DEEP LEARNING MEETS BLOCKCHAIN FOR AUTOMATED AND SECURE ACCESS CONTROL

International Journal of Security, Privacy and Trust Management (IJSPTM) Vol 12, No 3/4, November 2023
1
DEEP LEARNING MEETS BLOCKCHAIN FOR
AUTOMATED AND SECURE ACCESS CONTROL
Asma Jodeiri Akbarfam1
, Sina Barazandeh2
, Deepti Gupta3
, and Hoda Maleki1
1
School of Computer and Cyber Sciences, Augusta University, Augusta, GA, USA
2
Department of Computer Engineering, Bilkent University, Ankara, Turkey
3
Subhani Department of Computer Information Systems, Texas A&M University-
Central Texas, TX, USA
ABSTRACT
Access control is a critical component of computer security, governing access to system resources. How-
ever, designing policies and roles in traditional access control can be challenging and difficult to maintain
in dynamic and complex systems, which is particularly problematic for organizations with numerous
resources. Furthermore, traditional methods suffer from issues such as third-party involvement,
inefficiency, and privacy gaps, making transparent and dynamic access control an ongoing research
problem. Moreover detecting malicious activities and identifying userswho are not behaving appropriately
can present notable difficulties. To address these challenges, we propose DLACB,a Deep Learning Based
Access Control Using Blockchain, as a solution to decentralized access control. DLACB usesblockchain to
provide transparency, traceability, and reliability in various domains such as medicine, finance, and gov-
ernment while taking advantage of deep learning to not rely on predefined policies and eventually
automate access control. With the integration of blockchain and deep learning for access control, DLACB
can provide a general frame-work applicable to various domains, enabling transparent and reliable logging
of all transactions. As all data is recorded on the blockchain, we have the capability to identify malicious
activities. To expedite this process, we store a list of malicious activities in the storage system and employ a
verification algorithm to cross-reference it with the blockchain. Weconduct measurements and comparisons
of the smart contract processing time for the deployed access control system incontrast to traditional access
control methods, determining the time overhead involved. The processing time of DLBAC demonstrates
remarkable stability when exposed to increased request volumes. This approach is particularly usefulfor
organizations seeking to automate access control while simultaneously detecting and preventing data
breaches to enhance security.
KEYWORDS
Blockchain, Deep learning, Access control, Authentication, Security
1. INTRODUCTION
Blockchain is a peer-to-peer network [40] and a method of storing information. It prevents un-
wanted modification and manipulation of data while providing a decentralized ledger for transac-
tions. It is also used for authenticating, maintaining, and synchronizing the content of a transaction
ledger replicated across multiple users. Blockchain facilitates decentralized transactions and data
management and offers an environment where no third party controls the data, and no trust is re-
quired between stakeholders [2]. The features of blockchain make it tamper-resistant, providing
integrity and a transparent view of what is happening to the data [13,59]. Therefore, blockchain is
useful in various domains for data management such as Internet of Vehicles (IoV), energy systems,
and supply chain management [55] or for ensuring security and privacy [2,39] such as medical sys-
tems [23], internet of things [31], finance [21] and access control [51].

2
In the context of access control, blockchain technology can be used to record and store access
control policies and access requests [18]. Each subject can have their own blockchain-based iden-
tity that contains their attributes and access rights. When an access request is made, the policy is
evaluated against the access context, and the decision is recorded on the blockchain. This ensures
transparency and audibility of access decisions. Blockchain technology enables secure and decen-
tralized transfer of access rights. Instead of relying on a central authority to transfer access rights,
blockchain allows for the peer-to-peer transfer of access rights through smart contracts. Decen-
tralization of access control eliminates the need for intermediaries and reduces the risk of fraud
[20].
Traditional approaches to access control, such as Access Control Lists (ACLs) [10], Role- Based
Access Control (RBAC) [16] , and Attribute Based Access Control (ABAC) [22] have beenutilized
in blockchains [18,36,44]. However, these types of access control have limitations such as over-
provisioning and under-provisioning of access, complexity in policy and attribute engineer-ing,
poor generalization to new situations, and utilizing a large amount of memory in the system[41].
Therefore, there is an increasing demand for innovative access control approaches, includingthe
utilization of deep learning-based models. Although few studies [33,42,15] have proposed deep
learning-based access control methods, they have not yet been integrated with blockchain technol-
ogy. These approaches can automate and dynamically manage access control decisions. However,
they may face challenges such as a central point of failure or lack of transparency, making them
unsuitable for organizations to detect and prevent data breaches and insider attacks. To overcome
these challenges, there is a need for new approaches that provide a secure and transparent access
control mechanism.
We propose Deep Learning Based Access Control Using Blockchain (DLACB), which com-bines
the benefits of blockchain and deep learning to provide a secure and transparent access control
mechanism. DLACB relies on smart contracts to verify users and integrates a deep learning
model for precisely assessing each user’s access level within the system. The utilization of deep
learning automates access control and eliminates the need for policy development. Addition-ally,
the framework employs smart contracts to consider organizational priority rules, detect ma-
licious users, and enforce bans when necessary. The framework is based on a private blockchain,
which only allows trusted nodes to validate transactions. This ensures the security and integrity of
the access control mechanism by preventing unauthorized access to the blockchain while not re-
quiring third-party authentication service. In addition, Each user has their own blockchain-based
identity, which ensures that only authorized users can access the system resources. DLACB al-
lows users to make access requests by generating new transactions, which are then logged on the
blockchain. This enables access transparency and helps detect malicious user activities in retriev-
ing resources. Moreover, we have incorporated a simple method to lock out malicious users after
their initial attempt and maintain an easily retrievable log of malicious activities stored securely
in the system. To further ensure the accuracy of this log, we employ a verification method that
leverages blockchain technology, offering fast retrieval of malicious activities while preserving
data integrity.
The integration of blockchain and deep learning in DLACB creates a resilient and secure access
control framework. Utilizing blockchain technology ensures transparency, authentication,
immutability, detection of malicious users, and secure communication with users and storage.
Simultaneously, the incorporation of deep learning enhances access control by enabling automated
decision-making, adapting to dynamic environments, improving accuracy, generalizing to new
scenarios, and minimizing human error.
The main contributions of this paper are as follows.
– We identify a research gap in traditional access control models, which rely on predefined

3
policies and permissions.
– We propose DLACB, which is a framework that utilizes deep learning access control to deter-
mine a user’s permissions on a given resource while having the capability of adding organiza-
tional rules for access. The framework also authenticates the users and logs the access requestson
the blockchain to recognize malicious users.
– We integrate a methodology for storing malicious user activities, retrieving them, and cross-
referencing them with the blockchain for verification purposes.
– We present the blockchain communication process, such as consensus protocols and smart
contracts to develop this framework and its cryptographic techniques, including hash functions,
digital signatures, and encryption algorithms.
– We demonstrate the security analysis a perform measurements and comparisons of the smart
contract processing times between the deployed access control system and traditional access
control methods, aiming to ascertain the extent of time overhead incurred.
Our proposed framework can serve as a basis for the development of secure and efficient access
control systems in various domains, including finance, healthcare, and government. By leveraging
the power of Deep Learning and Blockchain technologies, our framework provides a robust and
transparent solution to the critical issue of access control. The rest of this paper is structured as
follows. Section 2 discusses an overview of blockchain and access control. Section 3 explores the
related work. Section 4 discusses the importance of the method, describes the framework andthe
communication between entities. Section 5 describes the details of the protocols, Section 6
analyzes the security of the platform, Section 7 illustrates how the system was implemented and
the details of the used deep learning model. Section 8 contains the evaluation. Section 9 delves
into the proposed approach, providing insights into the reasons behind certain decisions made and
future work. The paper concludes with section 10.
2. BACKGROUND
2.1. Blockchain
Blockchains are peer-to-peer networks and timestamped chains of blocks [3]as shown in Figure 1
maintained by participating nodes and transactions that are accumulated in blocks. Every block is
chained to the previous block and cryptographically linked by including the previous block’s hash
value [12,5]. Blockchain provides immutable data storage as new blocks which can only be ap-
pended to the end of the chain, meaning existing transactions cannot be updated or deleted. Thus,
transactions can be trusted without the assistance of any third parties. The immutable chain of his-
torical transactions guarantees non-repudiation of historical transactions. In order to prove identity and
authenticity and secure read and write access to the blockchain, cryptography primitives suchas
digital signatures, hash functions, and ciphers are used.
Fig. 1: Chain of Blocks
A vital aspect of a blockchain network is the consensus algorithm that ensures integrity and
security. The consensus is a procedure that enables all participants in a distributed computing

4
environment to reach an agreement on the ledger’s current data state and be able to trust peers
[54]. Furthermore, fraud and data tampering is prevented because data cannot be changed without
the permission of a majority of the stakeholders in a blockchain system. Blockchain ledgers allow
for sharing but not for modification.
There are various types of blockchains, including public and private [17,53,58]. A distributed
ledger system without constraints and permissions is known as a public blockchain, with examples
of Bitcoin and Ethereum. Anyone with internet access can join the network as an authorized node
and become a part of the blockchain. It is permitted for a node or user who is a part of the public
blockchain to view recent and old records, confirm transactions and engage in mining. The mining and
trading of cryptocurrencies is the most fundamental usage of public blockchains.
A restricted or private blockchain [58,9] that may only be used in a closed network is a private
blockchain. A private blockchain is typically utilized within businesses or organizations where
only a small group of people are allowed to participate in the blockchain network. The governing
organization controls the level of security, authorizations, permissions, and accessibility. In other
words, private blockchains are similar to public blockchains, but with restricted access [32]. Private
blockchains are mostly used for cases where centralized control of the network is needed, such as
asset ownership, digital identity, supply chain management, voting, etc.
Some blockchains, including Ethereum-based ones, consist of smart contracts, which are modular,
reusable, and automatically executed programs [38]. A smart contract responds to information sent
to it, receives and stores the values, and transfers the information and values to other participants
in the blockchain network. Once a smart contract is launched, the compiler generates a byte code
that is stored on the blockchain, and it is assigned a storage address. All network nodes execute
the opcodes produced by compiling the contract scripts when a condition specified in the contract
takes place. The execution results are then written to the blockchain by the network nodes once a
transaction is sent to the contract address.
2.2. Access Control
Access control is a crucial component of security that restricts access to resources on a system to
authorized users while preventing unauthorized access. By implementing restrictions and permis-
sions, access control reduces the risks of data breaches and other security threats to organizations
or businesses. Security compliance programs rely heavily on access control policies and technol-
ogy to protect sensitive information, such as customer data.
There are various traditional access control types, including attribute-based access control
(ABAC), role-based access control (RBAC), and rule-based access control. In ABAC, every user
has an attribute defined by the administrator, and their request can be accepted or denied based on
these attributes [26,14,27]. RBAC defines the level of access based on the user’s responsibilities
and departmental jobs [49]. Rule-based access control restricts access to different parts of a system
based on predefined rules for each area [24]. However, traditional access control techniques have
some drawbacks, including the provision of excessive or insufficient access, challenges in creating
policies and attributes, limited adaptability to new scenarios, and significant memory usage in the
system [42].Consequently, there is an increasing demand for innovative access control methods,
such as models based on deep learning, to overcome these limitations.

5
3. RELATED WORK
Blockchain technology has demonstrated its effectiveness in diverse sectors. In supply chain man-
agement, it enhances traceability, transparency, product authenticity, and combats counterfeiting,
facilitating decentralized information sharing [4]. In healthcare, ensuring the security and privacy
of electronic medical records remains a challenge [29]. blockchain ensures secure and efficient
medical information exchange, preserving data integrity and confidentiality while addressing data
management challenges [52]. For voting systems, blockchain guarantees secure, transparent, and
immutable transactions, boosting the integrity of the process [28]. Blockchain also has the poten-
tial to be utilized with insurance solutions against political risks [57]. In recent years, there has
been a significant increase in the adoption of blockchain technology in access control systems. This
adoption provides a decentralized and secure framework for implementing access control policies.
Within this section, we delve into various research endeavors that have put forward blockchain-
based access control solutions tailored to different purposes and with the utilization of various
types of access control.
RBAC on blockchain: In the context RBAC, several studies have explored its application within
blockchain technology[16,47]. Sun et al. [50] introduced an RBAC model using a private
blockchain. They utilized a multi-index table and a key-value database to formulate access control
policies and store summarized access control data on the blockchain. Additionally, they employeda
smart contract to handle user access queries and adopted the distributed proof of stake (DPOS)as
the consensus algorithm within the blockchain. This private blockchain was combined with a
Mandatory Access Control (MAC) policy. Kim et al. [30] utilized RBAC on a blockchain for sharing
filmed images in closed-circuit television systems for crime prevention. Similarly, Cruz et al.The
authors in [8] introduce an approach to secure access control in a private blockchain. It empha- sizes
the initial deployment of a RBAC, to effectively categorize users based on predefined roles,
restricting their access to specific parts of the blockchain. This is further fortified by the integration
of a Machine Learning (ML) model, enabling the system to detect intricate user behavior patterns
and potential security threats in real-time. [16] applied RBAC and defined a smart contract-based
authentication method suitable for trans-organizational utilization of roles. However, traditional
access control models like RBAC can lead to errors in policies and require significant memory
for storing access lists, limiting their ability to handle unforeseen access requests. RBAC-based
blockchain solutions may encounter problems related to policy engineering, memory usage, and
difficulties in handling novel access requests.
ABAC on blockchain: In the realm of ABAC, blockchain has also found applications [48]. Given
the swift evolution of the Internet of Things (IoT) [6], blockchain proves especially bene-ficial in
this sphere [56], along with data-sharing situations. Ding et al. [19] and Zaidi et al. [56] employed
blockchain for IoT access management using ABAC. Ding’s approach stored attributedistribution
on the blockchain, while Zaidi utilized smart contracts for fine-grained access controlin IoT data
sharing. This ABAC-based approach enables more flexible and granular access control. Abdi et al. [1]
presented a hierarchical blockchain architecture for ABAC in the IoT environment, aiming to
reduce network overhead and transaction latency. They introduced various types of managers and
employed practical Byzantine fault tolerance (PBFT) as the consensus algorithm. This approach,
while more robust, may require extensive attribute engineering. ABAC-focused blockchain
solutions may face challenges related to attribute engineering and policy definition, potentially
leading to time-consuming processes and policy errors.
Reinforcement Learning and Dynamic Access Control on blockchain: Outchakoucht et al.
[43] introduced a dynamic and decentralized security framework that leverages blockchain tech-
nology and reinforcement learning (RL) algorithms. This approach eliminates the need for cen-
tralized authorities in access control and continuously adapts and optimizes security measures.

6
However, it exhibits limitations concerning blockchain privacy and block validation duration.
Privacy-Preserving Healthcare Data Sharing: Blockchain has also been applied to secure
healthcare data sharing. Madine et al. [34] proposed a decentralized data management model in
which patients control their data, including the ability to remake re-encryption keys. Similarly, Al-
hajri et al. [7] introduced a human-centric decentralized consent system for fitness data sharing via
blockchain. Zou et al. [59] presented a medical data sharing and privacy-preserving eHealth system
(SPChain) using proxy re-encryption schemes to allow patients to share data while maintaining
privacy. Challenges in healthcare data sharing solutions include key management, deployability,
and authentication rules, as well as the risk of incorrect access decisions.
The DLACB framework stands out for its use of deep learning to enhance access control
decision-making. It automates permission grants and eliminates the need for predefined policies,
and policy engineering. Additionally, It helped with fast detection of malicious activity and adapt-
ability to limiting malicious users. Overall, the research landscape in blockchain-based access
control systems offers a spectrum of approaches, each with its own unique set of advantages and
challenges.
4. DLACB FRAMEWORK
In organizations, outsourcing access control functions to third parties has become popular. This
eliminates the need to configure and maintain complex systems that could incur high acquisition
and operating costs [35]. These systems require log production that usually reveals the system’s
regular events and assists in detecting malicious behavior or attacks. Analyzing these logs helps
to catch security breaches or incident discovery and subsequent damage control [46]. Therefore,
these logs need to be immutable to prevent attackers from altering them and causing analysis to be
incorrect.
This paper proposes the DLACB; a novel approach for implementing access control services by
utilizing blockchain technology. Our framework creates immutable system logs and eliminatesthe
need for trusting a third party while carrying the mentioned advantages. As part of our approach,
an access control service is built on top of a private blockchain. It leverages blockchain
technology for configuring access control policies based on a decision engine. In other words, the
decision engine evaluates an access request for a resource This decision engine combines a deep
learning model with priority rules. It utilizes deep learning for access control policies, facilitating
informed decisions based on user and resource metadata across various operations. This approach
enables dynamic adaptability, improved accuracy, reduced human errors, and enhanced overall
efficiency, allowing it to seamlessly navigate evolving environments and effectively generalize to
new scenarios. By leveraging deep learning for access determination, the reliance on predefined,
rigid policies and rules is eliminated, ultimately enhancing the precision of access decisions. Addi-
tionally, although not mandatory, specific priority rules can be applied within the decision engine,
tailored to conditions within a given network or organization, ensuring the effective adaptation of
the access control policy to predefined circumstances.
4.1. Motivations and Objectives
DLACB offers several notable advantages over prior approaches:
1. Elimination of the need for third-party trust: Blockchain integration removes the necessity
for third-party reliance, allowing for decentralized and transparent access control decisions
verified through network consensus. The tamper-resistant and immutable ledger characteristic of

7
blockchain enables autonomous user authentication, access rights management, and a secure
environment without external dependencies.
2. Enhanced Security: DLACB’s utilization of blockchain technology bolsters security, provid-
ing robust protection for sensitive data such as access control policies and user permissions.
3. Streamlined Administrative Processes: DLACB simplifies access control procedures, liber-
ating administrators to allocate their resources to other areas.
4. Adaptability: DLACB’s dynamic access determination powered by deep learning enables
swift adaptation to new scenarios and emerging threats.
5. Prevention of Replay Attacks: DLACB incorporates preventive measures, including nonces,
to thwart replay attacks. This ensures that each request or transaction is processed only once,
reinforcing system security.
6. Malicious Activity Detection: The framework maintains comprehensive logs of user requests
and access decisions, facilitating the detection of malicious activities and unauthorized access
attempts. This transparency is pivotal for security monitoring and incident response.
7. Consistent Performance Under Heavy Loads: DLACB’s processing time remains remark-
ably stable even when subjected to high request volumes. This analysis underscores DLACB’s
suitability for larger enterprises, effectively addressing intricate access control requirements.
4.2. DLACB Architecture
The DLACB framework uses blockchain to manage access control policies and data retrieval; it
first authenticates the user and then provides the user with the appropriate level of access to the
requested resources. It also keeps logs of the requests and their results, which are immutable based
on the nature of the blockchain. As illustrated in Figure 2, our framework consists of four entities;
user, node, decision engine, and storage.
4.1.1. Blockchain: A private blockchain, responsible for access control, is employed, utilizing
trusted nodes and adopting the Proof-of-Authority consensus algorithm. This configuration
ensures that only authorized validator nodes have the exclusive authority to validate transactions
andgenerate blocks within the blockchain.
4.1.2. User: An entity, such as an employee, interested in storing, manipulating, or accessing
data and interacts with our application.
4.1.3. Storage: An entity responsible for storing the user’s data. The storage interacts with the
blockchain when a request for storing or retrieving data is made by a user.
Access Data
Fig. 2: DLACB Framework
Users
Blockchain Storage
Access Request Access Decision
Link
Link Utilization
Malicious user list and verification
Authentication Authorization
User Authentication
Binary representation
of the request
Decision regarding
access permissions
Using Deep Learning
and Rules
1001

8
4.3. Blockchain Structure
The DLACB framework relies on a private blockchain structure consisting of trusted validator
nodes and a crucial decision engine for access control.
Validator Node: Validator nodes are the backbone of the DLACB blockchain. These nodes are
entrusted entities that play a pivotal role in the validation of transactions and the creation of
blockchain blocks. To ensure utmost trustworthiness, the network mandates that each val-idator
node undergoes identity verification before participating as a trusted node. Within ourframework,
validator nodes are implemented as standalone, physically secured servers.
Decision Engine: The decision engine is the heart of DLACB’s access control mechanism. It
serves the critical function of user authentication and determining the access level granted to
specific users, utilizing advanced deep learning models. To achieve this, the blockchain
incorporates two essential smart contracts, securely stored and executed when the relevant
transactions are initiated.
Authentication Smart Contract: This smart contract is triggered upon the reception of a data
retrieval request. It rigorously validates the user’s public key and confirms their network
membership.
Authorization Smart Contract: Following successful authentication, the authorization smart contract
takes over. This contract assesses whether the user is allowed to access a particular resource. It
harnesses a deep learning model to generate an output containing a set of operations available
for the specified resource, based on the provided inputs. Subsequently, it conducts a comparison
between this output and priority rules stored within the smart contract’s storage, thereby
determining any specific event preferences of the organization.
Every node on the blockchain is equipped with a public and private key pair, ensuring secure
transaction encryption and digital signature capabilities. To facilitate access to resources, whena
user initiates a request, the blockchain undertakes the authentication and access determination
process through its decision engine. Subsequently, the decision engine’s result is transmitted to
the storage subsystem, enabling the generation of an access link. This link is then shared with the
user, granting them access to the requested resource.
4.4. DLACB Transactional Workflow for Data Retrieval
Within the DLACB framework, a range of transactions can be performed on the blockchain, as
detailed in Table 1. These transactions serve key roles in managing access and data interactions.
Consequently, the TSetuptransaction is used to add a new node to the system, effectively expanding
the blockchain’s capabilities. Users employ the TAccReqtransaction to request access to specific re-
sources, specifying their data or operational requirements. On the other hand, the TLinktransaction
serves the purpose of providing users with an access link, granting them access to their requested
resources. When a user accesses a resource, the TStoragetransaction comes into play. It is utilized to
log the user’s interaction with the resource, ensuring the maintenance of an immutable record of
data access. Finally, the TVerifiedtransaction is created as an input to the smart contract within the
decision engine and plays a crucial role in the authentication and authorization processes.
•
•

9
Table 1: Transactions
Transaction Parameters
TSetup pkA, pkU , Time, SignskA
TAccReq (pkU , Time, ReqInfo), SignU
TLink EncpkU (Accesslink, Nonce, Time), SignS
TStorage (Nonce, Time, pkU ), SignS
TVerified (Time, pkUbinary , ReqInfobinary)
The system process for utilizing the transactions for data retrieval, as depicted in Figure 3,
unfolds as follows: The user initiates an access transaction within the blockchain, supplying their
public key, request details, and freshness indicator. Subsequently, the blockchain proceeds to au-
thenticate the user by harnessing the decision engine to determine the user’s access permissions.
At the core of the blockchain lies the decision engine, a fundamental component composed of two
smart contracts securely stored within the blockchain. These contracts execute in responseto
specific transactions. When an access request is received, it triggers the authentication smart
contract, which verifies the user’s public key and confirms their network membership. Upon suc-
cessful authentication, this contract generates a verification (TVerified) transaction, setting in motion
the authorization smart contract. The authorization smart contract evaluates whether the user is au-
thorized to access the requested resource. More precisely, it employs a sophisticated deep-learning
model explained in Section 7.2 to generate an output that enumerates permissible operations for
the specified resource based on the provided inputs. This output undergoes comparison with prior-
ity rules stored within the contract’s storage, determining if any specific organizational preferences
exist for the event.
The result of this process is then transmitted to the storage subsystem. In response, the storage
generates a TLinktransaction containing an access link, timestamp, and nonce, all encrypted with
the user’s public key. This information is relayed back to the blockchain, which, in turn, forwardsit
to the user. Upon receiving, the users utilize the provided access link and nonce. To maintaina
comprehensive record of the event, the storage subsystem generates a TStoragetransaction within the
blockchain, effectively logging the access event.
Fig. 3: Requesting Data from Storage by User

10
4.5. Malicious User Detection
At each stage of the process, the data is also sent to storage for the purpose of maintaining records.
This additional step plays a crucial role in automating the identification of suspicious users within
the system, thereby eliminating the need for manual blockchain queries to spot potentially harmful
actions. Here’s how this process operates:
1. When a user lacks proper authentication, possibly indicating their initial interaction with the
system, the storage logs the user’s request along with related information as suspicious activity.
2. If an authenticated user tries to access a file without the necessary permissions or attempts to
exceed their authorized access limits, the storage marks this behavior as suspicious.
3. In cases where an authenticated user tries to access an incorrect file, the system records this as
suspicious activity.
4. If an authenticated user repeatedly makes unsuccessful attempts to access an incorrect file,
the storage system triggers an action by sending a request to the authorization smart contracton
the blockchain. This request prompts the contract to adjust its predefined rules, ultimately
revoking the user’s access to all files within the system.
4.6. Enhancing Detection of Malicious Activity Queries
To improve the extraction of records related to malicious activities, two methods are employed.
The first method involves recording all transactions on the blockchain, including additional infor-
mation such as the user’s public key and request details. This facilitates the extraction of transac-
tions specifically associated with the targeted user or file. However, this approach may allow some
activities to go unnoticed.
To address this limitation, a second method is utilized, which involves storing records in an off-
chain storage system, enabling faster malicious activity detection. However, when extracting
records from this storage, it is crucial to verify their integrity. To achieve this, the system utilizes
the distributed Merkle root, as proposed in [5].
Whenever a request is denied by an authentication smart contract or authorization smart con-tract,
the entire transaction is sent to storage along with its results and added to the list of malicious
activities. A Merkle root is then generated and stored in the smart contract. This Merkle root is
calculated by hashing the previous Merkle root (denoted as Mprev Malicious) with the transactions
added in the current block (represented as Tblock) and the output of the smart contract is
Smartcontract Output:
MMalicious = Hash(Mprev Malicious ∥ (Tblock ∥ Smartcontract Output))
By hashing the previous Merkle root with the transactions in the current block, a tamper- evident
structure is created, linking the records of malicious activities. Through these two methods—
storing provenance records on the blockchain and using distributed Merkle roots—the sys-tem
enhances its ability to extract the information of the malicious activity from the storage and verify
with the blockchain.
5. DLACBPROCESS AND CRYPTOGRAPHIC TECHNIQUES
In this section, we will introduce cryptographic notions and techniques, underlying protocols, the
details of the transactions, and smart contracts.

11
←
∈
←
←
←
←
5.1. Cryptographic Notions
We will use standard cryptographic building blocks, including asymmetric encryption, digital sig-
natures, and a cryptographic hash function. The encryption scheme consists of a tuple (Gen, Een,
Dec)– a generator, an encryption algorithm, and a decryption algorithm, respectively. The digital
signature scheme is defined by the tuple (Gens, Sig, V er)– a generator, a signature, and a verifi-
cation algorithm, respectively. We denote the cryptographic hash function as H and call it a hash
function. Finally, we represent the blockchain memory as M; a key-value structured database that
contains the complete information stored on the blockchain.
5.2. Smart Contract Protocols
Our approach contains an auxiliary function, Access Verification, and two smart contracts, Authen-
tication and Authorization.
Algorithm 1 extracts the keys, timestamps, and signatures from the transactions. By checkingthe
signatures and timestamps, the function establishes if the node or the user is the claimed entityand
if the transaction is fresh.
Algorithm 1 Access Verification Check
1: procedure ACCESSVERIFICATION(TAccReq, M)2: a 0
3: (pkU , sign, time) Decompose(T )
4: if H(pkc) M[key] and Current(time) and verify(time) then
5: a 1
6: end if
7: return a
8: end procedure
Algorithm 2 is the first smart contract performed in the decision engine. The smart contract for
authentication is executed because of TAccReq. It extracts the user key from the transaction anduses
AccessVerification() to verify the user and determine if they were previously added and hadthe
authority to access the system. After the user is verified, and the freshness of the request is
evaluated, the smart contract converts the user data and resources into a binary representation and
issues TVerified, which triggers the authorization smart contract.
Algorithm 2 Authentication
1: procedure VERIFICATION(TAccReq, M)
2: (pkU , time, ReqInfo, sign) Decompose(TAccReq )
3: if AccessV erificationCheck(TAccReq, M) then
4: pkUbinary ← binaryRepr(pkU ) ReqInfobinary ← binaryRepr(ReqInfo)5
6: TVerified (time, pkUbinary ,
ReqInfobinary)
7: return TVerified
8: end if
9: end procedure
After receiving TVerified, Algorithm 3 is executed on the decision engine to determine access
according to the deep learning model for the user, and then compares the output with the priority
rules and sends the final decision to the storage.
Algorithm 3 Authorization
1: procedure DECISIONENGINE(TVerified, M)

12
←
←
←
←
← ∥
2: (time, pkU , ReqInfo) Decompose(TVerified)
3: if Current(time) then
4: accessList DecisionEngine(pkU , ReqInfo)
5: end if
6: return accessList
7: end procedure
5.3. Denied Access and Malicious Activity List Verification
We have adapted the algorithm originally presented in [5] to accommodate verification alongside
the storage system. The verification procedure, as detailed in Algorithm 4, is utilized to safeguard
the integrity of the Malicious Activity list stored in the system. This algorithm constructs a Merkle
tree and subsequently checks the result and Merkle root against the stored Merkle root. This com-
parison enables us to be certain whether any alterations have been made to the data.
Algorithm 4 Enhanced Storage Verification
procedure STORAGEVERIFICATION(sorted malicious list)
MMalicious null
for block in sorted record do
Mblock ← null
Mblock ← BlockMerkleRoot(block record)
MMalicious MMalicious Mblock
end for
stored merkle root Retrieve stored Merkle root
if MMalicious matches stored merkle root then
Data integrity verified
else
Data integrity compromised
end if
end procedure
6. SECURITY ANALYSIS
In this section, we perform an analysis of DLACB’s features that contribute to secure data sharing
with users.
6.1. Applicable Scenarios
To evaluate the effectiveness and practicality of the proposed DLACB framework, various exper-
iments were conducted to test its performance in different scenarios. The results of these experi-
ments, as well as all the requests, were logged on the blockchain for record-keeping purposes. The
scenarios tested were designed to cover all possible cases and include the following:
1. In the first scenario, a user who is not registered on the blockchain attempts to access a re-
source. In this case, the access control mechanism should deny the request, as the user’s publickey
is not present on the blockchain, and the system cannot authenticate the user.
2. The second scenario involved a user who is registered on the blockchain but is denied access
by the deep learning model. The deep learning model in this scenario predicts that the user’s
request should be denied, and the system should respond accordingly.
3. The third scenario tested the case where the deep learning model approves the user’s request,
but the decision engine’s policies reject it. This denial is based on static rules and policies defined
by the organization’s administrators.
4. The final scenario tested is where both the decision engine and the policies agree on authoriz-

13
ing the user to access the resource. This scenario represents the ideal situation where the useris
authorized to access the resource.
To ensure that the testing process is comprehensive, the model was tested with all the above sce-
narios, and the framework was able to handle each test case successfully. For testing purposes, a
dataset containing 100 users and their corresponding transactions was registered on the blockchain.
However, the deep learning model recognized the entire dataset, including all users. The test cases
for the scenarios were selected carefully to satisfy the description of each case and provide an
accurate representation of the system’s capabilities.
Moreover, the lightweight model was tested both locally and remotely, assigning a specific range
of ports for users and administrators. This testing process was conducted to verify the sys-tem’s
functionality and performance in various settings, including both local and remote environ-ments.
6.2. Security Features
DLACB incorporates several robust security features to ensure the protection of sensitive data and
maintain the integrity of its operations. These features work in tandem to safeguard the framework
from various threats and vulnerabilities.
Confidentiality and integrity: DLACB’s approach to ensuring confidentiality and integrity is a critical
feature that helps prevent unauthorized access to sensitive data. In the DLACB blockchain,every
entity has a pair of public and private keys that are used for secure communication. When
requests are exchanged between two framework entities, they are encrypted with the receiver’s
public key, ensuring that only the intended recipient can decrypt and access the content. For exam-
ple, when the storage sends a TLink, it is encrypted with the user’s public key, making it unreadable
to anyone who does not possess the user’s private key. Similarly, the Request Result is encrypted
with the storage’s public key, ensuring that only the storage can read the content. Furthermore,
DLACB also ensures integrity by requiring that the entities sending a request sign the content with
their private key. This signature acts as proof that the request originated from the entity and was
not modified in any way. This feature helps prevent tampering with the data and ensures that all
data transmitted across the network is authentic.
By combining confidentiality and integrity features, DLACB offers a robust solution for secure data
sharing. The encryption of data using the recipient’s public key ensures that only authorized
parties can access the content, while the signature ensures that the data’s integrity is maintained
and it has not been tampered with.
Preventing reply attack: Preventing replay attacks is crucial to maintaining the security of the
system. Without proper measures, an attacker could capture a message and then repeatedly send it
to the receiver, causing unwanted actions to be taken. One of the primary methods used toprevent
replay attacks is the use of a nonce. In DLACB, a nonce is generated for each request andproduced
link. A nonce is a random token that is unique for each request and is used to ensure that a
malicious user cannot reuse previously communicated links. The nonce is included in the request
and transaction messages, and the recipient verifies that the nonce in the message matches the
nonce that they expect to receive. By including a nonce in each message, DLACB ensures that
each request or transaction can only be processed once and cannot be reused by an attacker. By
preventing replay attacks, DLACB maintains the integrity of the system and ensures that each
request or transaction is processed only once. This makes it more difficult for an attacker to disrupt
the system or gain unauthorized access to the data. The use of nonces is a common technique in
blockchain systems to prevent replay attacks, and DLACB implements it effectively.
Preventing man in the middle attack: There are different scenarios in that a man-in-the-middle attack

14
is possible but prevented by encryption of the data, including the nonce and the signature ofthe
message inside the request or the transaction.
Detecting malicious activity: Another important feature of DLACB is its ability to detect malicious
activity. The framework maintains a log of all user requests and the results of those requests,
including which user requested the data and whether they were granted access. By logging this
information on the blockchain, the system can detect any unauthorized requests or attempts to
modify the data. In addition to the mentioned features, the DLACB uses PoA as a consensus al-
gorithm, in contrast to Proof-of-Work consensus, where an attacker needs to control 51% of the
network’s computing power; in PoA consensus, it is necessary to control 51% of the network
nodes. It is more difficult to control the nodes in a permissioned network than it is to acquire com-
putational power, leading the attack to be more challenging in PoA [37]. PoA is scalable, efficient,
and easy to implement; it can be scaled so that a blockchain using our model can have v validators
with v > 2. Validators, the trusted nodes, are the only nodes responsible for reaching consensus,
while the users are free to submit transactions.
7. IMPLEMENTATION
This section discusses the implementation details of the introduced DLACB framework, including
the blockchain and the Deep-Learning-based decision engine, and how these components work
together as a system. In the end, we provide details on the experimental results and discuss the
practicality of the framework from a technical perspective.
7.1. Blockchain Implementation
DLACB is proposed primarily for organizations that use private networks that require particular
transactions, authorization procedures, and smart contracts, as discussed in section 4.2. There-
fore, the blockchain is implemented from scratch, not inherited from any off-the-shelf blockchain
technology, and can be considered a baseline for future frameworks using Deep Learning for the
Access Control decision engine. To implement the framework as a baseline, as mentioned and as
proof of concept, we use Python3 as the backend language. The framework’s functionality is im-
plemented as an API using the Flask library. In addition, a simple interface is implemented using
plain HTML styled with CSS, and JavaScript is used to connect the interface and the backend by
sending and receiving the HTTP requests. The implementation used for the test was an interactive
web-based implementation that required the user/tester to perform each task by interacting with
the web page. In the backend of this implementation, different ports are used for trusted nodes and
users, and the blockchain runs on a defined port necessary for communication through the API.
The user’s public and private keys are generated as pairs, and their uniqueness is ensured. Here, we
use the widely used SHA-256 cryptographic hash function [45] as our encryption method forthe
entity and verification of digital signatures. The transactions are implemented as mentioned in
section 4.2, and the transactions are then extracted and distributed among the nodes for consensus.
The keys and the data are stored locally by serializing and restored by de-serializing. This method
is only suitable for the demo version. However, the deployed version of the framework requires
the use of a DBMS database management system such as MySQL.
7.2. Decision Engine Implementation
In this work, we use a machine learning model as our decision engine to allow or reject access toa
resource upon a user’s request. For this purpose, we use DLBACα [41], a deep learning model
inspired by the widely used [25] model, a convolutional Deep Neural Network successfully used
in various domains. In [41], the authors used a variety of datasets, including real-world and syn-

15
thetic datasets. The real-world datasets are the Amazon-Kaggle and Amazon-UCI datasets, which
contain historical access data and access information for a large number of users and resources,
respectively. Both real-world datasets are imbalanced, with the majority of tuples having grant
access. The Amazon-UCI dataset is also not ABAC in nature, meaning that there are some tuples
in the dataset where users with identical attribute values do not have the same access permissions.
The synthetic datasets were generated with varying numbers of users, resources, user and resource
metadata, and authorization tuples, each reflecting a varying degree of complexity. The synthetic
datasets were designed to be more challenging than the real-world datasets, with more overlapped
samples and higher complexity.
The model is implemented using popular Tensorflow and Keras machine learning libraries in
Python 3.9. We use a pre-trained version of the model publicly available on GitHub. The model
receives the user identification, which is an id, resource id, and returns four values for all the
possible operations the user might have requested for the specific resource. The decision is made
based on the requested operation. Nevertheless, the model is not limited to the use of userid and
resourceid and can function if trained on any attribute of the users or resources as well as user
roles. The size of the trained decision engine and the weights of the model is a small 1 MB file
and has little computational and memory overhead.
Deep learning models can be trained by utilizing any numerically representable feature. There- fore,
Depending on the use case, the attributes utilized to train the model can be derived from attributes
stored in a medical facility or, in general, an access control system containing users or resources
with various attributes. Nevertheless, a deep learning model does not consider the predefined rules of
the blockchain for accessing resources. To address this, we pass the request through a rule-checker, a
set of static rules defined to ensure the validity of the decision made by the engine. The rule-
checker consists of rules that apply to the user-resource pairs in specific cases.
8. EVALUATION
The evaluation of DLBACα in [41] compared it with various access control methods, demonstrat-
ing that deep learning-based techniques outperformed others. Higher F1 scores indicated improved
generalization and accurate access control for unseen attributes, while higher True Positive Rate
(TPR) and Precision suggested more efficient access grants. Furthermore, a lower False Posi- tive
Rate (FPR) indicated better unauthorized request denials, emphasizing DLBACα ’s superior
performance in making accurate access decisions compared to other methods. The details of the
discussed scores are present in the reference paper [41]. In this section, we assess the feasibil-
ity of DLACB, which integrates blockchain and DLBACα, by evaluating the processing time
of authentication and authorization smart contracts. We evaluate the processing time of both the
authentication and authorization smart contracts. Additionally, we compare the processing time
overhead introduced by the authorization smart contract in conjunction with the deep learning
model against other access control models, RBAC and ABAC. We run all these experiments on
the same computer with the specifications mentioned in the previous section. More details of this
test are provided in the following subsection. To evaluate the performance in different settings,
we run the smart contracts in single-thread and multi-thread modes and consider three conditions
per user attributes in RBAC and six conditions per resource attributes in ABAC, for both of which
these conditions are loaded from a file. For DLBAC, we assume that the weights of the model are
already loaded, and only the inference time is calculated.

16
8.1. Evaluation Results
To evaluate the performance of the smart contracts mentioned before, we run many instances of
these modules with different sets of inputs in different circumstances. Figure 4 provides insights
into the average processing times for Authorization and Authentication smart contracts. The au-
thorization access control corresponds to the DLACB’s average time for a single request, while
the Authentication access control represents the time required for a single database query.
Fig. 4: Smart Contract Processing Time
In Figure 5, we conduct a comparative analysis of the access control processing times across
different methodologies, encompassing deep learning-based, role-based (RBAC), and attribute-
based (ABAC) access control. This figure illustrates that, unlike RBAC and ABAC, the DLBAC
processing time remains relatively stable even when subjected to higher request volumes. The
analysis underscores the suitability of DLBAC for larger enterprises, where the design complexity
of RBAC and ABAC often leads to slower performance, especially in handling multiple concurrent
requests. Conversely, in smaller organizations with fewer users and resource attributes, RBAC and
ABAC might exhibit a processing time advantage.
In the multi-threaded scenario, as depicted in Figure 5, we present a graph illustrating the timetaken
by different access control methods to process requests. DLBAC, the deep learning-based
approach, and DLBAC’ with authorization overhead are contrasted with RBAC and RBAC’ for
24 and 72-core processors. Likewise, in attribute-based access control (ABAC) and ABAC’, we
show the processing times for 24 and 72-core processors. Notably, DLBAC demonstrates superior
performance, particularly under high request loads, owing to its neural network-based nature.
In part b of the figure, which corresponds to single-threaded operation, we depict similar pro-
cessing times when only a single thread is available, and requests are processed sequentially. The
results are analogous, with the distinguishing feature being that DLBAC outperforms other ap-
proaches even when handling fewer concurrent requests.
9. DISCUSSION AND FUTURE WORK
Our study’s results have demonstrated that using a pre-trained deep learning model can provide
an effective solution for predicting access control policies. However, it is important to note that
training the model online as the users are added to the organization is not preferred due to potential
attacks on the training procedure.

17
To mitigate these risks, we train the model offline using the existing dataset of users and their
access history and provide a pre-trained model to be used for inference. The model used here isa
lightweight model and modifying it for specific applications by adding more features will not
increase the required computation and storage substantially.
Looking toward the future, there are opportunities to further improve the performance of ac-cess
control policy prediction models. One potential avenue for improvement is to train the modelon
access control patterns specific to particular industries or types of organizations. For example,
models used for healthcare organizations can be designed according to the features and proper-
ties of the data available in this domain. In addition, private blockchains which are designed from
Fig. 5: Access Control Time Analysis in Authorization Smart Contract
scratch, can also provide advantages for organizations that require strict control over access to
their data. These private blockchains can be tailored to meet the specific needs of the organiza-
tion, such as implementing different consensus mechanisms or other features that are required for
their particular use case. However, it is important to note that these private blockchains may not
always be necessary. In cases where the needs of an organization are less specific, existing private
blockchains may be sufficient for their needs. It is important to carefully evaluate the requirements
of each use case and choose the appropriate solution based on those requirements.
10. CONCLUSION
Maintaining logs of access control patterns for detecting malicious behavior and determining the
proper amount of access for the users is essential. In access control methods such as ABAC and
RBAC, the administrator engineers attributes, roles, and rules for applying access control. The
administrator also ensures the policies are suitable for each organization and each user has the
correct amount of access to the resources, which can lead to many errors or insufficient or extensive
access of some users to the resource. These methods work for predefined cases within a broader
organization; therefore, they can not make the correct choice when encountering a new issue [11].
Consequently, there is a necessity for an alternative access control method.
In this paper, we presented DLACB, a blockchain-based access control platform that authen-
ticates the users and utilizes a decision engine that employs a deep learning model along with
priority rules to determine the extent of access required upon a user request. DLACB offers sev-
eral unique features that contribute to its secure and efficient performance. For instance, it ensures
confidentiality and integrity by encrypting requests exchanged between two framework entities
with the receiver’s public key. This prevents attackers from comprehending the content of a trans-

18
action or request and ensures the authenticity of the request’s origin. DLACB also prevents replay
attacks by issuing a nonce, a randomly generated token, for each request and produced link, which
ensures that a malicious user cannot reuse previously communicated links. Moreover, DLACB
prevents man-in-the-middle attacks by encrypting the data, including the nonce and the signature
of the message, inside the request or the transaction. Additionally, the platform keeps records of
the user’s requests to access specific data as well as their permission to access them. This pro-
cess is done by logging the requests and the results on the blockchain, enabling DLACB to detect
malicious requests and attacks. We also maintain a record of malicious actions within the stor-
age system and utilize a verification algorithm to compare it against the blockchain. We deployed
DLACB on a private blockchain, conducted a security analysis, and employed this framework to
assess the processing time of deep learning-based access control smart contracts in comparison to
traditional access control smart contract processing. The results revealed that the processing time
of DLBAC remains relatively stable even when subjected to higher request volumes, further
emphasizing its suitability for larger enterprises.
ACKNOWLEDGEMENT
This work was funded by NSF grant CCF-2131509 and Augusta University Provost’s office and
the Translational Research Program of the Department of Medicine.
REFERENCES
[1] Abdi, A.I., Eassa, F.E., Jambi, K., Almarhabi, K., Khemakhem, M., Basuhail, A., Yamin, M.:
Hierarchical blockchain-based multi-chaincode access control for securing iot systems. Electronics
11(5), 711 (2022)
[2] Abou Jaoude, J., Saade, R.G.: Blockchain applications–usage in different domains. IEEE Access 7,
45360–45381 (2019)
[3] Adhikari, R., Busch, C.: Lockless blockchain sharding with multiversion control. In: International
Colloquium on Structural Information and Communication Complexity. pp. 112–131. Springer
(2023)
[4] Ahmed, W.A., MacCarthy, B.L.: Blockchain-enabled supply chain traceability–how wide? how deep?
International Journal of Production Economics 263, 108963 (2023)
[5] Akbarfam, A.J., Heidaripour, M., Maleki, H., Dorai, G., Agrawal, G.: Forensiblock: A provenance-
driven blockchain framework for data forensics and auditability. arXiv preprint arXiv:2308.03927
(2023)
[6] Akbarfam, A.J., Motarjemi, K.K.: Proposing a new protocol for using device-to-device
communications in nar- rowband iot-based systems. In: 2021 11th smart grid conference (SGC). pp.
1–5. IEEE (2021)
[7] Alhajri, M., Rudolph, C., Shahraki, A.S.: A blockchain-based consent mechanism for access to fitness
data in the healthcare context. IEEE Access 10, 22960–22979 (2022)
[8] Almansoori, S., Alzaabi, M., Alrayssi, M., Puthal, D., Dutta, J., Al Shehhi, A.: Machine learning-
based adaptive access control mechanism for private blockchain storage. In: 2023 IEEE 47th Annual
Computers, Software, and Applications Conference (COMPSAC). pp. 1243–1248. IEEE (2023)
[9] Amiri, M.J., Agrawal, D., El Abbadi, A.: Permissioned blockchains: Properties, techniques and
applications. In: Proceedings of the 2021 International Conference on Management of Data. pp.
2813–2820 (2021)
[10] Barkley, J.: Comparing simple role based access control models and access control lists. In:
Proceedings of the second ACM workshop on Role-based access control. pp. 127–132 (1997)
[11] Bauer, L., Cranor, L.F., Reeder, R.W., Reiter, M.K., Vaniea, K.: Real life challenges in access-control
management. In: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems.
pp. 899–908 (2009)
[12] Bhutta, M.N.M., Khwaja, A.A., Nadeem, A., Ahmad, H.F., Khan, M.K., Hanif, M.A., Song, H.,
Alshamari, M., Cao, Y.: A survey on blockchain technology: evolution, architecture and security.
IEEE Access 9, 61048–61073 (2021)
[13] Casino, F., Dasaklis, T.K., Patsakis, C.: A systematic literature review of blockchain-based

19
applications: Current status, classification and open issues. Telematics and informatics 36, 55–81
(2019)
[14] Chari, S.N., Molloy, I.M.: Generation of attribute based access control policy from existing
authorization system (Feb 16 2016), uS Patent 9,264,451
[15] Chen, J.I.Z., Lai, K.L.: Internet of things (iot) authentication and access control by hybrid deep
learning method-a study. Journal of Soft Computing Paradigm (JSCP) 2(04), 236–245 (2020)
[16] Cruz, J.P., Kaji, Y., Yanai, N.: Rbac-sc: Role-based access control using smart contract. Ieee Access
6, 12240– 12251 (2018)
[17] Cui, Z., Fei, X., Zhang, S., Cai, X., Cao, Y., Zhang, W., Chen, J.: A hybrid blockchain-based identity
authentication scheme for multi-wsn. IEEE Transactions on Services Computing 13(2), 241–251
(2020)
[18] Di Francesco Maesa, D., Mori, P., Ricci, L.: Blockchain based access control. In: Distributed
Applications and Interoperable Systems: 17th IFIP WG 6.1 International Conference, DAIS 2017,
Held as Part of the 12th Interna- tional Federated Conference on Distributed Computing Techniques,
DisCoTec 2017, Neuchaˆtel, Switzerland, June 19–22, 2017, Proceedings 17. pp. 206–220. Springer
(2017)
[19] Ding, S., Cao, J., Li, C., Fan, K., Li, H.: A novel attribute-based access control scheme using
blockchain for iot. IEEE Access 7, 38431–38441 (2019)
[20] Ding, Y., Sato, H.: Bloccess: enabling fine-grained access control based on blockchain. Journal of
Network and Systems Management 31(1), 6 (2023)
[21] Egelund-Mu¨ller, B., Elsman, M., Henglein, F., Ross, O.: Automated execution of financial
contracts on blockchains. Business & Information Systems Engineering 59(6), 457–467 (2017)
[22] Ezhil Arasi, V., Indra Gandhi, K., Kulothungan, K.: Auditable attribute-based data access control
using blockchain in cloud storage. The Journal of Supercomputing 78(8), 10772–10798 (2022)
[23] Gordon, W.J., Catalini, C.: Blockchain technology for healthcare: facilitating the transition to patient-
driven interoperability. Computational and structural biotechnology journal 16, 224–230 (2018)
[24] Hayes-Roth, F.: Rule-based systems. Communications of the ACM 28(9), 921–932 (1985)
[25] He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of
the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)
[26] Hu, V.C., Ferraiolo, D., Kuhn, R., Friedman, A.R., Lang, A.J., Cogdell, M.M., Schnitzer, A., Sandlin,
K., Miller, R., Scarfone, K., et al.: Guide to attribute based access control (abac) definition and
considerations (draft). NIST special publication 800(162), 1–54 (2013)
[27] Hu, V.C., Ferraiolo, D.F., Chandramouli, R., Kuhn, D.R.: Attribute-Based Access Control. Artech
House (2017)
[28] Huang, J., He, D., Obaidat, M.S., Vijayakumar, P., Luo, M., Choo, K.K.R.: The application of the
blockchain technology in voting systems: A review. ACM Computing Surveys (CSUR) 54(3), 1–28
(2021)
[29] Innab, N.: Availability, accessibility, privacy and safety issues facing electronic medical records.
International Journal of Security, Privacy and Trust Management 7(1), 1–10 (2018)
[30] Kim, J., Park, N.: Role-based access control video surveillance mechanism modeling in smart
contract environment. Transactions on Emerging Telecommunications Technologies 33(4), e4227
(2022)
[31] Kshetri, N.: Can blockchain strengthen the internet of things? IT professional 19(4), 68–72 (2017)
[32] Lai, R., Chuen, D.L.K.: Blockchain–from public to private. In: Handbook of Blockchain, Digital
Finance, and Inclusion, Volume 2, pp. 145–177. Elsevier (2018)
[33] Lee, S.H.: Deep learning based face mask recognition for access control. Journal of the Korea
Academia-Industrial cooperation Society 21(8), 395–400 (2020)
[34] Madine, M.M., Battah, A.A., Yaqoob, I., Salah, K., Jayaraman, R., Al-Hammadi, Y., Pesic, S.,
Ellahham, S.: Blockchain for giving patients control over their medical records. IEEE Access 8,
193102–193115 (2020)
[35] Maesa, D.D.F., Mori, P., Ricci, L.: Blockchain based access control services. In: 2018 IEEE
International Con- ference on Internet of Things (iThings) and IEEE Green Computing and
Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE
Smart Data (SmartData). pp. 1379–1386. IEEE (2018)
[36] Maesa, D.D.F., Mori, P., Ricci, L.: A blockchain based approach for the definition of auditable access
control systems. Computers & Security 84, 93–119 (2019)
[37] Manolache, M.A., Manolache, S., Tapus, N.: Decision making using the blockchain proof of authority

20
consensus. Procedia Computer Science 199, 580–588 (2022)
[38] Mohanta, B.K., Panda, S.S., Jena, D.: An overview of smart contract and use cases in blockchain
technology. In: 2018 9th international conference on computing, communication and networking
technologies (ICCCNT). pp. 1–4. IEEE (2018)
[39] Nair, S.M., Ramesh, V., Tyagi, A.K.: Issues and challenges (privacy, security, and trust) in
blockchain-based applications. In: Research Anthology on Convergence of Blockchain, Internet of
Things, and Security, pp. 1101–1114. IGI Global (2023)
[40] Nakamoto, S.: Bitcoin: A peer-to-peer electronic cash system. Decentralized business review p.
21260 (2008)
[41] Nobi, M.N., Krishnan, R., Huang, Y., Shakarami, M., Sandhu, R.: Toward deep learning based access
control. In: Proceedings of the Twelveth ACM Conference on Data and Application Security and
Privacy. pp. 143–154 (2022)
[42] Nobi, M.N., Krishnan, R., Huang, Y., Shakarami, M., Sandhu, R.: Toward deep learning based access
control. In: Proceedings of the Twelveth ACM Conference on Data and Application Security and
Privacy. pp. 143–154 (2022)
[43] Outchakoucht, A., Hamza, E.S., Leroy, J.P.: Dynamic access control policy based on blockchain and
machine learning for the internet of things. International Journal of Advanced Computer Science and
Applications 8(7) (2017)
[44] Patil, P., Sangeetha, M., Bhaskar, V.: Blockchain for iot access control, security and privacy: a
review. Wireless Personal Communications 117, 1815–1834 (2021)
[45] Penard, W., van Werkhoven, T.: On the secure hash algorithm family. Cryptography in context pp. 1–
18 (2008)
[46] Putz, B., Menges, F., Pernul, G.: A secure and auditable logging infrastructure based on a
permissioned blockchain. Computers & Security 87, 101602 (2019)
[47] Rahman, M.U.: Scalable role-based access control using the eos blockchain. arXiv preprint
arXiv:2007.02163 (2020)
[48] Rouhani, S., Belchior, R., Cruz, R.S., Deters, R.: Distributed attribute-based access control system
using permissioned blockchain. World Wide Web pp. 1–28 (2021)
[49] Sandhu, R.S., Coyne, E.J., Feinstein, H.L., Youman, C.E.: Role-based access control models.
Computer 29(2), 38–47 (1996)
[50] Sun, S., Chen, S., Du, R., Li, W., Qi, D.: Blockchain based fine-grained and scalable access control
for iot security and privacy. In: 2019 IEEE Fourth International Conference on Data Science in
Cyberspace (DSC). pp. 598–603 (2019). https://doi.org/10.1109/DSC.2019.00097
[51] Troyer, D., Henry, J., Maleki, H., Dorai, G., Sumner, B., Agrawal, G., Ingram, J.: Privacy-preserving
framework to facilitate shared data access for wearable devices. In: 2021 IEEE International
Conference on Big Data (Big Data). pp. 2583–2592. IEEE (2021)
[52] Villarreal, E.R.D., Garc´ıa-Alonso, J., Moguel, E., Alegr´ıa, J.A.H.: Blockchain for healthcare
management systems: A survey on interoperability and security. IEEE Access 11, 5629–5652 (2023)
[53] Wu, M., Wang, K., Cai, X., Guo, S., Guo, M., Rong, C.: A comprehensive survey of blockchain:
From theory to iot applications and beyond. IEEE Internet of Things Journal 6(5), 8114–8154 (2019)
[54] Xiao, Y., Zhang, N., Lou, W., Hou, Y.T.: A survey of distributed consensus protocols for blockchain
networks. IEEE Communications Surveys & Tutorials 22(2), 1432–1465 (2020)
[55] Xu, Y., Xiao, S., Wang, H., Zhang, C., Ni, Z., Zhao, W., Wang, G.: Redactable blockchain-based
secure and accountable data management. IEEE Transactions on Network and Service Management
(2023)
[56] Zaidi, S.Y.A., Shah, M.A., Khattak, H.A., Maple, C., Rauf, H.T., El-Sherbeeny, A.M., El-Meligy,
M.A.: An attribute-based access control for iot using blockchain and smart contracts. Sustainability
13(19), 10556 (2021)
[57] Zgraggen, R.: Political risks and blockchain: Definitions and possible case examples. International
Journal of Security, Privacy and Trust Management (IJSPTM) Vol 9 (2020)
[58] Zheng, Z., Xie, S., Dai, H.N., Chen, X., Wang, H.: Blockchain challenges and opportunities: A
survey. International journal of web and grid services 14(4), 352–375 (2018)
[59] Zou, R., Lv, X., Zhao, J.: Spchain: Blockchain-based medical data sharing and privacy-preserving
ehealth system. Information Processing & Management 58(4), 102604 (2021)

21
AUTHORS
Asma Jodeiri Akbarfam holds a Bachelor’s degree from the University of Tabriz and a
Master’s degree in Computer Engineering from Sharif University of Technology, currently
pursuing a Ph.D. in Cyber and Computer Science atAugusta University. Her research interests
include blockchain, security, access control, and edge computing.
Sina Brazandeh received a Bachelor of Computer Engineering from Shiraz University. He is
now a Master’s student in Computer Engineering at Bilkent University. His research interests
include Computational Biology, Bioinfor- matics, and Deep Learning.
Deepti Gupta received her Ph.D. degree in Computer Science from the Uni- versity of Texas at
San Antonio (UTSA) and also received her M.S. degree in Computer Science from UTSA. Dr.
Deepti Gupta is currently an Assistant Professor at Texas A&M University-Central Texas. Dr.
Gupta’s research interests lie in the areas of security and privacy in the Internet of Things
(IoT) leveraging cloud and edge computing. Her research interests also include the application
of AI and Machine Learning to secure IoT and CPS infrastructures.
Hoda Maleki earned both her Bachelor’s and Master’s degrees at Tehran Poly- technique in
Iran. In 2019, she completed her Ph.D. at the University of Con- necticut. Currently, she holds
the position of Assistant Professor at Augusta University, specializing in applied cryptography
and system security research.

DEEP LEARNING MEETS BLOCKCHAIN FOR AUTOMATED AND SECURE ACCESS CONTROL

More Related Content

Similar to DEEP LEARNING MEETS BLOCKCHAIN FOR AUTOMATED AND SECURE ACCESS CONTROL (20)

Recently uploaded (20)

DEEP LEARNING MEETS BLOCKCHAIN FOR AUTOMATED AND SECURE ACCESS CONTROL