Content uploaded by Mangayarkarasi Ramaiah
Author content
All content in this area was uploaded by Mangayarkarasi Ramaiah on Oct 19, 2023
Content may be subject to copyright.
An Anomaly Detection on Blockchain Infrastructure using Artificial Intelligence
Techniques: Challenges and Future Directions – a Review
Vasavi Chithanuru1,Dr. Mangayarkarasi Ramaiah1
School of Information Technology and Engineering, Vellore Institute of Technology, Vellore 632014, India .
Abstract:
Recently, Blockchain cryptographic distributed transaction ledger technology finds its usage in
many applications. The application’s ledgers implemented through Blockchain, ensures tamper-
proof transactions, and in turn the applications became robust enough against cyber-attack But
still adversaries put forward their efforts in detecting the vulnerabilities in the infrastructure to
execute their ill intent. In the literature, many counter measures techniques are presented to
address the security breaches on the Blockchain. Detecting as well mitigating from the possible
anomalies against on blockchain infrastructure through AI techniques is the greatest attempt of
this article, and which is much needed now. Hence, this review article enlightens the readers
with the essence of cyber security, the security aspects of Blockchain, its infrastructure
vulnerabilities, various Blockchain-enabled use cases along with the their challenges. Primarily,
anomaly detection on Blockchain infrastructure through Artificial Intelligence Techniques is
focused. A detailed analysis of Artificial Intelligence Techniques in detecting the anomalies
with the help of Blockchain and also how these two technologies complement each other was
demonstrated with the help of suitable use cases. The merits, challenges along with the possible
future directions, while integrating Blockchain with Artificial Intelligence Techniques are
presented for the benefit of research community.
Keywords: Blockchain, cyber-attack, Anomaly Detection,Artificial Intelligence, Machine
Learning, Deep Learning, Federated learning.
1. Introduction
A cyber-attack is a deliberate exploitation of computer systems, resources, networks, and
technologies connected through the World Wide Web. [1] Nowadays, there is a surge in cyber-
attack because of the greater number of devices linked through the internet. [2] Cyber criminals
snoops the vulnerabilities through hacking and malicious bots to harm the system. Cyber
security is procedure for ensuring the integrity of networks, software, and data against various
types of attacks or harms, by preventing unauthorized access. cyber security features prevent
threats, detect vulnerabilities, and protect data from unauthorized users. Table 1 highlights the
features of cyber security briefly.
TABLE 1: Key aspects of cyber security
The key essentials of cyber security are mentioned in Figure 1.
Features of cyber security
Description
Protection
To protect the system from unauthorized code. Firewalls or Antivirus are
used.
Confidentiality
Data availability should lie only with the intended recipients.
Reliability
Deals with data preciseness,
Privacy
Deals with the privileges associated the user
Technologies/ Techniques
State-of-the-art software and hardware enhancements are used to give
security awareness.
FIGURE 1: Areas of cyber security
Ensuring cyber security should not be an event. Instead, protecting the organizations' assets
should be a continuous process. Cybercriminals are becoming savvy enough to find
vulnerabilities before the software detects them. Various cyber-attacks like DoS Attacks, DDoS
Attacks, MITM Attacks, Phishing Attacks, Brute Force Attacks, and SQL Injection are raised
against the network infrastructure. Every sector is threatened with different types of threats, all
of which lead to financial loss by leaking sensitive information Report presented in [3]
reviewed the incidence of cyber breaches, malicious code, session hijacking, flooding, traffic
analysis and monitoring, traffic jamming, and eavesdropping anticipated in mobile ad-hoc
networks. [4] The Internet of Things' contribution is phenomenal for the smart environment,
hence security threats against them are inevitable. IoT (Internet of Things) threats and
countermeasures were examined against IoT layered architecture (Transport and
Application/Network/Perceptron). Connecting physical and software entities to the Internet is
entitled to “The Internet of Things". The security mechanisms, cryptographic techniques, and
robust authentication techniques facilitate detecting and preventing DoS attacks, Jamming
attacks, Replay attacks, and MITM attacks. [5] Presents various solutions to detect Phishing
Attacks in Industry 4.0. Phishing attack detection is attempted with Machine and Deep
Learning and Data Mining techniques. URL scan and blacklisting, and whitelisting methods are
the other possible choices for designing a defense method meant for phishing attacks. [6]
Server Resident Web application Firewall (WAF) addressed to detect SQL Injection attacks on
web applications. The user/hacker connects to the website over the Internet, routed through the
Network Firewall, where the server receives the user's requests. All the Requests are processed
through the proposed Firewall (WAF) before it has been sent to the server. WAF able to
prevent the five forms of SQL Injection attacks: redundancies, union queries, warehoused
events, and piggy-backed and rationally improper hunts. The server will alone process the
request that does not belong to the above-mentioned category. An alert message will be popped
for the malicious request.
The study reported in [7] describes an experiment to evaluate the efficiency of proxy servers
and the security they offer. Authors are particularly curious in portraying the evolving nature of
the web which has been impacted through the functionality of proxy servers and the level of
protection they offered. To assess the effectiveness of proxy servers, access logs from three
different proxy servers over a range of period are collected. Records are examined through
Webalizer. Then the performance of the proxy servers assessed through performance metrics. A
Firewall is used to protect network assets. To do so, Firewall uses three approaches, circuit
proxy, application proxy and packet filtering. Firewall, filters incoming packets from different
sources. Upon approval of the incoming packet, the circuit proxy will change the address with
its intended destination address. To make a change in the header, it needs to process the
resources and the same has been considered as one of the demerits associated with the circuit
proxy. Appropriate authentication is carried out by application proxy, which detects the
possibility of threats. Application proxy acts as a gateway because host machines are running on
them. Therefore, unauthorized access can be prevented through Firewall and again security
measures are need to tightened against viruses and Trojan horses [8].
Proxys servers and Firewalls couldn't succeed in attempting all sorts of cyber threats, hence,
many researchers and commercial vendors are presented Intrusion Detection System (IDS)
software designed using signature, anomaly and hybrid methods. A detailed study of IDS can
be found in [9]. As discussed in the references [7,8,9], proxy servers, firewalls, and IDS also
couldn’t cope with the dynamic threats. In other context, Proxy servers like Virtual Private
Networks (VPNs) and firewalls are centralized systems records issues in terms of network
connection and minimal possibility of backup, but there is no graceful degradation of the
system. So, decentralized systems are considered as one of the alternative solutions to minimize
performance bottlenecks and ensure more autonomy and control over resources for the
centralized systems. The decentralized systems find difficulty in accomplishing big global tasks
and finding failed nodes in the network. Such issues can be mitigated through decentralized
system, establishes the peer-to-peer connectivity quickly identifies node connectivity failure.
Moreover, in the context of distributed system, IDS works better than that of proxy servers and
Firewall. But, still IDS unable to detect the network scams and also find difficulties in
analyzing the encrypted traffic [10]. Such issues can be solved through blockchain
technology[11]. With the merits of Blockchain, the network security vulnerabilities can be
reduced [12] and [13]. The benefits of including Blockchain technology is represented in
Figure 2.
FIGURE 2: Advantages of Blockchain
Blockchain has been flashing on the screen since 2008 [11], and bitcoin was also there in the
trend since 2009. The Distributed ledger technology connects nodes to process the service
request. The most frequent technical terms associated with the Blockchain technology are Peer-
to-Peer Network, Ledger, Node, Hash, Merkle Tree, Transactions, Smart Contracts, Mining,
Consensus Mechanism and Cryptography algorithms. Encryption, Decentralization, and
Consensus Principles are foundations of Blockchain Technology, ensuring transactions'
trustworthiness. While connecting blocks, chains are made more efficient by the Merkle Tree, a
security feature of Blockchain.
FIGURE 3:
Blockchain Structure
Each block consists of header and body as shown in the Figure 3. Header entails version,
previous block hash, Merkle root, Timestamp, Bits and Nonce. The body part consist of the list
of transactions. Every block is connected through cryptographic techniques to form an
unbreakable chain. Blockchain makes tamper-proof data which makes modification harder.
Only authorized persons can access the information. Such aspect ensures confidentiality and
privacy. Due to the inherent merits of Blockchain, Healthcare [14-16], Banking [17-19], Supply
Chain Management [20-22], and the Internet of Things [23-25] are relying on it to enhance the
security aspects. Figure 4 portrays the statistical data of Blockchain technologies used in
various sectors. By observing the details in Figure 4, the finance sector heavily utilized
distributed technology in a best possible ways. Though it rendered many security benefits,
distributed ledger technology is not exempted from Cyber-attacks. In the Blockchain platform,
Bitcoin is one of the famous cryptocurrencies, prone to malleability and time hijacking attacks.
Through malleability attack, attackers change the signature of the transactions while it's being
validated, and the changed signature leads to new transactions. Hence, adversary broadcasts the
two transactions, triggers Double Spending threat. Quantum attack targets the Blockchain's
cryptographic portion, leading to hash collisions. Attacks like Selfish-mining, stake-bleeding,
bribery, refund and balance mainly occurred, when mining the transactions through a consensus
mechanism. The possibility of Sybil attack is high during the identification of nodes in a peer-
to-peer network. A Sybil Attack might be used in several ways, including tampering the online
elections or disrupting P2P (Peer-to-Peer) network traffic. A P2P network can be tricked using
Sybil Attack by using numerous false identities. These various identities appear to the viewer to
be typical consumers, but in reality, a single entity manages all these fictitious entities
simultaneously.
FIGURE 4: Usage of BCT in Various Sectors
In blockchain, all the nodes retains its own copy of the data. Data decentralization attracts many
applications and is also very essential too. However, data decentralization imposes challenges
for hackers while accessing sensitive data. For example, a Domain Name System (DNS) Attack
is used to obtain the IP addresses and the name of the organization's website, making the
website inaccessible to the users. Decentralizing the website's DNS records through Blockchain
can prevent such threats and ensure the system's availability to legitimate users. In most of the
occasions hackers use Routing, Phishing, Sybil, and 51% Attacks to endanger the Blockchain
platform. Authors in [26] discusses the Blockchain concepts along with the inherent security
options. The strength of various security and privacy techniques suits for the Blockchain,
including Mixing Protocols, Attribute-Based Encryption (ABE), Anonymous Signatures,
Secure Multi-Party Computation, Non-Interactive Zero-Knowledge (NIZK) Proof,
Homomorphic Encryption (HE) and Game-based Smart Contracts are too discussed. The
vulnerabilities in cryptocurrencies, Bitcoin and Ethereum leads to huge financial loss [27]. The
logic of the Byzantine General's Problem is how to establish a mutual understanding among
multiple parties. The complexity of this problem goes high as we connect more computers to
the network. Internet of Things-enabled applications are increasing now; due to the
unavailability of consistent protocol and centralized storage, the possibility of cyber threats is
increased [28] . To solve the security and privacy issues from the malicious entity and data in
Transit in the context of IoT are enhanced using trusted distributed authentication. IoT
Scalability is addressed using a Distributed System. The critical elements like Digital
Signatures, Mining Processes and Smart Contracts of Blockchain technology significantly
improve IoT security issues. A DDoS Attacker can slow down the system by flooding the web
with more transactions. DDoS and rogue devices attack can be prevented by designing a
deterministic smart contract-based IoT device and server communication framework as
presented in [29]. The suggested system sets a static resource limit to distinguish between
trusted and untrusted devices. [30] In healthcare applications, the availability of sensitive
information poses severe privacy and security threats to patients and healthcare service
providers. Hyperledger Fabric is an appropriate solution for such kind of application to ensure
privacy.
The authors [31] presented a comprehensive review of the anticipated vulnerabilities on
Blockchain platforms. Every layer needs security concerns against various cyber threats in the
layered Blockchain architecture. Layer-wise attacks and the possible countermeasures are
portrayed well. For example, expected Attacks in Data Layer are Malleability Attacks, Time
Hijacking Attacks and Quantum Attacks and the Attacks in Network Layer are DDoS, Eclipse,
Sybil, BGP Routing, and Phishing. The possible Attacks in the Consensus and Incentive Layer
are a Selfish Mining Attack, Double Spending Attacks like Vector 76 Attacks, 51%
vulnerability, Race Attack, Balance Attack and Bribery Attack, Stake-Bleeding Attack, Refund
Attack, and Block Withholding Attack. And finally, cyber-attack in the Contract Layer are
Reentrancy, Short Address, and Integer Overflow.
Apart from the above discussed cyber-attacks, the factors like lack of cyber security skill
crisis,new inventions and updations of blockchain techology platforms for developers and
stakeholders are also greatly affecting the security aspects of Blockchain infrastruture.
In the literature, various countermeasure techniques are presented in multiple venues to
complement Blockchain operations and mitigate from the security breaches. To detect the
anomalies on the blockchain key components many researchers presented various solutions
using Artificial Intelligence(AI) Techniques. Deep Learning (DL) is widely used in many
applications [32]. DL is a subset of AI. In real-time applications of Deep Learning Technique-
enabled features excel in biometric security and face recognition applications. In recent years,
DL has attracted a lot of attention because of its ability to infer more insight and facilitate the
making of appropriate decisions. [33] In many use cases, DL systems relying on centralized
servers face challenges, while ensuring transparency, traceability, reliability, security, and
trustworthy data provenance. Learning models built upon centralized data may lead to a single
point of failure. The authors analyzed the current state of DL frameworks purely meant for
Blockchain and the merits and the demerits. The article reveals the challenges expected from
the Blockchain infrastructure: Platform scalability,transaction execution latency, ensuring
secure data Exchange, Big Data, platform interoperability support, safe economic models, and
computationally expensive consensus protocols.
The challenges incurred due to the shortage of limited training sample and imbalanced out class
label distribution are efficiently handled by the recurrent attention model [34] built upon Deep
Learning architecture. Due to included attention mechanism, the presented Deep Learning
model able to extract the latent features which are highly cruel for pattern classification. There
is a huge necessity for some use cases to analyze huge dataset to infer the spatial and temporal
features in the presence of noise. Recently,Convolutional Recurrent Network Architecture
demonstrated its merits in analyzing the EEG signal [35] to interpret human intentions. The
merits of CNN and RNN architectures are integrated for the development of robust BCI (Brain
computer interface). Another serious concern associated with the Anomaly Detection Model is
appropriate feature extraction in the presence of minimal samples. Deep Learning based Semi
Supervised models demonstrates its ability in extracting the local features [36] through small
scale dataset in order to enhance the Classification Process.
In many contexts, integrating Blockchain with DL is well appreciated [37]. While analyzing the
various research topic between January 2018 and August 2020, Blockchain and DL terms are
heavily used by multiple researchers. The search reveals that Blockchain technology is adopted
by various domains, including finance and trade, and logistics, smart contract, and information
security. The number of publications in 2018 was 6, 12 in 2019, and 25 in 2020. The number of
publications shows that Blockchain and DL-related aacademic topics have been increasingly
studied and published. The following venues IEEE Access, IEEE Internet of Things Journal,
IEEE Communications Magazine, IEEE Network, IEEE Transactions on Industrial Informatics,
and Neural Computing Applications publishes good work for the candidate topics blockchain
with DL. According to IEEE Access, which has published 10 articles connected to Blockchain
and DL, this implies that IEEE Access is the most popular magazine in publishing the article
related to Blockchain and DL research work.
while surfing the scopus database with the terms anomaly detection, blockchain and AI, the
search shows minimal number of articles. Hence, the presented article focus on summarizing
the strength of various anomaly detection techniques meant for blockchain infrastructure
through AI techniques.
Figure 5 shows the statistics about the number of documents published in the area of
Blockchain and DL from the Scopus database. From 2018 to 2023, a noticeable increase in the
volume of publications was distinguished. From the diagram, in 2018, there were 62 documents
published and in 2019 it increased to 109. But in 2020, 218 papers were published and
documents 344 will be published in 2021. In 2022, 347 papers are published. In 2023, the
numbers of documents are 13. Year by year, the number of articles is increasing in this area.
This indicates that there will be plenty of opportunities to work in this field in the coming years.
FIGURE 5: Year-wise Paper Publications
The key contributions of this article are as follows:
1. We present various cyber-attacks anticipated on blockchain key components along with the
analysis of appropriate counter measures for the benefit of early researchers.
2. We report the insight into the state of the art anomaly detection frameworks combines the
merits of blockchain and AI. Also we demonstrated how these two technologies
complementing each other.
3. We present multiple use cases benefited through the integration of Blockchain and AI
especially meant for smart environment.
4. We compare the performance of various anomaly detection using AI techniques in terms of
quantitative metrics.
5. We summarize the open research challenges and the possible extension wherever possible
may be tried upon the existing techniques.
Section 2 discussed the possible cyber-attacks on Blockchain key infrastructure along with the
mitigation techniques. Section 3 discusses Anomaly detection using the integration of AI
Techniques and Blockchain. Section 4 reports the various use cases benefitted through the
combination of AI and Blockchain. Section 5 highlights the merits of Anomaly Detection
Techniques through the obtained results along with the challenges and the future direction for
further research in the candidate topic. Finally, Section 6 concludes the review article.
2 .Cyber-attack on Blockchain Networks
This section reports the various cyber-attack possible against the critical components of
the Blockchain, along with the countermeasures. The Distributed ledger technology is prone to
51% Vulnerability, Double Spending, Transaction Privacy Leakage, Vulnerabilities in Smart
Contracts, etc. The flaws encountered in the Consensus,Transaction Verification, and
cryptocurrency application lead to security breaches. Bitcoin is more prone to cyber security
breaches than that other cryptocurrency like Ethereum, used by cybercriminals to execute their
illegal intent [38]. The user's private key generation is robust through the Elliptic Curve Digital
Signature algorithm. The possible security breaches may be anticipated from the intruder's end,
guessing the randomness used for the user's private key. In turn, it may lead to compromises the
confidentiality [39]. A distributed consensus approach was included to establish trust among
the participants. The chances of a 51% Attack was confirmed, when a single miner exceeded
50% of the network hashing power in a PoW-enabled Blockchain. The negative impact of such
kind of attack is that any transaction can be reversed and modified, and even new transactions
can be ordered.
A good analysis of Smart Contract vulnerability on Ethereum is presented in [40] and also
demonstrates some tools that could detect the vulnerabilities in the code. Every Smart Contract
should be designed in a deterministic manner. In Ethereum, every transaction has to be
processed by all the untrusted nodes in the network upon the Proof-of-Work. The smart
contracts produce the correct output unless the adversary controls it. The authors [40] list many
smart contracts, the possible vulnerabilities, and how attacks may exploit them to harm the
process through games. The title of "King of the Throne" is at stake in the game King of the
Ether Throne. Anyone who wishes to rule the world must pay for the contract and provide a
small amount of ether to the current ruler to obtain the title. Such a contract seems honest
initially, but it was designed with gasless send vulnerability. Governmental is a faulty contract.
To participate in the game, users must deposit a specified amount of ether into a contract. If no
one signs up within the next twelve hours, the last participant can claim all the money. The
possible vulnerabilities in King of the Throne are time-based constraints, stack size limits,
Immutable bugs, Exception disorder, Stack overflow, Unpredictable state and Timestamp
dependence. In another sample Rubix Ponzi Contract, participants benefit from the investment
of newcomers, a fake high-yield plan. Furthermore, the contract owner gains some money paid
to the agreement upon investing. Hacker exploited the flaw in "Immutable bugs" is used in this
hack to allow an attacker to steal some ether through the contract. Consider a contract with a
library of set operations that can dynamically update one of its components. As a result, the
contract can use the updated library version if a more efficient implementation of these
operations is developed or a defect is fixed. A single person cannot know which version of the
library will be used to complete a transaction. The Dynamic Libraries attack exploits the
"Unpredictable State" vulnerability presents a review report [41] on various anomaly detection
methods for Blockchain. Authors made extensive discussion on the possible anomalous entities
in Blockchain layered architecture. The presented framework uses statistical and ML
techniques to detect the anomalies. There are uncommon Attacks on Blockchain technology,
mainly based on Accounts, Smart Contracts, Consensus, Transactions and Systems which is
displayed in Table 2.
Attacks based on Blockchain
Problems
Account
Wallet Theft, Key Theft, Dusting, Crypto Jacking and Anomalous
Peers
Smart Contract
Honeypots, Coin Freezing, Reentrancy Dependent Tx Execution,
Bytecode Based Bugs
Consensus
Race Attack, Eclipse Attack, 51% Attack, Balance Attack, Fork
Formation
Transactions
Tx Tampering, Currency/Token Theft, Malicious Tx on Ledger,
Money Mixing, Double Spending
System
BGP Jacking, Modifying Logs, DDoS Attack, Divergent Paths,
Malicious Network Requests
TABLE 2: Blockchain Component Attacks
Cyber-attack need to be detected as early as possible. There is a significant loss in
cryptocurrencies due to Attacks in the Blockchain. [42] In 2013, and 2016, wallet attacks
occurred. With this attack, the US lost 70 million dollars. A "Dusting Attack" is used to
compromise the privacy of cryptocurrency users by sending a small amount of crypto to their
wallets. A Double Spending Attack was defined as spending the same currency for different
transactions. Because of this attack, there has been a rapid drop in bitcoin prices to the amount
of $175 million in March 2013. Another type of attack is BGP Hijacking. The complete form is
Border Gateway Protocol (BGP), a deserter protocol. Such an Attack primarily keeps track of
the recipient system's IP address to send packets that cause the network to be disrupted. From
this Attack in 2014 the attackers gained $83000 currency. There are two variants of such kinds
of Attacks; the first one is a Node-Level Attack, and the second one is a Network-Level Attack
[42] .
A Spam attack is a flood attack designed to interrupt Blockchain networks with a vast number
of transactions. Detecting spam in Blockchain networks is a challenging task. Still, networks
use different techniques like charging high transaction fees, prioritizing the transaction based
on throughput or reputation, etc., to mitigate the spam transaction attack. The DAO
(Decentralization Autonomous Organization) is a Blockchain-enabled organization meant to
accomplish collaborative tasks. Such DAO is vulnerable; hackers gain access to such an
organization by designing a smart contract with the title investor and investing some ether to
DAO, which later invokes the withdraw function. As per report in [42] due to the DAO attack,
money worth $60 million loss incurred in the US.
A Denial-of-Service(DoS) Attack is attempted on the availability of services for the legitimate
users by consuming the system resources beyond its capacity. DDoS (Distributed Denial-of-
Service) attack is similar to a DoS attack; however, it was launched from a succession of hosts
instead through a single host. The incidence of such an attack in the US incurred 20 hours of
network downtime and led to a financial loss worth $123,00. The Selfish Mining attack occurs
because of non-honest miners, which helps them secure rewards by wasting real-time. Through
this mining, the attacker gets more benefits like illegal rewards and the wasting of simple
miners. Here, the non-honest miners kept their miner's block private, while the honest miners
kept it in public mode. This attack led to $90,000 worth of capital loss in May 2018 [42] .
Cyber-attack anticipated in Blockchain networks and the possible countermeasures are
summarized in Table 3. Such attacks not only incur financial loss but also cause the system to
compromise the security aspects of sensitive data. Hackers steal money worth $3.78 billion
with 122 numbers of attacks. While analyzing the Blockchain-related cyber-attack, the statistics
reveal that one-third are targeted on Blockchain platforms in the total amount of hack time [43].
Eradicating the Hackers presence from the picture is impossible. Hacking was aimed with
different intents, but in most cases, it is meant for monetary purposes. Out of 100 %, 76% of the
attacks were aimed at Blockchain-enabled Finance Applications in the year 2021, which
incurred more than $1 billion loss in the third quarter alone. According to an analysis from
SlowMist, 20% more Blockchain-based hacking incidents happened in the third quarter of 2021
and 2020 [44]. Tabel 3 describes about various attacks and their countermeasures on BCT
platform.
Authors
Attack/malicious action
Methodology Used
[45]
DDoS Attack
DDoS attack detection using Mempool optimization allows legitimate
users discarding spam transactions.
[46]
Sybil Attack
Cyber threat intelligence is delivered using Blockchain technology to
prevent erroneous data and increase resilience against Sybil attacks.
[47]
Mining Pool Attacks
Presents a novel protocol SMARTPOOL for decentralizing the mining
pool. Protocol gives transaction selection control back to miners to reduce
the transaction fee.
[48]
51% Attack
A security-based framework had introduced with five (Penalty System,
DPoW, PirlGuard, ChainLocks and Merged mining) techniques to
alleviate the weakness in the consensus mechanisms.
[49]
Double spending Attack
Machine accepts the bitcoins with a OR code (QR Scan) to raise alarm on
the occurrence of double spending.
[50]
Silkroad Trader attack,
Marketplace trader attack
To prevent both attacks, propose a change to the BIP70 standard that
indicates the merchant with publicly verifiable evidence.
[51]
Transaction Malleability
Attack.
A technique to create a resilient refund transaction which making any
change in the Bitcoin to prevent the financial loss is presented.
[52]
Eclipse Attack
Adversary with enough IP addresses and time can raise eclipse attack.
Countermeasures inspired by bot-net architectures are present.
[53]
Eclipse Attack
Presents an anomaly detection tool, creates a database with malicious
forks to detect the attacks
[11]
Double spending Attack
System for electronic transaction without depending on trust is presented.
PoW records history of public transactions which are computationally not
feasible for an adversary to change when a trust node controls the majority
of CPU power.
[54]
Sybil, DoS Timing-Based
Inference Attacks
Presents a two-party mixing decentralized protocol to address the security
breaches.
[55]
Stake-bleeding attack
Possible incidence of long-range attack on PoS protocols without
checkpoints. Key evolving cryptography is sufficient for mitigation
[56]
Selfish mining
The Poisson property of Bitcoin's proof-of-work is used in the timestamp-
TABLE 3: Various Attacks on BCT platform along with the countermeasures
3.Anomaly Detection using integration of Artificial Intelligence Techniques and
Blockchain
Artificial techniques find its application in various domains due to its inbuilt merits in learning
anomalies. This section discusses how Blockchain-based applications could benefit through AI
techniques for enhancing security by detecting abnormalities. Applying AI techniques on the
Blockchain ensures the preciseness in Data Scalability, Predictability, Accuracy, and Reliable
decision-making. Blockchain is a well-known peer-to-peer transaction processing technology
that includes the AI-based Learning Models' decisions that may be easily validated because of
the traceability feature. As a result, the highly accurate AI technique-based learning model
gained the stakeholder's trust.
Earlier, AI dealt with the centralized repository. After the invention of Blockchain,AI
techniques, the resultant collaborative models are updated to improve scalability. Centralized
storage is not the solution for the application intent to ensure privacy and security, because it is
incredibly vulnerable, primarily when it deals with personal information about users' locations,
activities, health information, and finances. Furthermore, processing massive datasets by AI
applications on centralized infrastructure is hampered by scaling and capacity constraints.
Blockchain-based decentralized storage infrastructure makes storing data in an encrypted
format easier for the participating networks. An overview of distributed storage on Blockchain
technologies, its challenges, and possible solutions through AI techniques are presented.
Moreover, the challenges incurred in integrating AI with Blockchain are briefly summarized
[57].
The concepts of AI and Blockchain technologies vastly differ from each other in terms of logic.
Still, both of its intentions are to increase the quality of the use cases. One of the Blockchain
infrastructures, such as smart contract efficiency, can be improvised via AI techniques like
formal verification and search-based software testing. Likewise, through AI, the security
aspects can be wildly improvised. Unfortunately, scalability problems prevent a Blockchain
system from extending its more significant benefits.
Figure 6 displays the various AI Techniques used along with Blockchain technology for
detecting the various cyber-attack through anomaly detection.
free technique.
FIGURE 6: Learning techniques meant for Blockchain technology
3.1: Blockchain with Machine Learning(ML)
The unlimited merits of ML Algorithms are explored for many applications. However, the
results obtained through machine algorithms can be tampered by intruders accessing the
network. Hence, running ML applications on Blockchain networks ensures tamper-proof and
other security aspects. This section brought some of the recent benchmarking techniques to
portray the advantages of Blockchain technology in combination with ML Techniques
especially in detecting the anomaly on Blockchain technologies wherever possible.
Actual incidents of DDoS attacks against Bitcoin-related services are the source of the data for
the research [58]. Feature Extraction is done through Principal Component Analysis (PCA).
The experiment considers the following features, the name of the service, date, category of
service, number of posts, generated block during the attack and, eventually, details of the
transaction. DDoS attack detection precisely carried out through DL Models [59].
DDoS attacks are raised using different types of protocols used. To handle DDoS attacks, the
business application relies on third parties. ML techniques are suitable for judging, whether the
incoming packet is legitimate or not, and also to generate a blacklist of IP addresses in
Blockchain networks. Blacklisted IP addresses are effectively recorded using Random Forest
(RF), K-Nearest Neighbors (KNN) and Decision Trees (DT) in the Blockchain. The tested
results reveal that the Random Forest Approach outperforms the other two algorithms in terms
of performance metrics [60] .
The DDoS attack creates a significant threat to network security. One of the most destructive
DDoS Attacks is the DNS Amplification attack. In the context of software-defined networks,
authors proposed Brainchain, a scalable and effective method to defend permissioned
Blockchain nodes from the DDoS attack. Brainchain consists of four modules. First, the Flow
Statistics(FS) collection strategy is used to gather information about flows utilizing the flow
protocol. Second, the Entropy-based Technique (ES) automatically discovers network
anomalies. Bayes Network based Filtering scheme (BF). Finally, the DNS Mitigation (DM)
technique can effectively identify and mitigate DNS Amplification attacks using FS, ES, and
BF. According to experimental findings, the BrainChain has a high accuracy and a low false
positive rate when detecting and mitigating attacks (such as DNS Amplification attacks) [61].
Another security breach for the Blockchain enabled application is from DAO attack. An
unsupervised encoder decoder prototype developed along with the benefits of SDN (Software
defined networking) is presented and experimented on Ethereum classic dataset [62]. In
contradiction to the existing supervised ML methods, the proposed technique uses an
unsupervised model built upon irregular anomaly detection. Moreover, the experiment focuses
on sensing Zero-Day attacks, which are highly unpredictable. The authors disclosed that the
presented technique is superior in detecting DAO and 51% attacks and proved the same
through ROC(Receiver Operating Characteristic) metrics. Several organizations in private and
government sectors are collaborating with consortium-based Blockchain networks. Such
combination may anticipate a prevalent threat, from Majority attack provided collision occurs.
A supervised ML technique with game theory is presented to detect majority attacks on a
Consortium Blockchain [63].
Intrusion detection framework enriched with ML is another alternative solution for detecting
the anomalies on Blockchain network. A Blockchain-based intrusion detection framework
using DELM (Deep Extreme Learning Machine) is beneficial [64] for detecting adversaries
precisely. DELM model is built with back-propagation to adjust the weights, thereby reducing
the error rate. In terms of efficiency, the DELM framework outperforms the one built upon the
machine learning methods such as SVM, KNN and Decision Trees. The presented technique
was experimented with using the NSL-KDD dataset and the KDD-CUP-99 dataset and obtained
93.9% and 94.6% as its accuracy value [64] .
3.2: Blockchain with Deep Learning(DL)
DL models produce foreseeable precise predictions while detecting anomalous entities, whereas
Blockchain is to ensure data security. Characteristics of Blockchain are Decentralization,
Immutability, Data Integrity, and ultimately, the candidate applications built upon this became
attack resistant. How the Blockchain and DL Techniques complement each other in various
aspects are reviewed in [33]. A detailed taxonomy of categories for integrating Blockchain with
DL using seven parameters is analyzed, along with challenges and possible future directions.
Moreover, Blockchain-based DL frameworks categorization upon the Blockchain type, DL
techniques, dataset, and Consensus protocols are summarized to address the merits of these two
technologies [33] .
An efficient Stochastic Gradient Descent (SGD) algorithm for analyzing the anomalies in Big
Data in perception of security is presented in [65]. The proposed L-nearest aggregation scheme
ensures privacy for multi-party data and primarily protects against Byzantine attacks through
the global gradient aggregation process. The presented framework is experimented using
Python and Ethereum Platforms. The experimental results regarding the performance of the l-
nearest gradient reveal that the proposed work produces a lower test error rate, with 40% of the
Byzantine data holders, and the error rate goes high with the increase in the percentage of
Byzantine data holders.
ML Techniques work in an integrated manner, whereas Blockchain is a P2P network, hence
necessitating the concept of parallel ML Framework is attempted in [66]. A brief roadmap for
combining AI with Blockchain technology is presented. A framework that integrates the DL
technique with Blockchain technology is demonstrated. The hyper-parameter for DL is derived
through a meta-heuristic algorithm, and the preciseness purely depends on the nodes'
cooperation. The claimed merits are the communication delay incurred while transferring data
and model and the reduced waiting time for synchronization along with the security aspects.
Cyber-attack mitigation has becoming challenging day by day in Blockchain networks.
Especially. Most of the security breaches are occurred using DDoS, 51%, Transaction
malleability and Sybil attacks. Many researchers present technique to address such breaches to
some extent. DDoS gained accessed through the system using Blockchain Network layer,
which is expected to leave high level impact on resources., Hence, to mitigate from the DDoS
attacks, a CMCNN (Comprehensive Multilayer Convolutional Neural Network) model is used
by the author in [67] , which yields higher performance.
A robust anomaly detection framework meant for intelligent power networks is presented in
[68]. The proposed framework ensured privacy using two levels of privacy preservation and
anomaly detection module. Two privacy preservation groups provide legitimate transactions by
transforming all the data into a new format ePoW method in the first level. In the second level,
VAE (Variational Autoencoder) detects an abnormal pattern in the transformed data. Finally, an
LSTM-based anomaly detection is included to find out the irregularities in the data. The
presented framework uses the power dataset and UNSW dataset to complete the task; the
proposed techniques consider the power parameters from the power dataset and the network
parameters from the UNSW dataset. Hence, the presented framework is an appropriate solution
for smart power networks.
3.3: Blockchain with Federated Learning(FL)
FL is a new way to train AI models without touching the data, and it also paves the new way to
use ML techniques in a decentralized manner to ensure privacy. The volume of data in each
block makes it difficult for traditional centralized data mining techniques to handle. However,
cutting-edge AI algorithms, like FL, may absorb knowledge from diverse data sources, offering
an ideal solution for the Blockchain system. Moreover, Blockchain implementation issues with
AI techniques are applied [69] . As a result, IIoT produces more data, facing challenges while
preventing data leakage. A Blockchain-enabled data sharing framework [70] using FL explores
a new way of data security. The data sharing problem is considered a ML task by integrating
the decision obtained from all the nodes. Instead of sharing the data across the nodes, the
technique shares the trained ML model to prevent data leakage, thereby ensuring data privacy
among the multiple parties involved in the transaction. The presented framework is assessed
using two real datasets (the Reuters dataset and 20 newsgroups dataset); the experimented
results reveal that the proposed data-sharing scheme obtains high efficiency, good accuracy and
improved security.
As the number of operations increases, it causes latency; Blockchain-based Federated Learning
(Block FL) architecture may facilitate the exchange of ML model decisions gathered from
various nodes. However, the wireless system focuses on ensuring latency and reliability. Hence
the idea of on-device ML models is preferred. To make a precise decision, training samples are
shared across the nodes in the network. Each miner node communicates and validates its
decision through Proof-of-Work. Each miner in BlockFL broadcasts their local model’s
decision to other miners so that it can be validated. An optimal block production rate is
analyzed by considering communication, computation, and consensus delays [71].
One of the merits associated with FL, the exchange of raw data doesn’t require. FL offers the
potential to avoid direct data sharing, hence lowering the risks of data leakage by training
models locally at each client and aggregating learning models at a central server. However, the
conventional FL system is highly dependent on a single central server and can fail if that server
exhibits malicious behavior. To solve the single point of failure problem, clients are
safeguarded from malicious clients using the Blockchain-Assisted Decentralized FL (BLADE-
FL) framework, and a stable and self-driven learning environment is ensured [72].
A Federated learning-based system doesn't have any mechanism to detect client anomalies [73]
which may make the context more prone to Byzantine attack. Moreover, there are cases in the
model aggregation collection server that may become an untrust node, where there is no
guarantee that the parameter passed by them is legitimate. Hence, this may reduce the model's
prediction accuracy and encourage data leakage. A Blockchain-based federated learning
introduces the aggregation mechanism agreed upon by all the nodes to alleviate the need for a
centralized server and make the system robust enough against Byzantine attacks. Privacy
preservation is essential for the Internet of Vehicles (IoV) to prevent data tampering [74], and
also need to collect data from different sources. Hence FL would be the appropriate choice, but
still, some issue is associated with the FL-based system. Blockchain assists well in detecting
anomalies more than the conventional FL-based system while preventing the IoV from causing
road accidents.
ML and DL techniques are explored in detecting anomalies by various researchers. Now a
day's, adversaries are very smart enough to find vulnerable places for executing their illegal
tasks. The FL data maintained in multiple sites may be brought together for better monitoring.
Still, the biggest challenge associated with the FL model results performed on different nodes is
collected in a centralized server. There is a chances intruder may feed the data, which might be
inappropriate. In such cases, the inclusion of Blockchain is an excellent solution to enhance the
outcome of the FL-based system. [75] Solution to record all participants' model updates on the
distributed would solve the problem. Though none of the models' results can be evaded,
identity will also be preserved. Blockchain domain name service is difficult to address because
of the inherent challenges, high data, and lack of labels. Integrating Blockchain with FL [76] is
an excellent choice to find anomalies in Blockchain domain name service. The Blockchain with
an FL-based system ensures the security model transfer and data storage.
4.Integration of Blockchain with Artificial Intelligence- use cases
Now a days, Blockchain technology is considered as most promising platform for developing
secured applications. This section describes various benchmarking use cases, built upon the
combination of Blockchain and Artificial Intelligence(AI) techniques. Though tightened
security measures are in full swing for smart applications meant for Industry 4.0, preventing
security breaches is still challenging. The advancements in AI and Blockchain are considered
promising solutions to implement the security aspects.
A novel distributed consensus protocol Proof-of-Learning using ML systems to validate
Blockchain transactions. WekaCoin is a cryptocurrency based on the Proof-of-Learning
distributed consensus system that works on a peer-to-peer basis. This framework consists of
three components, suppliers who publish the ML task. Validators are responsible for ranking
the models and suggesting new chains of blocks. The trainer with the best model is rewarded
with a set of WekaCoins. Thus the presented work removes the computational waste incurred
while solving the puzzle [77].
Cryptocurrency plays a vital part in today's world and is widely accepted as a means of paying
and exchanging money. However, cryptocurrency is known for its tremendous volatility and
price variations over time. Bagging, Stacking, and Ensemble-averaging are three Ensemble
Learning approaches described by the authors, along with sophisticated DL models, including
Long Short-Term Memory(LSTM), Bi-directional Long Short-Term Memory, and
Convolutional Layers for predicting cryptocurrency hourly prices. Through DL techniques,
hourly price forecasting is done for Bitcoin, Ethereum and XRP. Performance is calculated
using metrics, and the results can also be summarized [78].
The authors designed an anomaly detection framework for industrial control systems (ICS)
attacks like the final attack. This framework has two stages. In the first stage, Blockchain-based
log management is attempted, and in the second phase, a deep learning-based Secure Water
Treatment (SWaT) is built upon the features collected from physical and network-level traffic.
The resultant framework obtained 95% in terms of precision [79].
The E-voting system is one of the internet infrastructures enabling vote casting. Internet
increases the possibility of security breaches, especially in Identity verification. For securing
the E-Voting system, integration of Blockchain with ML models is proposed. The presented
work uses personal and public Blockchains to facilitate voter registration. A public Blockchain
is used to ensure data integrity for voters. The proposed work use ML methods, Gaussian
Vector Support Machine (GVSM) and Linear Vector Support Machine(LVSM), to detect the
intrusion while exchanging information through the network. The efficacy of the intruder
detection is measured through the metrics AUC (Area Under the Curve) and Accuracy [80]. A
decentralized solution is more viable than the centralized one to ensure Transparency. A secure
decentralized framework for a smart city is explained in [81]. The Blockchain is implemented
at the fog layer to achieve decentralization and security advantages. DL is used at the cloud
layer to improve autonomous data processing and boost the connection bandwidth intended for
smart factories and manufacturing applications. And the presented framework experiments with
a car manufacturing case study, and its results are compared with those from other frameworks.
Intelligent peer-to-peer energy trading between prosumer and consumer is introduced in [82].
The presented energy trading is built upon a permission Hyperledger Fabric platform. Data
Mining techniques are used to analyze energy consumption data patterns. The decentralized
crowdsourced energy trading transaction records, real-time energy trading, day-ahead energy
trading, predictive short-term energy results, and individual user records are all protected
through Blockchain. Predictive analysis is also carried out using ML techniques to predict
short-term energy demand. Various DL approaches like Recurrent Neural Networks, LSTM
and Bi-directional LSTM for prediction approaches are also used for the candidate task. In
terms of Regression Performance measures, the proposed model performed well. The
prediction outcomes demonstrate that the tested results with ML models using time-series data
produce higher MAPE values than the results through LSTM.
The security of an intelligent transportation system (ITS) faces the issues like data integrity and
trust. To ensure these issues, the authors proposed Blockchain based Intelligent Transportation
system with Outlier Detection for Smart City (BITS). An anomaly ML model is built upon the
dataset with 10% of outliers. The system ensures 100% in terms of precision and recall [83].
Applications of IoT are very high, and ensuring privacy is the biggest concern for IoT-enabled
Smart Applications. The authors presented a Privacy-Preserving and Secure Framework (PPSF)
for IoT-driven smart cities. This framework is mainly used to avoid smart city issues like
privacy (e.g., performing data poisoning and inference attacks), centralization, scalability,
security and transparency. In terms of the privacy preservation scheme, PPSF has two levels.
The first level involves Blockchain and an enhanced Proof of Work (ePoW) approach based on
smart contracts using the Ethereum network. Also authenticates the IoT data records and
prevents modified original data through data poisoning attacks. The second level, Principal
Component Analysis (PCA), avoids Inference Attacks while transforming the authenticated IoT
data into a new encoded format. To measure the potentiality of the PPSF framework in finding
the smart city network cyber-attack, the authors designed a system called Gradient Boosting
Anomaly Detector (GBAD) and applied it for training and evaluating two-level privacy scheme
based on two datasets, ToN-IoT and BoT-IoT. The authors suggested a Blockchain-
InterPlanetary File System (IPFS) integrated with fog-cloud architecture to deploy the proposed
framework. With the transformed dataset, GBAD used two datasets; the results have been
achieved better in terms of performance metrics. GBAD using ToN-IoT dataset, Accuracy-
98.3%, Precision-95.32, Recall-94.35%, F1 score-94.80%. GBAD using BoT-IoT dataset,
Accuracy-99.9%,Precision-98.01%,Recall-97.24%, F1 score-97.59%. While detecting attacks,
the proposed model got an average of 91%-100%, which is higher when compared to others.
After employing the suggested two-level privacy preservation strategy (transformed dataset),
the proposed GBAD performs better than existing models like Random Forest, Decision Tree,
and Naive Bayes [84].
The authors suggested a wastewater recycling control system to effectively manage the
wastewater and coordinate it among the companies and the government. To store data and
create an incentive scheme to promote wastewater reuse, Blockchain technology has been
implemented. Industries will receive tokens based on the quantity and quality of recycled
wastewater. A Smart Contract is used for issuing and trading these tokens. Furthermore,
anomaly detection techniques are utilized to detect potential fraud in the system due to IoT
metre data tampering. The system uses IoT metres to measure the volume of wastewater
produced and reused and quality indicators like hardness, oil content and pH. To detect the
tampering, multiple ML algorithms like Polynomial Regression (PR), Density-Based Spatial
Clustering of Applications with Noise (DBSCAN), AutoEncoders (AE) and Long Short-Term
Memory (LSTM) networks. Performance is measured in terms of metrics Accuracy, Precision,
Recall and F1-score. Among all the ML models, DBSCAN got better results with
94%,94%,99.69%,96.93% [85].
Blockchain technology can assist in billing, patient and healthcare records, and drug
management. Blockchain and ML can better automate various healthcare sectors' operations.
The authors presented an efficient use case for managing healthcare data [86] . The presented
application uses a Blockchain database to ensure data integrity and supervised ML techniques
are used to extract the relevant features of patients' healthcare parameters. Details are
categorized in terms of diseases to assist healthcare professionals. Another use case [87] meant
disease diagnosis using DL Techniques and secure distributed data transmission is achieved
through Blockchain. Another sector, which may highly benefit from Blockchain technology is
Agriculture. Authors in [88] presented a Tea Traceability system using Blockchain and ML
techniques to manage all the information about the whole process life cycle. Data security is
included using Blockchain data collected from various IoT devices. ML techniques are used
with blocks to enhance the performance of the system. Summarization of various use cases
powered by Blockchain and AI techniques can be found in Table 4.
Ref no
Applied sector
Blockchain Type
Framework/AI/Technique
[64]
Smart Home
Public Blockchain
DELM
[68]
Smart Power
Private Blockchain
VAE,LSTM
[79]
ICS,Smart Manufacturing
Private Blockchain
Multi-Source Deep Learning
[80]
E-Voting
Personal and Public Blockchain
GSVM & LSVM
[81]
Smart Factories and
Manufacturing
Consortium Blockchain
Deep Learning
[82]
Energy Trading
Hyperledger Fabric
RNN,LSTM and biLSTM
[83]
ITS
Distributed Ledger
SVM,RF,MLP,K-means and
Isolation Forest Model
[84]
Smart City
Ethereum
PCA,GBAD
[85]
Wastewater reuse
Hyperledger Fabric
PR,DBSCAN,AE and LSTM
[86]
Health care Data
Hyperledger Fabric
BOW(Bag Of Words)+ML
[87]
Health care data transmission
Hyperledger Fabric
MFO(Moth Flame Optimization)
+ResNet-v2 + SVM
[88]
Agriculture
Ethereum
Ensemble Method
TABLE 4: Blockchain Enabled Use Cases
5. Results and Discussions
This section analyses the performance of various anomaly detection techniques
implemented for Blockchain infrastructure using AI discussed in section 3. A brief discussion
about the Malicious attacks on a Blockchain key infrastructure and how the security threats are
prevented through different AI techniques is portrayed along with the challenges.
Autoencoder's technique for detecting Blockchain Anomalies is presented in [89]. Recurrent
Autoencoders are applied to the Ethereum classic network to identify cyber-attack developed
using the dataset collected from Ethereum classic between July 2015 to July 2019 and available
on Kaggle [90]. The experimented dataset consists of seven tables for storing details regarding
transactions, blocks, logs, tokens, traces, contracts, and token transfers. Among the seven
tables, the authors had chosen the tables consisting of blocks and transaction details alone. The
data is preprocessed, and the features provided gas average, block size average, transaction
average per block, transactions number, block difficulty average, and gas used sum are chosen
to design the model. The most probable attacks anticipated on the Ethereum classic networks
are DAO and 51%. Therefore, the selected features are facilitated well in detecting the attacks.
The experiment result on detecting the anomalies at distinct points in time is displayed in
Figure 7. The figure states that the x-axis is the outlier scores, and on the y-axis, timestamps are
taken where the anomalies are detected. Day 1255 (where the anomaly score is the highest) and
day 1291 also appear to be the days where all the frozen activities were recorded on the
network, thus representing the actual moment the anomaly was injected in the Blockchain.
FIGURE 7: Identification of anomalies in the Blockchain using RAE [89]
Cryptocurrency plays a significant role in digital transactions, gaining more attraction because
of its included cryptography techniques. Though the security aspects are ensured using the
cryptographic method, cryptocurrencies are still prone to DDoS attacks, 51% vulnerability,
Double Spending Attacks, Selfish Mining, and other threats. Authors in [91] proposed two
techniques: Resilience to DDoS,51% vulnerability, and Double Spending attacks. One class
Support Vector Machine and K-means clustering experiment on the dataset consisting of
bitcoin transaction details to classify the attacks. The authors took Bitcoin Transaction
information from a Blockchain-based data source. The experimented model obtains 93% as its
attack detection accuracy. The potential researcher can give attempt via some other techniques
so that the attack detection accuracy may well improvise.
Anomaly detection on financial networks using Bitcoin is attempted by designing three
unsupervised ML methods [92]. The model is built upon a dataset that comprises the details of
users and their executed transactions. Due to the enormous size of the features of a dataset, the
parts are analyzed using user and transaction graphs. The graph-based visualization also
facilitates anomaly detection. Individually, anonymous user and anonymous transaction
detection became possible, but finding the intruder entity in both cases became impossible. An
unsupervised model using K-means clustering, Support Vector Machine (SVM) and Mahala-
nobis distance is presented to check the reliability of the results obtained through the graph-
based analysis. The authors compute the ratio obtained from the anomaly distances and their
respective centroids upon the top hundred abnormal entities to calculate the model's efficacy.
A model for detecting potential security holes in the Ethereum Blockchain was highlighted [93].
The proposed framework consists of two phases. In the first phase of the framework, the
candidate task is attempted using a Deep Neural Network and in the later stage tried with ML
techniques. While assessing the efficacy of the deep neural-based model, based on quantitative
metrics produces 97%, 99%, 98% and 97% against the Accuracy, Precision, Recall and F1-
score, respectively. When attempting the exact Attack Detection through ML Techniques, K-
Means and DT (Decision Tree) performed better than other methods. The tested results against
the Performance Metrics are as follows, in terms of accuracy, the attempted model produces
99.4%, and in terms of recall, precision, and F1-score, the designed model makes 99.5%.
The Authors attempted to reveal the characteristics of the nodes that participated in the
transaction on the Blockchain platform [94]. The assumption is the participated node doesn’t
reveal its identity to preserve anonymity. Hence, the presented framework demonstrates the
possibility of revealing the identity of nodes based on the transaction they executed through
supervised ML. To illustrate the proposal, 9,000 bitcoin addresses are considered with the
following target: Mining pools, miners, gambling, coinjoin, exchanges and services. After
collecting the addresses to compute the features of lessons, two techniques were used based on
the address statistics and node2vec embeddings. Feature importance is calculated through RF
(Random Forest). Results from machine models are summarized using a variable number of
transactions and features. The efficacy of the system is measured through F-score. According to
the experimental findings, RF produces superior results than the other approaches. Another
insight from the experimentation is, usually, as we increase the number of transactions, the
results will be more precise, but in contradiction to this fact, though the experiment extended
with 1000 transactions, after 500 numbers of transactions, the model didn’t show any
improvements in the F-score. The discussion on results also confirms that the RF classifies the
addresses very precisely into appropriate classes with the 100 numbers of transactions, and the
corresponding numerical F-score is 96%. Through this study, it is apparent that the possibility
of de-anonymization in some contexts exists in Blockchain networks, and countermeasures
must be developed to address the same.
The authors demonstrated that ClaimChain is a promising solution and also a cutting-edge
alternative for the NICB/ISO database architecture. [95] Presented a model for application-level
security built with ML models like K-Nearest Neighbours (KNN), Logistic Regression (LR),
XGBoost and Random Cut Forest (RCF) and assessed using various performance metrics.
Dataset consists of information about the insurance vehicle claims regarding undisclosed
insurance, where 15,530 samples are legitimate claims, 924 claims are fraudulent, and each
claim record consists of 31 attributes. In addition, the XGBoost-based ML model obtains 98%
Accuracy, Precision, Recall and F-score in fraud detection through simulation.
The authors invented the shared child nodes concept to address Smart Contract vulnerabilities
[96]. And also presented a novel ML-based analysis approach for training and classification and
evaluated the model's effectiveness. The data is collected from three datasets, i.e., Smart bugs-
wild, SolidiFi-benchmark, and Smart bugs. The authors constructed an Abstract Syntax Tree
(AST) from Smart Contracts in the dataset by python third-package that are py-solc-x and
solidity-parse. ASTs are obtained from the experimental dataset(X) and malicious dataset(Y).
Acquiring the shared nodes between X and Y and these nodes are used to extract feature
vectors. ML algorithms like KNN and Stochastic Gradient Descent (SGD) are used to evaluate
the dataset. The results obtained from KNN performed are good in terms of the highest
Accuracy, Precision and Recall with above 90% for each vulnerability in the Smart Contract.
Re-entrancy Vulnerability -Accuracy with 95.45%, Recall with 95.45%, Precision with 95.8%.
Arithmetic Vulnerability-Accuracy 95.54%, Recall 90.90%, Precision 91.1%. Access Control
Vulnerability-Accuracy 95%, Recall 95%, Precision 96.67%. Bad Randomness Vulnerability-
Accuracy 90.1%, Recall 93.1%, Precision 91.1%. DoS Vulnerability-Accuracy 90.2%, Recall
90.26% Precision 90.56%. Unchecked low-level Vulnerability-Accuracy 90.1%, Recall 90.1%,
Precision 90.1%. Short Address Vulnerability-Accuracy 91.1%, Recall 91.1%, Precision
92.85%. Front Running Vulnerability-Accuracy 95.45%, Recall 95.45%, Precision 96.1%.
The authors proposed a deep-learning-based scheme called DeepCoin, used to detect network
attacks and fraudulent transactions in the Blockchain-based Energy network [97]. In this,
Blockchain is combined with DL using a truncated BackPropagation Through Time (BPTT) for
the smart grid. The authors got results based on the experiments using TensorFlow on three
different datasets, the web robot (Bot) - Internet of Things (IoT) dataset with an Accuracy of
98.2%, the Power System dataset with 96.52% and the CICIDS2017 dataset with 98.23%.
Most of the research studies analysis anomalies in ledger applications, cryptocurrencies and
specific types of attacks like DoS, Sybil or Double Spending etc. A machine learning-based
security technique using Blockchain network traffic statistics is attempted in [98].The presented
method consists of two phases, data collection and anomaly detection. Since the engine gathers
the network traffic features and creates the instances, features are extracted from the message
type. One_class SVM(Support Vector Machine) and AE(AutoEncoder) are the used semi-
supervised ML techniques to detect the anomaly. Features are compared with an estimated
threshold value. Feature having value beyond the threshold concluded as an anomaly. Mainly,
the method considers flow-based features. However, bitcoin was used as a prototype to
simulate the attack; the authors claimed that the presented technique could be used to detect
anomalies in any Blockchain application.To summarize the merits obtained through the
integration of Blockchain with AI techniques, the results obtained by their combinations are
portrayed in terms of quantitative metrics for readers perusal in Table 5.
Ref.
No.
Name of the Attack
Algorithms Used
Result
(Performance Metrics)
[59]
DDoS attack
PCA&MLP
50% attack detection
70% legitimate
[60]
DDoS Attack
DT,RF and KNN
RF
95.19 (Accuracy)
[64]
Malicious Attack
DELM
94.6%
(Accuracy)
[68]
Abnormal patterns
VAE,LSTM
99.80%
(Detection Rate)
[79]
Final attack
Multi-source deep learning
95%
(Precision)
[80]
DoS Attack
GSVM and LSVM
94.9% and 95.2%
(Accuracy)
[83]
Malicious Activity
K-Means,SVM,RF,MLP,Isolation Forest Model
Isolation Forest Model -
79.87% (Accuracy)
[91]
Selfish Mining,
51% attacks, DDoS,
Double Spending
OCSVM and k-means clustering
93%-for K-mean
90% for OCSVM
(Accuracy)
[94]
Mining pool attack
RF, DT, SVM, LR,KNN,Neural Networks
SVM-91%
(F-score)
[96]
Smart contract
vulnerabilities
KNN and SGD
KNN-95.45
(Accuracy)
[97]
DDoS,Ecllipse,Sybil and
Phishing
RNN using BPTT
98.23% (Accuracy)
TABLE 5: Tested Results of various Attack Defense Mechanism for Blockchain
Here in the Table 5, the results of various attack countermeasures are measured using accuracy
and F-Score metrics. When observing the results, RNN model works better in detecting DDoS,
Eclipse, Sybil and Phishing attacks [97] and obtains 98% as its Accuracy. The models in [59]
and [60] built upon ML techniques to detect the DDoS Attacks and the later one succeed well
with higher accuracy than that of the earlier one [59]. Detecting DoS attack is also very
important for complementing the Blockchain based applications. Techniques presented in [80]
produce good results in terms of accuracy. Most of the anomaly detection methods summarized
in Table 5 are all designed using customized ML techniques. Model presented in [68], [79] and
[97] use DL techniques for detecting the abnormal entities. Apart from accuracy and F-score
there are other metrics, to evaluate the anomaly detection techniques. Hence, we summarize the
methods which use the other metrics, Precision, Recall in Table 6.
Ref.No.
Technique
Recall
Precision
F1-score
Accuracy
[84]
GBAD USING BoT-IoT
97.24
98.01
97.59
99.9
[84]
GBAD using ToN-IoT
94.35
95.32
94.80
98.38
[85]
DBSCAN
99.69
94.33
96.937
94.15
[93]
DNN
99.34
97.84
98.57
97.72
[93]
K-MEANS+DT
99.5
99.5
99.5
99.4
[95]
XGBOOST
98
98
98
98
TABLE 6: Performance Metrics
While observing the results in Table 6, in terms of Accuracy metrics, the framework
experimented with BoT-IoT dataset in [84] produces 99.9%, which is really appreciable. In
terms of Recall, the method presented in [85] DBSCAN leading the other techniques. While
evaluating the techniques using Precision metrics, the technique envisioned in [93] leading the
other techniques. By observing the overall results against all the metrics, technique in [93] tried
well while detecting the anomalies than others. And the same data is represented graphically in
Figure 8.
FIGURE 8: Performance Metrics
A graphical visualization about DDoS attack detection using various mentioned in Table 5 is
shown in Figure 9. The observation reveals that the DL Technique presented in [97] works
better than that of the others. And the same kind of visualization for DoS attack is displayed in
Figure 10.
FIGURE 9: Accuracy of techniques while detecting the DDoS Attacks
FIGURE 10: Accuracy of Algorithms to avoid DoS Attacks
Apart from DDoS and DoS attacks, there are other attacks attempted and succeed in Blockchain
infrastructure. Figure 11 portrays the other types of cyber-attack detection through the anomaly
detection techniques performance in terms of accuracy metric.
FIGURE 11: Accuracy of Algorithms to avoid Different types of Attacks
Apart from the quantitative metrics used in Table 5 and Table 6 to quantify the efficacy of the
anomaly detection models built upon AI techniques, there is one more metric called running
time to measure the model’s performance. The running time of a model considers the amount of
time consumed by the model for completing a particular task. In the case of DL models, the
time complexity computation depends on the type of architecture CNN, RNN or LSTM etc, the
type of activation function, size of a model like number of dense layers, filters, pooling layers,
optimization algorithms intend for the candidate task, number of epochs and finally size of the
dataset. In terms of dataset, number of samples, or size of the data in terms of GB or TB,
number of output class labels. A demonstration on computing the time complexity associated
with the convolutional model can be found in [99]. Results reveal that the designed model
records the tradeoff between the model’s accuracy versus running time. Models with a greater
number of layer and filter compute the task quickly than that of models with minimal number
of layer and filter for the same size of dataset, but the obtained accuracy value is appreciable
while recognizing the running time.While analyzing the conventional ML techniques, apart
from the dataset parameters, models hyper-parameter contribution plays an crucial role in
determining the models running time. LSTM based anomaly detection model [68] built upon
power system and UNSW datasets accomplished the intended task in 73 seconds, the results
against the other metrics are also appreciable.
The models analyzed in Table 5 and Table 6 in perception of Anomaly Detection on
Blockchain platform is designed and validated using the offline as well as online dataset.
Gradient Boosting based Anomaly Detection model is built upon the BoT-IoT and ToN-IoT
datasets. While analyzing the size of the dataset used for training the AI based models in Table
5 and Table 6, the BoT-IoT and ToN-IoT size is bigger than that all the datasets.[84] Earlier
dataset consists of 73360900 attack and 9543 normal samples and the latter has 1498334 Attack
and 79053 normal samples. In terms of confusion matrix-based metrics, both the models’
results are appreciable for the detection of security breaches while ensuring the privacy aspects
as well. Hence, GBAD models would be the effective models in solving big datasets in real-
time deployment.
5.1 Challenges and future directions
Blockchain innovation is purely meant for enhancing the security aspects. Initially, Blockchain
technology was developed to support cryptocurrency called Bitcoin, used for transaction. Later
on, it finds its applications in other domains like education, the Internet of Things (IoT),
healthcare, pharmaceutical supply chains, government, agriculture, energy, food distribution
manufacturing and Industry 4.0 [100]. Blockchain still has serious technological issues, despite
its many benefits. When more users and transactions are taken into account, this enormous
technology confronts significant challenges. As more numbers of users get connected to this
network, in turn, transaction speed got decreased, which will not be appreciated in many use
cases. Due to an increase in the number of connected users, transaction processing times are
growing longer in Bitcoin and Ethereum networks. A 51% attack is attempted by an intruder
who wishes to controls the Blockchain mining ability. Through the 51% attack, miners can stop
some of the chosen transactions, to execute their intention. Double Spending attack is raised via
reversing the previous transaction. Most anticipated challenges ahead of Blockchain is
represented graphically in Figure 12.
FIGURE 12: Challenges of Blockchain
Another critical challenge ahead of Blockchain technology is through the Smart Contracts,
which enable the untrusted connected parties to agree upon the drafted condition to complete
the transaction. A Smart Contract facilitates the automation of network operations without the
intervention of a centralized, trusted third party. Despite the merits of smart contracts for
decentralized applications, attacker can observe the flaws that are hidden, which may
jeopardize the system. Hence, more focus must be given at the time of designing Smart
Contracts for every use case. The Smart Contract should be deterministic and must consider all
possible test cases. Challenges confronting the Smart Contract are presented in [101] while
enforcing ensuring privacy and legal issues. The author in [96] presents a ML-based Smart
Contract vulnerabilities detection system. The AST represents the statements in the Smart
Contract. The features to detect the vulnerabilities are derived through shared child nodes
created by the ASTs from the statements in the Smart Contract with and without vulnerabilities.
Finally, the derived feature vectors are processed, labeled and classified through ML models.
The authors reveal that the presented model focuses only on vulnerabilities of the Ethereum
platform. Such future study, however, might be taken into account as a substitute for other
kinds of blockchain systems. In addition, apart from detecting the security vulnerabilities, it
would be good to find the underlying cause, so that security breaches can be prevented
precisely in the near future.
Sustainability is an essential factor for ensuring a better future. Every sector needs to focus on
three forms of sustainability: environmental, economic and social. Blockchain technology acts
as a promising tool for increasing the sustainability of many use cases by monitoring messages
transmitted across the nodes. The authors in [102] implements technical sustainability for the
supply chain management. The considered use case demands huge amount of data for retrieving
the insights, but consuming more memory space is another big concern. Hence, supply chain
data were uploaded to the cloud to alleviate the issues in storage space requirements. To
demonstrate the presented technical sustainable Blockchain-enabled framework, the author
considers the maritime risk assessment analysis. The routing details are updated through smart
contracts, and Machine models are designed using Microsoft Azure to forecast the routing
information. A detailed analysis can be attempted on technical sustainability of supply chain
management along with suitable quantitative metrics. Technical sustainability by compromising
the environment and social sustainability is not a favorite solution. The upcoming researchers
can attempt to improvise the system by developing a customized ML model for the specified
task.
DDoS attack detection technique on the Bitcoin ecosystem is presented in [59]. The proposed
method considers the features obtained from the block and transaction details to detect the
attack. The technique may be improvised by considering the network flow features. In addition,
considering the characteristics of a block in the Blockchain, both in the presence and absence of
a DDoS attack, may complement the process. An attempt to detect and mitigate the DDoS
attack using the CICDDoS-2019 dataset presented [60]. The dominant features are selected
Tree-based classifier, the malicious packet is detected along with the IP address, and the same
has been stored in the Blockchain nodes through Smart Contracts, hence, tampering IP
addresses became impossible for the intruder. Thus, security is ensured. The presented
technique can be improvised by applying the other feature engineering concept. The
classification of IP addresses may also be attempted with DL techniques since existing work is
carried out through ML models.
Feature importance is computed through Random Forest; an attempt can be expected to use
other feature selection techniques to verify the preciseness in node characterization [94]. The
presented focuses only on the node addresses derived from Bitcoin powered by Blockchain
technology. Hence the presented proposal may be applied to non–cryptocurrency-based
Blockchains. Moreover, the technique which considers the features of node addresses that
participated in the transaction and consideration of network features is also a better option for
establishing the presented work. Recently, the finance applications have shown a keen interest
in distributed storage and implementing transparency. Still, the chances of attacks on
blockchain infrastructure are high especially through DoS attacks. Hence, authors in [103]
present various mitigation techniques meant to alleviate the negative impacts received through
DoS. This review article pinpoints the importance of detecting DoS attack incidents, as the
attack leaves the high negative impacts on Blockchain infrastructure. Hence, a robust
mechanism to detect them at early stage is essential. Moreover, the generation of a new set of
DoS vectors should be given the utmost priority task before train the ML model
6.Conclusion
Blockchain technology has conquered in all the sectors, due to its unique nature in storing data
without compromising the security aspects. However, Blockchain infrastructure security
aspects may be compromised through DDoS, Sybil, Eclipse, DAO, Majority, Byzantine attack
and so on. This review paper presents the key aspects of cyber security and merits of
Blockchain in complementing the security aspects. Various cyber-attack incidences against the
Blockchain infrastructure and possible countermeasures presented in various venues are
reported. Detecting security breaches at the earliest possible is mandate for the Blockchain
enabled applications, since the technology is extensively adapted in financial sector hence,
detecting anomalies is very important to prevent the financial loss. Then, we present a bench
marking frameworks or techniques comprises the Blockchain and Artificial Intelligence
techniques, especially in detecting the anomalies are presented to record the merits of these
two trending technologies. A brief discussion on various use cases built upon the Blockchain
and Artificial Intelligence techniques are summarized. To highlight the advantages of the
inclusion Artificial Intelligence techniques in improvising the anomaly detection, the tested
results are analyzed in terms of various quantitative metrics. Eventually, the open challenges
incurred in the integration of Artificial Intelligence techniques and Blockchain technology and
research gaps yet to be filled is also included for the benefit of potential researchers. We
believe that in the literature there is no such review article exists to report the anomaly
detection anticipated in Blockchain key infrastructure, and we hope this report will be very
useful for the research community.
References:
1. Von Solms, R., & Van Niekerk, J. (2013). From information security to cyber security.
computers & security, 38, 97-102.
2. https://www.upguard.com/blog/biggest-data-breaches
3. Raja, L., & Baboo, S. S. (2014). An overview of MANET: Applications, attacks and
challenges. International journal of computer science and mobile computing, 3(1), 408-
417.
4. Abosata, N., Al-Rubaye, S., Inalhan, G., & Emmanouilidis, C. (2021). Internet of things
for system integrity: a comprehensive survey on security, attacks and countermeasures
for industrial applications. Sensors, 21(11), 3654.
5. Sadiq, A., Anwar, M., Butt, R. A., Masud, F., Shahzad, M. K., Naseem, S., & Younas,
M. (2021). A review of phishing attacks and countermeasures for internet of
things‐based smart business applications in industry 4.0. Human Behavior and
Emerging Technologies, 3(5), 854-864.
6. Harefa, J., Prajena, G., Alexander, A. M., Dewa, E. V. S., & Yuliandry, S. (2021). Sea
waf: The prevention of sql injection attacks on web applications. Advances in Science,
Technology and Engineering Systems, 6, 405-411.
7. Abiona, O., Oluwaranti, A., Oluwatope, A., Bello, S., Onime, C., Sanni, M., & Kehinde,
L. (2014). Proxy Server Experiment and Network Security with Changing Nature of the
Web. International Journal of Communications, Network and System Sciences, 7(12),
519.
8. Abie, H. (2000). An overview of firewall technologies. Telektronikk, 96(3), 47-52.
9. https://cybersecurity.att.com/blogs/security-essentials/open-source-intrusion-detection-
tools-a-quick-overview
10. Othman, S. M., Alsohybe, N. T., Ba-Alwi, F. M., & Zahary, A. T. (2018). Survey on
intrusion detection system types. International Journal of Cyber-Security and Digital
Forensics, 7(4), 444-463
11. Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system. Decentralized
Business Review, 21260.
12. Sayeed, S., & Marco-Gisbert, H. (2020). Proof of adjourn (PoAj): A novel approach to
mitigate Blockchain attacks. Applied Sciences, 10(18), 6607.
13. Wani, S., Imthiyas, M., Almohamedh, H., Alhamed, K. M., Almotairi, S., & Gulzar, Y.
(2021). Distributed denial of service (DDoS) mitigation using Blockchain—A
comprehensive insight. Symmetry, 13(2), 227.
14. Alzahrani, A. G., Alhomoud, A., & Wills, G. (2022). A Framework of the Critical
Factors for Healthcare Providers in Data-Sharing Using Blockchain. IEEE Access.
15. Zou, L., Chen, J., Lan, Q., Zhou, Z., Ma, C., & Yang, Z. (2022, January). Application of
Blockchain Digital Identity Technology in Healthcare Consumer Finance System. In
2022 IEEE 2nd International Conference on Power, Electronics and Computer
Applications (ICPECA) (pp. 1212-1219). IEEE.
16. Sadri, S., Shahzad, A., & Zhang, K. (2021, February). Blockchain traceability in
healthcare: Blood donation supply chain. In 2021 23rd International Conference on
Advanced Communication Technology (ICACT) (pp. 119-126). IEEE.
17. Ramchandra, M. V., Kumar, K., Sarkar, A., Mukherjee, S. K., & Agarwal, K. (2021).
Assessment of the impact of Blockchain technology in the banking industry. Materials
Today: Proceedings.
18. Pokrovskaia, N. N., Rodionova, E. A., Fomina, I. G., Epshtein, M. Z., & Fedorov, D. A.
(2022, January). Blockchain and Smart Contracting in the Context of Digital
Transformation of Service. In 2022 Conference of Russian Young Researchers in
Electrical and Electronic Engineering (ElConRus) (pp. 1727-1731). IEEE.
19. Joseph, S., & Karunan, S. (2021, November). A Blockchain Based Decentralized
Transaction Settlement System in Banking Sector. In 2021 Fourth International
Conference on Microelectronics, Signals & Systems (ICMSS) (pp. 1-6). IEEE.
20. Pardeshi, K. (2021). An Overview of Blockchain in Supply Chain Management:
Benefits and Issues. IUP Journal of Supply Chain Management, 18(3).
21. Cordova, R. S., Maata, R. L. R., Epoc, F. J., & Alshar'e, M. (2021). Challenges and
Opportunities of Using Blockchain in Supply Chain Management. Global Business &
Management Research, 13(3).
22. Agrawal, D., Minocha, S., Namasudra, S., & Gandomi, A. H. (2022). A robust drug
recall supply chain management system using hyperledger Blockchain ecosystem.
Computers in biology and medicine, 140, 105100.
23. Siddiqui, S. T., Ahmad, M. O., Khamruddin, M., Gupta, A. K., & Singha, A. K. (2022,
January). Blockchain and IoT for Educational Certificates Generation and Verification.
In 2022 2nd International Conference on Computing and Information Technology
(ICCIT) (pp. 298-303). IEEE.
24. Alderazi, F. (2022, January). Security of Internet of Things: A review of Challenges
with integrating Blockchain with IoT. In 2022 2nd International Conference on
Computing and Information Technology (ICCIT) (pp. 154-160). IEEE.
25. Paturi, M., Puvvada, S., Ponnuru, B. S., Simhadri, M., Egala, B. S., & Pradhan, A. K.
(2021, December). Smart Solid Waste Management System Using Blockchain and IoT
for Smart Cities. In 2021 IEEE International Symposium on Smart Electronic Systems
(iSES)(Formerly iNiS) (pp. 456-459). IEEE
26. Zhang, R., Xue, R., & Liu, L. (2019). Security and privacy on Blockchain. ACM
Computing Surveys (CSUR), 52(3), 1-34.
27. https://www.ibm.com/in-en/topics/Blockchain-security
28. Mohanta, B. K., Satapathy, U., Panda, S. S., & Jena, D. (2019, December). A novel
approach to solve security and privacy issues for iot applications using Blockchain. In
2019 International Conference on Information Technology (ICIT) (pp. 394-399). IEEE.
29. Javaid, U., Siang, A. K., Aman, M. N., & Sikdar, B. (2018, June). Mitigating loT device
based DDoS attacks using Blockchain. In Proceedings of the 1st Workshop on
Cryptocurrencies and Blockchains for Distributed Systems (pp. 71-76).
30. Antwi, M., Adnane, A., Ahmad, F., Hussain, R., ur Rehman, M. H., & Kerrache, C. A.
(2021). The case of hyperledger fabric as a Blockchain solution for healthcare
applications. Blockchain: Research and Applications, 2(1), 100012.
31. Wen, Y., Lu, F., Liu, Y., & Huang, X. (2021). Attacks and countermeasures on
Blockchains: A survey from layering perspective. Computer Networks, 191, 107978.
32. Pouyanfar, S., Sadiq, S., Yan, Y., Tian, H., Tao, Y., Reyes, M. P., ... & Iyengar, S. S.
(2018). A survey on deep learning: Algorithms, techniques, and applications. ACM
Computing Surveys (CSUR), 51(5), 1-36.
33. Shafay, M., Ahmad, R. W., Salah, K., Yaqoob, I., Jayaraman, R., & Omar, M. (2021).
Blockchain for Deep Learning: Review and Open Challenges.
34. Chen, K., Yao, L., Zhang, D., Wang, X., Chang, X., & Nie, F. (2019). A semisupervised
recurrent convolutional attention model for human activity recognition. IEEE
transactions on neural networks and learning systems, 31(5), 1747-1756.
35. Zhang, D., Yao, L., Chen, K., Wang, S., Chang, X., & Liu, Y. (2019). Making sense of
spatio-temporal preserving representations for EEG-based human intention recognition.
IEEE transactions on cybernetics, 50(7), 3033-3044.
36. Luo, M., Chang, X., Nie, L., Yang, Y., Hauptmann, A. G., & Zheng, Q. (2017). An
adaptive semisupervised feature analysis for video semantic recognition. IEEE
transactions on cybernetics, 48(2), 648-660.
37. Zhang, Y., Liu, Y., & Chen, C. H. (2020, December). Survey on Blockchain and Deep
Learning. In 2020 IEEE 19th International Conference on Trust, Security and Privacy in
Computing and Communications (TrustCom) (pp. 1989-1994). IEEE.
38. Li, X., Jiang, P., Chen, T., Luo, X., & Wen, Q. (2020). A survey on the security of
Blockchain systems. Future Generation Computer Systems, 107, 841-853.
39. H. Mayer, Ecdsa security in bitcoin and ethereum: a research survey, 2016
URL http://blog.coinfabrik.com/wp-content/uploads/2016/06/ECDSA-Security-in-
Bitcoin -and-Ethereum-a-Research-Survey.pdf
40. Atzei, M. Bartoletti, T. Cimoli, A survey of attacks on ethereum smart contracts (sok),
in: International Conference on Principles of Security and Trust, 2017
41. Hassan, M. U., Rehmani, M. H., & Chen, J. (2021). Anomaly Detection in Blockchain
Networks: A Comprehensive Survey. arXiv preprint arXiv:2112.06089.
42. Tareq, A. H., & Sultana, M. (2018). Blockchain Attacks and A Model for Double
Spending Attack (Doctoral dissertation, Daffodil International University).
43. https://aithority.com/technology/Blockchain/Blockchain-hackers-stole-3-8-billion-in-
122-attacks-in-2020/
44. https://www.techtarget.com/searchsecurity/tip/Top-Blockchain-security-attacks-hacks-
and-issues
45. Saad, M., Thai, M. T., & Mohaisen, A. (2018, May). POSTER: deterring DDoS attacks
on Blockchain-based cryptocurrencies through mempool optimization. In Proceedings
of the 2018 on Asia Conference on Computer and Communications Security (pp. 809-
811).
46. Gong, S., & Lee, C. (2020). Blocis: Blockchain-based cyber threat intelligence sharing
framework for sybil-resistance. Electronics, 9(3), 521.
47. Luu, L., Velner, Y., Teutsch, J., & Saxena, P. (2017). Smartpool: Practical decentralized
pooled mining. In 26th {USENIX} Security Symposium ({USENIX} Security 17) (pp.
1409-1426).
48. Sayeed, S., & Marco-Gisbert, H. (2019). Assessing Blockchain consensus and security
mechanisms against the 51% attack. Applied Sciences, 9(9), 1788.
49. Bamert, T., Decker, C., Elsen, L., Wattenhofer, R., & Welten, S. (2013, September).
Have a snack, pay with Bitcoins. In IEEE P2P 2013 Proceedings (pp. 1-5). IEEE.
50. McCorry, P., Shahandashti, S. F., & Hao, F. (2016, February). Refund attacks on
Bitcoin’s payment protocol. In International Conference on Financial Cryptography and
Data Security (pp. 581-599). Springer, Berlin, Heidelberg.
51. Andrychowicz, M., Dziembowski, S., Malinowski, D., & Mazurek, Ł. (2013). How to
deal with malleability of bitcoin transactions. arXiv preprint arXiv:1312.3230.
52. Kendler, E. H. A., Zohar, A., & Goldberg, S. (2015). Eclipse Attacks on Bitcoin’s Peer-
to-Peer Network. Cryptology ePrint Archive.
53. Signorini, M., Pontecorvi, M., Kanoun, W., & Di Pietro, R. (2018, July). ADVISE:
Anomaly Detection tool for Blockchain Systems. In 2018 IEEE World Congress on
Services (SERVICES) (pp. 65-66). IEEE.
54. Bissias, G., Ozisik, A. P., Levine, B. N., & Liberatore, M. (2014, November). Sybil-
resistant mixing for bitcoin. In Proceedings of the 13th Workshop on Privacy in the
Electronic Society (pp. 149-158).
55. Gaži, P., Kiayias, A., & Russell, A. (2018, June). Stake-bleeding attacks on proof-of-
stake Blockchains. In 2018 Crypto Valley Conference on Blockchain Technology
(CVCBT) (pp. 85-92). IEEE.
56. Solat, S., & Potop-Butucaru, M. (2016). Zeroblock: Preventing selfish mining in bitcoin.
arXiv preprint arXiv:1605.02435.
57. Salah, K., Rehman, M. H. U., Nizamuddin, N., & Al-Fuqaha, A. (2019). Blockchain for
AI: Review and open research challenges. IEEE Access, 7, 10127-10149.
58. Chen, X., Ji, J., Luo, C., Liao, W., & Li, P. (2018, December). When machine learning
meets Blockchain: A decentralized, privacy-preserving and secure design. In 2018 IEEE
international conference on big data (big data) (pp. 1178-1187). IEEE.
59. Baek, U. J., Ji, S. H., Park, J. T., Lee, M. S., Park, J. S., & Kim, M. S. (2019,
September). DDoS attack detection on bitcoin ecosystem using deep-learning. In 2019
20th Asia-Pacific Network Operations and Management Symposium (APNOMS) (pp.
1-4). IEEE.
60. Manikumar, D. V. V. S., & Maheswari, B. U. (2020, July). Blockchain based DDoS
mitigation using machine learning techniques. In 2020 Second International Conference
on Inventive Research in Computing Applications (ICIRCA) (pp. 794-800). IEEE.
61. Abou El Houda, Z., Hafid, A., & Khoukhi, L. (2020, June). BrainChain-A Machine
learning Approach for protecting Blockchain applications using SDN. In ICC 2020-
2020 IEEE International Conference on Communications (ICC) (pp. 1-6). IEEE.
62. Gaba, S., Budhiraja, I., Makkar, A., & Garg, D. (2022, May). Machine Learning for
Detecting Security Attacks on Blockchain using Software Defined Networking. In 2022
IEEE International Conference on Communications Workshops (ICC Workshops) (pp.
260-264). IEEE
63. Dey, S. (2018, September). Securing majority-attack in Blockchain using machine
learning and algorithmic game theory: A proof of work. In 2018 10th computer science
and electronic engineering (CEEC) (pp. 7-10). IEEE.
64. Khan, M. A., Abbas, S., Rehman, A., Saeed, Y., Zeb, A., Uddin, M. I., ... & Ali, A.
(2020). A machine learning approach for Blockchain-based smart home networks
security. IEEE Network, 35(3), 223-229.
65. Chen, X., Ji, J., Luo, C., Liao, W., & Li, P. (2018, December). When machine learning
meets Blockchain: A decentralized, privacy-preserving and secure design. In 2018 IEEE
international conference on big data (big data) (pp. 1178-1187). IEEE.
66. Tsai, C. W., Chen, Y. P., Tang, T. C., & Luo, Y. C. (2021). An efficient parallel
machine learning-based Blockchain framework. ICT Express, 7(3), 300-307.
67. Dai, Q. Y., Zhang, B., & Dong, S. Q. (2022). A DDoS-Attack Detection Method
Oriented to the Blockchain Network Layer. Security and Communication Networks,
2022.
68. Keshk, M., Turnbull, B., Moustafa, N., Vatsalan, D., & Choo, K. K. R. (2019). A
privacy-preserving-framework-based Blockchain and deep learning for protecting smart
power networks. IEEE Transactions on Industrial Informatics, 16(8), 5110-5118.
69. Marwala, T., & Xing, B. (2018). Blockchain and artificial intelligence. arXiv preprint
arXiv:1802.04451.
70. Lu, Y., Huang, X., Dai, Y., Maharjan, S., & Zhang, Y. (2019). Blockchain and
federated learning for privacy-preserved data sharing in industrial IoT. IEEE
Transactions on Industrial Informatics, 16(6), 4177-4186.
71. Kim, H., Park, J., Bennis, M., & Kim, S. L. (2019). Blockchained on-device federated
learning. IEEE Communications Letters, 24(6), 1279-1283.
72. Ma, C., Li, J., Ding, M., Shi, L., Wang, T., Han, Z., & Poor, H. V. (2020). When
federated learning meets Blockchain: A new distributed learning paradigm. arXiv
preprint arXiv:2009.09338.
73. Liu, S., & Shang, Y. (2022, June). Federated Learning with Anomaly Client Detection
and Decentralized Parameter Aggregation. In 2022 52nd Annual IEEE/IFIP
International Conference on Dependable Systems and Networks Workshops (DSN-W)
(pp. 37-43). IEEE.
74. Islam, A., Morol, M. K., & Shin, S. Y. (2022, March). A Federated Learning-Based
Blockchain-Assisted Anomaly Detection Scheme to Prevent Road Accidents in Internet
of Vehicles. In Proceedings of the 2nd International Conference on Computing
Advancements (pp. 516-521).
75. Preuveneers, D., Rimmer, V., Tsingenopoulos, I., Spooren, J., Joosen, W., & Ilie-Zudor,
E. (2018). Chained anomaly detection models for federated learning: An intrusion
detection case study. Applied Sciences, 8(12), 2663.
76. Wang, H., Zheng, D., Liu, F., & Xiao, N. (2022). BlockFAD: A Federated Learning
Based I/O Anomaly Detection Method for Blockchain Domain Name System. In
International Conference on Artificial Intelligence and Security (pp. 668-679). Springer,
Cham.
77. Bravo-Marquez, F., Reeves, S., & Ugarte, M. (2019, April). Proof-of-learning: a
Blockchain consensus mechanism based on machine learning competitions. In 2019
IEEE International Conference on Decentralized Applications and Infrastructures
(DAPPCON) (pp. 119-124). IEEE.
78. Livieris, I. E., Pintelas, E., Stavroyiannis, S., & Pintelas, P. (2020). Ensemble deep
learning models for forecasting cryptocurrency time-series. Algorithms, 13(5), 121.
79. Jadidi, Z., Dorri, A., Jurdak, R., & Fidge, C. (2020, December). Securing manufacturing
using Blockchain. In 2020 IEEE 19th International Conference on Trust, Security and
Privacy in Computing and Communications (TrustCom) (pp. 1920-1925). IEEE.
80. Cheema, M. A., Ashraf, N., Aftab, A., Qureshi, H. K., Kazim, M., & Azar, A. T. (2020,
November). Machine Learning with Blockchain for Secure E-voting System. In 2020
First International Conference of Smart Systems and Emerging Technologies
(SMARTTECH) (pp. 177-182). IEEE.
81. Singh, S. K., Azzaoui, A. E., Kim, T. W., Pan, Y., & Park, J. H. (2021).
DeepBlockScheme: A deep learning-based Blockchain driven scheme for secure smart
city. Hum.-Centric Comput. Inf. Sci, 11, 12.
82. Jamil, F., Iqbal, N., Ahmad, S., & Kim, D. (2021). Peer-to-peer energy trading
mechanism based on Blockchain and machine learning for sustainable electrical power
supply in smart grid. IEEE Access, 9, 39193-39217.
83. Maskey, S. R., Badsha, S., Sengupta, S., & Khalil, I. (2020, March). Bits: Blockchain
based intelligent transportation system with outlier detection for smart city. In 2020
IEEE International Conference on Pervasive Computing and Communications
Workshops (PerCom Workshops) (pp. 1-6). IEEE
84. Kumar, P., Kumar, R., Srivastava, G., Gupta, G. P., Tripathi, R., Gadekallu, T. R., &
Xiong, N. N. (2021). PPSF: a privacy-preserving and secure framework using
Blockchain-based machine-learning for IoT-driven smart cities. IEEE Transactions on
Network Science and Engineering, 8(3), 2326-2341
85. Iyer, S., Thakur, S., Dixit, M., Katkam, R., Agrawal, A., & Kazi, F. (2019, July).
Blockchain and anomaly detection based monitoring system for enforcing wastewater
reuse. In 2019 10th International Conference on Computing, Communication and
Networking Technologies (ICCCNT) (pp. 1-7). IEEE.
86. Jain, S., Anand, A., Gupta, A., Awasthi, K., Gujrati, S., & Channegowda, J. (2020,
February). Blockchain and Machine Learning in Health Care and Management. In 2020
International Conference on Mainstreaming Block Chain Implementation (ICOMBI)
(pp. 1-5). IEEE.
87. Neelakandan, S., Rene Beulah, J., Prathiba, L., Murthy, G. L. N., Irudaya Raj, E. F., &
Arulkumar, N. (2022). Blockchain with deep learning-enabled secure healthcare data
transmission and diagnostic model. International Journal of Modeling, Simulation, and
Scientific Computing, 2241006.
88. Wu, Y., Jin, X., Yang, H., Tu, L., Ye, Y., & Li, S. (2022). Blockchain-Based Internet of
Things: Machine Learning Tea Sensing Trusted Traceability System. Journal of
Sensors, 2022.
89. Scicchitano, F., Liguori, A., Guarascio, M., Ritacco, E., & Manco, G. (2020). A Deep
Learning Approach for Detecting Security Attacks on Blockchain. In ITASEC (pp. 212-
222).
90. https://www.kaggle.com/bigquery/crypto-ethereum-classic
91. Sayadi, S., Rejeb, S. B., & Choukair, Z. (2019, June). Anomaly detection model over
Blockchain electronic transactions. In 2019 15th International Wireless
Communications & Mobile Computing Conference (IWCMC) (pp. 895-900). IEEE.
92. Pham, T., & Lee, S. (2016). Anomaly detection in bitcoin network using unsupervised
learning methods. arXiv preprint arXiv:1611.03941.
93. Rabieinejad, E., Yazdinejad, A., & Parizi, R. M. (2021, October). A deep learning
model for threat hunting in ethereum Blockchain. In 2021 IEEE 20th International
Conference on Trust, Security and Privacy in Computing and Communications
(TrustCom) (pp. 1185-1190). IEEE.
94. Michalski, R., Dziubałtowska, D., & Macek, P. (2020). Revealing the character of
nodes in a Blockchain with supervised learning. Ieee Access, 8, 109639-109647.
95. Bhamidipati, N. R., Vakkavanthula, V., Stafford, G., Dahir, M., Neupane, R., Bonnah,
E., ... & Calyam, P. (2021, December). ClaimChain: Secure Blockchain Platform for
Handling Insurance Claims Processing. In 2021 IEEE International Conference on
Blockchain (Blockchain) (pp. 55-64). IEEE.
96. Xu, Y., Hu, G., You, L., & Cao, C. (2021). A Novel Machine Learning-Based Analysis
Model for Smart Contract Vulnerability. Security and Communication Networks, 2021.
97. Ferrag, M. A., & Maglaras, L. (2019). DeepCoin: A novel deep learning and
Blockchain-based energy exchange framework for smart grids. IEEE Transactions on
Engineering Management, 67(4), 1285-1297.
98. Kim, J., Nakashima, M., Fan, W., Wuthier, S., Zhou, X., Kim, I., & Chang, S. Y.
(2022). A Machine Learning Approach to Anomaly Detection based on Traffic
Monitoring for Secure Blockchain Networking. IEEE Transactions on Network and
Service Management.
99. Shah, B., & Bhavsar, H. (2022). Time Complexity in Deep Learning Models. Procedia
Computer Science, 215, 202-210.
100. Ramaiah, M., Chithanuru, V., Padma, A., & Ravi, V. (2023). A Review of
Security Vulnerabilities in Industry 4.0 Application and the Possible Solutions Using
Blockchain. cyber security Applications for Industry 4.0, 63-95.
101. Khan, S. N., Loukil, F., Ghedira-Guegan, C., Benkhelifa, E., & Bani-Hani, A.
(2021). Blockchain smart contracts: Applications, challenges, and future trends. Peer-to-
peer Networking and Applications, 14(5), 2901-2925.
102. Wong, S., Yeung, J. K. W., Lau, Y. Y., & So, J. (2021). Technical sustainability
of cloud-based Blockchain integrated with machine learning for supply chain
management. Sustainability, 13(15), 8270.
103. Chaganti, R., Boppana, R. V., Ravi, V., Munir, K., Almutairi, M., Rustam, F., ...
& Ashraf, I. (2022). A Comprehensive Review of Denial of Service Attacks in
Blockchain Ecosystem and Open Challenges. IEEE Access.
