd international
conference on open and big data (OBD) 2016 (pp. 25-
30). IEEE.
[55] Liu J, Peng S, Long C, Wei L, Liu Y, Tian Z.
Blockchain for data science. In proceedings of the
international conference on blockchain technology
2020 (pp. 24-8).
[56] Zhang D. Big data security and privacy protection. In
8th international conference on management and
computer science (ICMCS 2018) 2018 (pp. 275-8).
Atlantis Press.

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1393
[57] Rahouti M, Xiong K, Ghani N. Bitcoin concepts,
threats, and machine-learning security solutions. IEEE
Access. 2018; 6:67189-205.
[58] Nassif AB, Talib MA, Nasir Q, Dakalbab FM.
Machine learning for anomaly detection: a systematic
review. IEEE Access. 2021; 9:78658-700.
[59] Bulusu S, Kailkhura B, Li B, Varshney PK, Song D.
Anomalous example detection in deep learning: a
survey. IEEE Access. 2020; 8:132330-47.
[60] Bartoletti M, Pes B, Serusi S. Data mining for
detecting bitcoin ponzi schemes. In 2018 crypto valley
conference on blockchain technology (CVCBT) 2018
(pp. 75-84). IEEE.
[61] Ostapowicz M, Żbikowski K. Detecting fraudulent
accounts on blockchain: a supervised approach. In
international conference on web information systems
engineering 2020 (pp. 18-31). Springer, Cham.
[62] Bamhdi AM, Abrar I, Masoodi F. An ensemble based
approach for effective intrusion detection using
majority voting. Telkomnika (Telecommunication
Computing Electronics and Control). 2021; 19(2):664-
71.
[63] Awang MK, Makhtar M, Udin N, Mansor NF.
Improving customer churn classification with
ensemble stacking method. International Journal of
Advanced Computer Science and Applications. 2021;
12(11).
[64] Awang MK, Makhtar M, Mamat AR. Ensemble
selection and combination based on cost function for
UCI datasets. Journal of Theoretical and Applied
Information Technology. 2021; 99(16): 4015-25.
[65] Rosly R, Makhtar M, Awang MK, Hassan H, Rose
AN. Deep multi-classifier learning for medical data
sets. International Journal of Engineering Trends and
Technology (IJETT). 2020.
[66] Mendes-moreira J, Soares C, Jorge AM, Sousa JF.
Ensemble approaches for regression: a survey. ACM
Computing Surveys. 2012; 45(1):1-40.
[67] Monamo P, Marivate V, Twala B. Unsupervised
learning for robust Bitcoin fraud detection. In 2016
information security for South Africa (ISSA) 2016
(pp. 129-34). IEEE.
[68] Kumar P, Gupta GP, Tripathi R. TP2SF: a trustworthy
privacy-preserving secured framework for sustainable
smart cities by leveraging blockchain and machine
learning. Journal of Systems Architecture. 2021.
[69] Toyoda K, Ohtsuki T, Mathiopoulos PT. Identification
of high yielding investment programs in bitcoin via
transactions pattern analysis. In GLOBECOM 2017
(pp. 1-6). IEEE.
[70] Camino RD, State R, Montero L, Valtchev P. Finding
suspicious activities in financial transactions and
distributed ledgers. In international conference on data
mining workshops (ICDMW) 2017 (pp. 787-96).
IEEE.
[71] Monamo PM, Marivate V, Twala B. A multifaceted
approach to bitcoin fraud detection: global and local
outliers. In 2016 15th IEEE international conference
on machine learning and applications (ICMLA) 2016
(pp. 188-94). IEEE.
[72] Harlev MA, Sun YH, Langenheldt KC, Mukkamala R,
Vatrapu R. Breaking bad: de-anonymising entity types
on the bitcoin blockchain using supervised machine
learning. In proceedings of the 51st hawaii
international conference on system sciences 2018.
[73] Chen W, Zheng Z, Cui J, Ngai E, Zheng P, Zhou Y.
Detecting ponzi schemes on ethereum: towards
healthier blockchain technology. In proceedings of the
2018 world wide web conference 2018 (pp. 1409-18).
[74] Toyoda K, Mathiopoulos PT, Ohtsuki T. A novel
methodology for hyip operators’ bitcoin addresses
identification. IEEE Access. 2019; 7:74835-48.
[75] Zola F, Bruse JL, Eguimendia M, Galar M, Orduna
UR. Bitcoin and cybersecurity: temporal dissection of
blockchain data to unveil changes in entity behavioral
patterns. Applied Sciences. 2019; 9(23):1-20.
[76] Kanemura K, Toyoda K, Ohtsuki T. Identification of
darknet markets’ bitcoin addresses by voting per-
address classification results. In 2019 IEEE
international conference on blockchain and
cryptocurrency (ICBC) 2019 (pp. 154-8). IEEE.
[77] Lin YJ, Wu PW, Hsu CH, Tu IP, Liao SW. An
evaluation of bitcoin address classification based on
transaction history summarization. In international
conference on blockchain and cryptocurrency (ICBC)
2019 (pp. 302-10). IEEE.
[78] Song J, He H, Lv Z, Su C, Xu G, Wang W. An
efficient vulnerability detection model for ethereum
smart contracts. In international conference on
network and system security 2019 (pp. 433-42).
Springer, Cham.
[79] Momeni P, Wang Y, Samavi R. Machine learning
model for smart contracts security analysis. In 2019
17th international conference on privacy, security and
trust (PST) 2019 (pp. 1-6). IEEE.
[80] I MJZ, Suárez-varela J, Barlet-ros P. Detecting
cryptocurrency miners with NetFlow/IPFIX network
measurements. In international symposium on
measurements & networking (M&N) 2019 (pp. 1-6).
IEEE.
[81] Victor F, Hagemann T. Cryptocurrency pump and
dump schemes: quantification and detection. In
international conference on data mining workshops
(ICDMW) 2019 (pp. 244-51). IEEE.
[82] Baek H, Oh J, Kim CY, Lee K. A model for detecting
cryptocurrency transactions with discernible purpose.
In eleventh international conference on ubiquitous
and future networks (ICUFN) 2019 (pp. 713-7). IEEE.
[83] La MM, Mei A, Sassi F, Stefa J. Pump and dumps in
the bitcoin era: real time detection of cryptocurrency
market manipulations. In 2020 29th international
conference on computer communications and
networks (ICCCN) 2020 (pp. 1-9). IEEE.
[84] Camino R, Torres CF, Baden M, State R. A data
science approach for detecting honeypots in ethereum.
In 2020 IEEE international conference on blockchain
and cryptocurrency (ICBC) 2020 (pp. 1-9). IEEE.
[85] Liao Q, Gu Y, Liao J, Li W. Abnormal transaction
detection of bitcoin network based on feature fusion.
In 2020 IEEE 9th joint international information

Sabri Hisham et al.
1394
technology and artificial intelligence conference
(ITAIC) 2020 (pp. 542-9). IEEE.
[86] Ngo QD, Nguyen HT, Tran HA, Nguyen DH. IoT
botnet detection based on the integration of static and
dynamic vector features. In 2020 IEEE eighth
international conference on communications and
electronics (ICCE) 2021(pp. 540-5). IEEE.
[87] Cheema MA, Qureshi HK, Chrysostomou C, Lestas
M. Utilizing blockchain for distributed machine
learning based intrusion detection in internet of things.
In 16th international conference on distributed
computing in sensor systems (DCOSS) 2020 (pp. 429-
35). IEEE.
[88] Fan S, Fu S, Xu H, Zhu C. Expose your mask: smart
ponzi schemes detection on blockchain. In 2020
international joint conference on neural networks
(IJCNN) 2020 (pp. 1-7). IEEE.
[89] Dashevskyi S, Zhauniarovich Y, Gadyatskaya O,
Pilgun A, Ouhssain H. Dissecting android
cryptocurrency miners. In proceedings of the tenth
ACM conference on data and application security and
Privacy 2020 (pp. 191-202).
[90] Huang D, Chen B, Li L, Ding Y. Anomaly detection
for consortium blockchains based on machine learning
classification algorithm. In international conference on
computational data and social networks 2020 (pp. 307-
18). Springer, Cham.
[91] Kumar N, Singh A, Handa A, Shukla SK. Detecting
malicious accounts on the Ethereum blockchain with
supervised learning. In international symposium on
cyber security cryptography and machine learning
2020 (pp. 94-109). Springer, Cham.
[92] Alarab I, Prakoonwit S, Nacer MI. Competence of
graph convolutional networks for anti-money
laundering in bitcoin blockchain. In proceedings of the
2020 5th international conference on machine learning
technologies 2020 (pp. 23-7).
[93] Lašas K, Kasputytė G, Užupytė R, Krilavičius T.
Fraudulent behaviour identification in ethereum
blockchain. In CEUR workshop proceedings
[Electronic Resource]: IVUS 2020, information
society and university studies, kaunas, lithuania:
proceedings. Aachen: CEUR-WS 2020.
[94] Lorenz J, Silva MI, Aparício D, Ascensão JT, Bizarro
P. Machine learning methods to detect money
laundering in the bitcoin blockchain in the presence of
label scarcity. In proceedings of the first ACM
international conference on AI in Finance 2020 (pp. 1-
8).
[95] Poursafaei F, Hamad GB, Zilic Z. Detecting malicious
Ethereum entities via application of machine learning
classification. In 2020 2nd conference on blockchain
research & applications for innovative networks and
services (BRAINS) 2020 (pp. 120-7). IEEE.
[96] Farrugia S, Ellul J, Azzopardi G. Detection of illicit
accounts over the Ethereum blockchain. Expert
Systems with Applications. 2020.
[97] Nerurkar P, Busnel Y, Ludinard R, Shah K, Bhirud S,
Patel D. Detecting illicit entities in bitcoin using
supervised learning of ensemble decision trees. In
proceedings of the 2020 10th international conference
on information communication and management 2020
(pp. 25-30).
[98] Badawi E, Jourdan GV, Bochmann G, Onut IV. An
automatic detection and analysis of the bitcoin
generator scam. In 2020 IEEE european symposium
on security and privacy workshops (EuroS&PW) 2020
(pp. 407-16). IEEE.
[99] Bhowmik M, Chandana TS, Rudra B. Comparative
study of machine learning algorithms for fraud
detection in blockchain. In 2021 5th international
conference on computing methodologies and
communication (ICCMC) 2021 (pp. 539-41). IEEE.
[100]Wen H, Fang J, Wu J, Zheng Z. Transaction-based
hidden strategies against general phishing detection
framework on ethereum. In 2021 IEEE international
symposium on circuits and systems (ISCAS) 2021 (pp.
1-5). IEEE.
[101]Gouda DK, Jolly S, Kapoor K. Design and validation
of blockeval, a blockchain simulator. In 2021
international conference on communication systems &
networks (COMSNETS) 2021 (pp. 281-9). IEEE.
[102]Balagolla EM, Fernando WP, Rathnayake RM,
Wijesekera MJ, Senarathne AN, Abeywardhana KY.
Credit card fraud prevention using blockchain. In 2021
6th international conference for convergence in
technology (I2CT) 2021 (pp. 1-8). IEEE.
[103]Sousa JE, Oliveira VC, Valadares JA, Vieira AB,
Bernardino HS, Villela SM, et al. Fighting under-price
DoS attack in ethereum with machine learning
techniques. SIGMETRICS Perform. Eval. Rev. 2021;
48(4):24-7.
[104]Al-e’mari S, Anbar M, Sanjalawe Y, Manickam S. A
labeled transactions-based dataset on the ethereum
network. In international conference on advances in
cyber security 2020 (pp. 61-79). Springer, Singapore.
[105]Mittal R, Bhatia MP. Detection of suspicious or un-
trusted users in crypto-currency financial trading
applications. International Journal of Digital Crime
and Forensics (IJDCF). 2021; 13(1):79-93.
[106]Agarwal R, Barve S, Shukla SK. Detecting malicious
accounts in permissionless blockchains using temporal
graph properties. Applied Network Science. 2021;
6(1):1-30.
[107]Yang X, Chen Y, Qian X, Li T, Lv X. BCEAD: a
blockchain-empowered ensemble anomaly detection
for wireless sensor network via isolation forest.
Security and Communication Networks. 2021.
[108]Chen B, Wei F, Gu C. Bitcoin theft detection based
on supervised machine learning algorithms. Security
and Communication Networks. 2021.
[109]Lacruz F, Saniie J. Applications of machine learning
in fintech credit card fraud detection. In international
conference on electro information technology (EIT)
2021 (pp. 1-6). IEEE.
[110]Ibrahim RF, Elian AM, Ababneh M. Illicit account
detection in the ethereum blockchain using machine
learning. In 2021 international conference on
information technology (ICIT) 2021 (pp. 488-93).
IEEE.

International Journal of Advanced Technology and Engineering Exploration, Vol 9(95)
1395
[111]Liu X, Jiang F, Zhang R. A new social user anomaly
behavior detection system based on blockchain and
smart contract. In 2020 IEEE international conference
on networking, sensing and control (ICNSC) 2020
(pp. 1-5). IEEE.
[112]Shahin R, Sabri KE. A secure IoT framework based
on blockchain and machine learning. International
Journal of Computing and Digital System. 2021.
[113]Jatoth C, Jain R, Fiore U, Chatharasupalli S.
Improved classification of blockchain transactions
using feature engineering and ensemble learning.
Future Internet. 2021; 14(1):1-12.
[114]Brinckman E, Kuehlkamp A, Nabrzyski J, Taylor IJ.
Techniques and applications for crawling, ingesting
and analyzing blockchain data. In international
conference on information and communication
technology convergence (ICTC) 2019 (pp. 717-22).
IEEE.
[115]Sayadi S, Rejeb SB, Choukair Z. Anomaly detection
model over blockchain electronic transactions. In 2019
15th international wireless communications & mobile
computing conference (IWCMC) 2019 (pp. 895-900).
IEEE.
[116]Yang L, Dong X, Xing S, Zheng J, Gu X, Song X. An
abnormal transaction detection mechanim on bitcoin.
In international conference on networking and
network applications (NaNA) 2019 (pp. 452-7). IEEE.
[117]Keshk M, Turnbull B, Moustafa N, Vatsalan D, Choo
KK. A privacy-preserving-framework-based
blockchain and deep learning for protecting smart
power networks. IEEE Transactions on Industrial
Informatics. 2019; 16(8):5110-8.
[118]Sureshbhai PN, Bhattacharya P, Tanwar S. KaRuNa:
a blockchain-based sentiment analysis framework for
fraud cryptocurrency schemes. In international
conference on communications workshops (ICC
Workshops) 2020 (pp. 1-6). IEEE.
[119]Bhargavi MS, Katti SM, Shilpa M, Kulkarni VP,
Prasad S. Transactional data analytics for inferring
behavioural traits in ethereum blockchain network. In
2020 IEEE 16th international conference on intelligent
computer communication and processing (ICCP) 2020
(pp. 485-90). IEEE.
[120]Fan S, Fu S, Xu H, Cheng X. Al-SPSD: anti-leakage
smart ponzi schemes detection in blockchain.
Information Processing & Management. 2021; 58(4).
[121]Yilmaz I, Kapoor K, Siraj A, Abouyoussef M. Privacy
protection of grid users data with blockchain and
adversarial machine learning. In proceedings of the
2021 ACM workshop on secure and trustworthy
cyber-physical systems 2021 (pp. 33-8).
[122]Hu T, Liu X, Chen T, Zhang X, Huang X, Niu W, et
al. Transaction-based classification and detection
approach for Ethereum smart contract. Information
Processing & Management. 2021; 58(2).
[123]Zarpelão BB, Miani RS, Rajarajan M. Detection of
bitcoin-based botnets using a one-class classifier. In
IFIP international conference on information security
theory and practice 2018 (pp. 174-89). Springer,
Cham.
[124]Bach LM, Mihaljevic B, Zagar M. Comparative
analysis of blockchain consensus algorithms. In 2018
41st international convention on information and
communication technology, electronics and
microelectronics (MIPRO) 2018 (pp. 1545-50). IEEE.
Sabri Hisham earned his Bachelor's
degree in computer science (Industrial
Computing) from Universiti Teknologi
Malaysia (UTM) in 2001 and his
Master's degree in software engineering
from Universiti Malaysia Pahang
(UMP) in 2014. He is currently a PhD
student at Universiti Sultan Zainal
Abidin's Department of Computer Science in the Faculty of
Computing and Informatics at Terengganu, Malaysia. He is
also the Head of Infostructure at Universiti Malaysia
Pahang's Information and Technology Department. He is
also a Blockchain Solidity Smart Contract and Ethereum
Expert and certified professional. Blockchain, Bitcoin, ML,
IoT, Mobile App, Web Application, SCADA and
Telemetry System are among his current research interests.
Email: sabrihisham@ump.edu.my
Prof. Mokhairi Makhtar earned his
PhD in 2012 from the University of
Bradford in the United Kingdom. He is
currently a Professor at Universiti
Sultan Zainal Abidin in Terengganu,
Malaysia, in the Department of
Computer Science. Machine Learning,
Ensemble Method, Data Mining, Soft
Computing, Timetabling and Optimisation, Natural
Language Processing, E-Learning, and Deep Learning are
some of his current research interests.
Email: mokhairi@unisza.edu.my
Mr. Azwa Abdul Aziz earned a
degree in computer science in 2002
from Malaysia's Universiti Teknologi
Mara. He completed his studies at the
Bachelor's level and graduated from
Universiti Teknologi Mara, Malaysia,
in 2004. Then, in 2010, he earned a
master's degree in computer science
from Malaysia's University of Malaysia Terengganu. He is
recently a lecturer at the Department of Computer Science
at Sultan Zainal Abidin University in Terengganu,
Malaysia. Included in his research interests are Big Data
Analytics, Text Mining, Business Intelligence, and
Machine Learning.
Email: azwaaziz@ unisza.edu.my

Sabri Hisham et al.
1396
Appendix I
S. No.
Abbreviations
Descriptions
1
AI
Artificial Intelligence
2
API
Application Programming
Interface
3
AUC
Area Under the ROC Curve
4
BT
Blockchain Technology
5
CA
Centralized Authority
6
DApps
Decentralized Applications
7
DeFi
Decentralized Finance (DeFi)
8
DL
Deep Learning
9
DLT
Distributed Ledger Technology
10
DoS
Denial of Service
11
DPOS
Delegated proof of stake
12
DT
Decision Trees
13
ELM
Ensemble Learning Methods
14
EM
Ensemble Method
15
EVM
Ethereum Virtual Machine
16
FN
False Negative
17
FNR
False Negative Rate
18
FP
False Positive
19
FPR
False Positive Rate
20
GCN
Graph Convolutional Network
21
IF
Isolation forest
22
IoT
Internet Of Things
23
KNN
K-Nearest Neighbour
24
LR
Logistic Regression
25
LSTM
Long short-term memory
26
ML
Machine Learning
27
MLOps
Machine Learning Operations
28
MSP
Membership Service Provider
29
NB
Naïve Bayes
30
NFT
Non-fungible token
31
NN
Neural Network
32
OCVM
One-Class Support Vector
Machine
33
P2P
Peer to Peer
33
PBFT
Practical Byzantine Fault
Tolerance
34
PCA
Principal Component Analysis
35
PoA
Proof of Authority
36
PRISMA
Reporting Items for Systematic
Reviews and Meta-Analysis
37
PoS
Proof of Stake
38
PoW
Proof of Work
39
RF
Random Forest
40
RL
Reinforcement Learning
41
SD
Standard Deviation
42
SCM
Supply Chain Management
43
SVM
Support Vector Machines
44
TNR
True Negative Rate
45
TPR
True Positive Rate
46
UAV
Unmanned Aerial Vehicle
47
WoS
Web of Science

© 2022. This work is published under
http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding
the ProQuest Terms and Conditions, you may use this content in accordance
with the terms of the License.