Decision Analytics Journal 10 (2024) 100419

Contents lists available at ScienceDirect

Decision Analytics Journal

journal homepage: www.elsevier.com/locate/dajour

A blockchain privacy-conserving framework for secure medical data

transmission in the internet of medical things
Shrabani Sutradhar a,b , Sudipta Majumder a ,∗, Rajesh Bose c , Haraprasad Mondal a ,∗,
Debnath Bhattacharyya d
a
Dibrugarh University Institute of Engineering and Technology, Dibrugarh University, Dibrugarh, Assam 786004, India
b
Department of Computational Sciences, Brainware University, West Bengal, India
c Department of Computer Science & Engineering; JIS University, West Bengal, India
d Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Vaddeswaram, Andhra Pradesh, India

ARTICLE INFO ABSTRACT

Keywords: The Internet of Medical Things (IoMT) has transformed healthcare, collecting and transmitting vast medical
Blockchain data. This study proposes an innovative solution, integrating blockchain into IoMT within a fog-cloud
Fog computing computing framework for secure medical data transmission. The blockchain-based zero-trust system ensures
Internet of Medical Things (IoMT)
reliable data auditing and Electronic Health Repository (EHR) protection. New blockchain entries atop prior
Quad Merkle tree
blocks deter tampering, with Quad Merkle tree and zero-knowledge proof encryption ensuring data integrity
Zero-knowledge proof encryption
Markov decision process
and privacy. The multi-critic deep deterministic policy gradient algorithm optimizes task-offloading decisions
in the fog-cloud layer. Security analyses validate its effectiveness, improving computing efficiency and ensuring
data fidelity and privacy. These findings position the system as a promising solution for enhancing medical data
security and integrity in edge-fog cloud environments, aligning with evolving healthcare technology demands.

1. Introduction fortifies data integrity and validation, while privacy concerns are ad-
dressed through Zero-knowledge proof encryption techniques [2]. The
The rapid evolution of technology, particularly the Internet of Med- framework also strategically incorporates the Multi Critic Deep De-
ical Things (IoMT), has ushered in a transformative era in healthcare, terministic Policy Gradient (DDPG) algorithm with Markov Decision
revolutionizing the collection, transmission, and utilization of extensive Process (MDP) [4,5] for optimal decision-making in task offloading
medical data. This paradigm shift aligns with Weiser’s vision of ubiqui- within the fog-cloud layer.
tous computing [1], aiming for seamless integration into daily life and To handle real-time medical data efficiently [6,7], our solution
fostering efficient, adaptable, and convenient ambient environments. integrates the Multi Critic DDPG algorithm with MDP, optimizing
At the core of this evolution lies E-Healthcare, an integral facet iden- task offloading and maximizing system performance. Our work inter-
tified by the World Health Organization (WHO) [2], striving not just sects visionary computing [8], healthcare tech, and IoMT challenges,
for the absence of illness but encompassing complete physical, mental, contributing to this dynamic field’s transformation. Grateful for the
and social well-being. However, challenges faced by IoMT, including opportunity to advance secure, efficient IoMT frameworks.
resource limitations, privacy concerns, unauthorized access, network- The proposed solution offers an array of contributions, including
level attacks, and data breaches during data transmission, present enhanced security, privacy preservation, optimized task offloading,
critical obstacles. To address these challenges, the study proposes an scalability in IoMT deployments, and efficient learning mechanisms.
innovative system integrating blockchain into the IoMT framework Positioned as a beacon in the realm of E-Healthcare and IoMT appli-
within a fog-cloud computing architecture. cations, this framework sets the stage for the development of smarter,
The proposed system operates within a fog-cloud infrastructure, integrated, and secure healthcare systems, promising a transformative
ensuring secure medical data transmission and fostering a zero-trust impact on the future of healthcare technology.
system for Electronic Health Repository (EHR) protection. It addresses The following Section 2 of this paper will explore the literature
the single point of failure issue associated with centralized cloud stor- survey. Section 3 outlines the suggested methodology, and provides
age [3] by leveraging a decentralized blockchain-based design. The detailed descriptions of the architecture accompanied by proper di-
integration of smart contracts and the Quad Merkle tree algorithm agram. Section 4 contains detailed working & pseudo code for each

∗ Corresponding authors.
E-mail addresses: [email protected] (S. Sutradhar), [email protected] (S. Majumder), [email protected] (R. Bose),
[email protected] (H. Mondal), [email protected] (D. Bhattacharyya).

https://doi.org/10.1016/j.dajour.2024.100419
Received 28 September 2023; Received in revised form 15 December 2023; Accepted 7 February 2024
Available online 12 February 2024
2772-6622/© 2024 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

system component. In Section 5, a comprehensive comparison between and data sharing while leveraging blockchain, fog computing, and
our proposed system with state-of art. A challenge for deploying in intelligent edge technologies. Efforts have been made to design secure
real world setup is described in Section 6. Sections 7 & 8 details the and privacy-preserving frameworks for sharing medical data within
simulation setup, various case studies, and the required software and IoMT networks [23,31–36]. Researchers like Zhang et al. [23], Wang
hardware. Discussion of result and objectives of our research work et al. [33], and Zhou et al. [37] have developed blockchain-based
are depicted into Section 8, concludes with future research scope solutions integrating encryption techniques to ensure secure and con-
considerations are drawn into the remaining sections. trolled sharing of sensitive medical information while maintaining
patient confidentiality. Addressing scalability challenges in large-scale
2. Related work IoMT deployments has been a focal point of research [12,31,38–41].
Contributions by Qian et al. [38], Jin et al. [32], and Farouk et al. [12]
The healthcare sector has seen significant advancements during have proposed scalable blockchain-based frameworks that improve
and post-COVID-19, marked by technology-driven healthcare systems. data transmission efficiency, enhance security, and support patient-
Telemedicine has emerged as a solution to manage patient data over- centric healthcare services. Researchers have investigated secure au-
load, reduce internet traffic, and lower costs. The Internet of Medical thentication protocols and identity management systems for IoMT de-
Things (IoMT), comprising biomedical devices and internet-connected vices utilizing blockchain technology [42–45]. Zhu et al. [43], Jiang
applications, has gained prominence for its potential to address stan- et al. [44], and Sutradhar et al. [13,14] introduced blockchain-driven
dardization and scalability challenges stated by Khan et al. [9]. Recent authentication mechanisms and dynamic encryption algorithms to es-
research has brought fresh perspectives to the domain of IoMT. Am- tablish secure communication and identity verification among IoMT
ponsah et al. [10] introduced a novel fraud detection and prevention devices. The aforementioned studies collectively contribute to the ad-
method using machine learning and blockchain in healthcare claim vancement of secure, efficient, and privacy-preserving frameworks for
processing. Chakraborty et al. [11] proposed a multi-criteria decision managing medical data within IoMT ecosystems, addressing critical
analysis model for selecting optimal customer service chatbots un- challenges related to security, scalability, task optimization, and data
der uncertainty. Farouk et al. [12] envisioned a blockchain platform sharing while leveraging blockchain, fog computing, and intelligent
for industrial healthcare, opening new vistas for future opportunities. edge technologies.
Sutradhar et al. [13,14] contributed to enhancing identity and ac- Research Gaps:
cess management in healthcare using Hyperledger Fabric and OAuth The identified research gaps encompass a diverse range of crucial
2.0, alongside proposing a dynamic step-wise Tiny Encryption Algo- aspects within the field of the Internet of Medical Things (IoMT). These
rithm for improving Quality-of-Service. Huang et al. [15] put forward gaps highlight the need for advanced research and development to ad-
a blockchain-based eHealth system for auditable Electronic Health dress key challenges and propel the IoMT landscape forward. Key areas
Records (EHRs) manipulation in cloud environments. Numerous stud- include the exploration of robust privacy-preserving architectures, inte-
ies have explored integrating blockchain technology to bolster the gration of emerging technologies like 5G and AI, scalability challenges
security, privacy, and integrity of medical data within IoMT ecosys- in large-scale deployments, secure user authentication and identity
tems proposed by Zubaydi et al. [16], Khezr et al. [17], and Sharma management, and understanding the human factors influencing IoMT
et al. [18]. These approaches emphasize benefits in ensuring data acceptance. The gaps also emphasize the importance of compliance
integrity, secure sharing, and patient privacy. Notably, blockchain- with healthcare regulations, optimization of task offloading techniques,
enabled smart contracts and encryption methods like Zero-Knowledge real-world implementations and evaluations, dynamic adaptability to
Proof play crucial roles in securing medical data [19–23]. Incorporat- evolving healthcare scenarios, and interoperability challenges among
ing fog computing and edge intelligence into IoMT systems has been IoMT devices. Additionally, ethical considerations, security in fog com-
pivotal for addressing data transmission, computational efficiency, and puting layers, long-term reliability studies, community engagement, ed-
resource constraints proposed by Pan et al. [24]. Blockchain and fog ucation, and assessing the environmental impact of IoMT technologies
computing frameworks have been proposed to enhance data processing, emerge as critical areas requiring focused attention for comprehensive
security, and resource optimization in IoMT environments suggested advancements in the field.
by Mohanty et al. [25] & Karmakar et al. [26]. Additionally, studies
have explored reinforcement learning algorithms like Multi Critic Deep 3. Proposed system architecture
Deterministic Policy Gradient (DDPG) combined with Markov Decision
Processes (MDP) for efficient task offloading [27–30]. Efforts have In this section we propose a system to enhance the privacy and
been made to design secure and privacy-preserving frameworks for security challenges in the context of data transmission, network secu-
sharing medical data within IoMT networks [23,31–36]. Researchers rity and secure, smooth data offloading from Edge medical devices to
like Zhang et al. [23], Wang et al. [33], and Zhou et al. [37] have de- the cloud. This paper constructs a blockchain based zero-trust security
veloped blockchain-based solutions integrating encryption techniques model for the healthcare industry. Fig. 1 shows that, the system is
to ensure secure and controlled sharing of sensitive medical informa- working on four layers. Every layer has a specific functionality, and
tion while maintaining patient confidentiality. Addressing scalability design to achieve four system goals: System integrity, Data privacy-
challenges in large-scale IoMT deployments has been a focal point of preservation, load-balancing, secure data off-loading. In first Layer,
research [12,31,38–41] Qian et al. [38], Jin et al. [32], and Farouk Newly created fog nodes undergo verification for the Blockchain net-
et al. [12] have proposed scalable blockchain-based frameworks that work. The Quad Merkle Hash tree algorithm is applied to verify the
improve data transmission efficiency, enhance security, and support integrity of the node. After successful verification, the new fog node is
patient-centric healthcare services. Researchers have investigated se- included in the network. Inclusion is facilitated by deploying a smart
cure authentication protocols and identity management systems for contract, ensuring consistency on the Blockchain. The system becomes
IoMT devices utilizing blockchain technology [42–45] Zhu et al. [43], immutable and tamper-proof, eliminating ambiguity. The process al-
Jiang et al. [44], and Sutradhar et al. [13,14] introduced blockchain- lows for autonomous execution based on predefined rules, reducing
driven authentication mechanisms and dynamic encryption algorithms the potential for human error or manipulation. In second layer, each
to establish secure communication and identity verification among data set undergoes further encryption using the Zero-Knowledge Proof
IoMT devices. The aforementioned studies collectively contribute to cryptographic technique. This technique, commonly used in Blockchain
the advancement of secure, efficient, and privacy-preserving frame- networks, enhances security and preserves privacy. Data can be verified
works for managing medical data within IoMT ecosystems, addressing without revealing the actual information, ensuring a robust layer of
critical challenges related to security, scalability, task optimization, confidentiality. Further in third layer, Verified and encrypted data

2
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

Fig. 1. Architecture of the proposed framework.

is stored in regional fog nodes located closer to the network edge. optimal offloading policy and appropriate node within fog nodes or
Finally in fourth layer, The Markov decision process, combined with the fog-to cloud data offloading.
Multi-critic Deep Deterministic Policy Gradient algorithm, is utilized
for optimal data offloading decisions. This involves determining the 3.6. Novelty of the proposed architecture
optimal offloading policy and selecting the appropriate node within fog
nodes or initiating fog-to-cloud data offloading. The novelty in this research lies in the comprehensive and innova-
tive approach to addressing privacy and security challenges in health-
3.1. Node verification care data transmission, network security, and data offloading from
Edge medical devices to the cloud. The proposed system introduces
The system facilitates the verification of newly created fog node for a blockchain-based zero-trust security model designed to achieve four
the Blockchain network. For this system the Quad Merkle Hash tree key system goals: system integrity, data privacy-preservation, load bal-
algorithm has been applied using the root of the quad Merkle hash ancing, and secure data offloading. The utilization of Blockchain tech-
tree to verify the integrity of the node, which significantly improves nology with Quad Merkle Hash Tree, smart contracts, Zero-knowledge
computing and storage efficiency. proof cryptographic tools, a simple load-balancing algorithm, and the
integration of the Markov decision-making algorithm with the Multi-
3.2. Node inclusion critic Deep Deterministic Policy Gradient (MCDDPG) algorithm sets
this research apart. The detailed sections on node verification, node
After verification new fog node is included into the network by inclusion, data authorization, load balancing within fog nodes, and data
deploying a smart contract, which eliminates ambiguity, ensures con- offloading contribute to the novelty by providing a comprehensive and
sistency on the Blockchain, and makes it immutable and temper-proof. well-integrated solution. The use of Quad Merkle Hash tree for node
It facilitates autonomous execution according to predefined rules and verification, the deployment of smart contracts for node inclusion, and
conditions and eliminates the need for intermediaries or third-party the application of Zero-Knowledge Proof cryptographic technique for
intervention, thereby reducing the potential for human error or ma- data authorization showcase the innovative application of cutting-edge
nipulation. technologies. Additionally, the incorporation of a simple load balancing
technique to handle encrypted data and the use of the Markov decision
3.3. Data authorization process with the MCDDPG algorithm for optimal offloading policy
determination further contribute to the uniqueness of this research.
After inclusion, each data set is further encrypted using the Zero- The holistic and layered approach presented in the system architecture
Knowledge Proof cryptographic technique. Zero-Knowledge Proof cryp- demonstrates a novel and effective strategy for enhancing healthcare
tographic technique is a widely used Blockchain, decentralized network data security and privacy in edge-fog cloud environments.
for enhancing security and preserving privacy. This technique allows
the data to be verified without revealing the information itself. 4. System components

3.4. Load balancing within Fog-Nodes The proposed methodology integrates several algorithms and tech-
niques to enhance various aspects of a fog-based blockchain system.
Verified and encrypted data is stored into the regional fog nodes To design and develop these algorithms, several foundational methods
located closer to the network edge. Here we include a simple load and concepts were used. Here are some of the key methodologies and
balancing technique to deal with a huge crowd of encrypted data. concepts that serve as the basis for the proposed system:
Additionally, single node failure problems which can lead to downtime
and disruption of emergency services can be taken care of with the help 1. Blockchain Technology: The foundation of the proposed sys-
of simple load balancing techniques. tem lies in blockchain technology. Blockchain provides a de-
centralized and immutable ledger that ensures secure and trans-
3.5. Data offloading parent transactions. Concepts like block creation, hashing, con-
sensus mechanisms (such as Proof of Work - PoW or Proof of
Finally, Markov decision process combined with Multi-critic Deep Stake - PoS), and smart contracts are integral parts of blockchain
Deterministic Policy Gradient algorithm is used to determine the technology.

3
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

Fig. 2. Structure of Quad Merkle hash tree — A binary or multinomial structure using hash functions.

2. Merkle Trees: The use of Merkle trees is fundamental in the pro- 4.1. Proposed node verification algorithm
posed system for efficient data verification and authentication.
Merkle trees provide a hierarchical data structure that enables A strong node verification algorithm is crucial for ensuring the
quick and secure verification of data integrity, reducing the integrity and security of the fog nodes in the blockchain network. The
number of hashes exchanged for large datasets. algorithm should be able to verify the authenticity and reliability of
3. Cryptography: Various cryptographic techniques are used newly created fog nodes. In the proposed system architecture, the Quad
throughout the proposed algorithms. These techniques include Merkle Hash tree algorithm is used for node verification. This algorithm
encryption, decryption, hashing functions, digital signatures, utilizes the root of the quad Merkle hash tree to verify the integrity of
and Zero-Knowledge Proofs (ZKPs). Cryptography ensures data the node, providing improved computing and storage efficiency. The
confidentiality, integrity, authenticity, and privacy. specific steps of the node verification algorithm can be outlined as
4. Distributed Systems: The concept of distributed systems, specif- follows:
ically within fog computing, forms the backbone of the pro-
posed algorithms. Fog computing utilizes decentralized resources
closer to end-users, enhancing efficiency and reducing latency.
5. Machine Learning and Reinforcement Learning: The Data
Offloading algorithm incorporates concepts from the field of
machine learning, specifically reinforcement learning techniques
like the Multi-critic Deep Deterministic Policy Gradient (MCD-
DPG). These algorithms learn from observed states, rewards, and
actions to make optimal decisions regarding data offloading.
6. Load Balancing Techniques: Load balancing principles within
fog nodes draw from established load balancing algorithms to
evenly distribute incoming encrypted data among fog nodes.
Concepts like monitoring node loads and assigning data based
on load calculations are utilized.
7. Markov Decision Process (MDP): MDP is a mathematical
framework used in the Data Offloading algorithm to model
decision-making problems with stochastic transitions. It assists
in determining the optimal offloading policy by considering
states, actions, and rewards.
8. Smart Contracts and Autonomous Execution: Smart contracts,
a key component of blockchain technology, facilitate the au-
tonomous execution of predefined rules and conditions. These
contracts ensure transparency, immutability, and tamper resis-
tance within the proposed system.
9. System Optimization Techniques: Optimization techniques are
employed to improve system performance, resource utilization,
and decision-making. These techniques aim to maximize effi-
ciency, minimize resource wastage, and adapt dynamically to
varying conditions.

These foundational methodologies and concepts provide the basis

for the proposed algorithms and systems, enabling the design and
development of a secure, efficient, and reliable fog-based blockchain
network within the Internet of Medical Things (IoMT) architecture.

4
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

Fig. 3. Graphical structure of node inclusion in blockchain with the help of smart contract.

The proposed system uses the Quad Merkle Hash tree algorithm to
boost trust and reliability of fog nodes in the blockchain. Quad Merkle
trees store transaction details in blockchains, grouping transactions in
sets of four to form leaf nodes. This arrangement forms the complete
Merkle tree, and recipients can confirm transaction inclusion in a block
using the Merkle tree root and interim hashes from different clients.
The credibility of the client supplying interim hashes does not matter,
as tampering with them or the block header is costly and leads to
verification failure (see Fig. 2).
Leaf nodes store data block hashes, while non-leaf nodes hold
concatenated children’s hash values, with the root representing the
whole dataset’s hash. Node A’s value is calculated from nodes n,
n2, n3, and n4. Leaf nodes (n1 to n4) store data block hashes (k1
to k4), while non-leaf nodes like node A store combined children
hashes. These trees ensure data integrity and authenticity, quickly
detecting tampering when data is transmitted over insecure channels by
comparing sender and recipient Merkle roots. Matching roots indicate
unaltered data, while differences suggest tampering. Merkle trees offer
efficient integrity verification, requiring minimal hash exchange for
large datasets. They excel at tamper detection since any change affects
the entire structure, helping identify tampered data points by tracking
hash discrepancies. Quad Merkle trees are crucial in blockchains, digital
signatures, and distributed file systems, ensuring data security and
consistency due to their simple yet effective design.

4.2. Proposed node inclusion algorithm

The system’s node inclusion algorithm guarantees smooth and se-

cure integration of newly verified fog nodes into the blockchain net-
work. Fig. 3 shows that smart contracts automate rule execution,
ensuring consistency, immutability, and tamper resistance. This pro-
cess, using consensus and smart contracts, ensures secure, transparent,
and autonomous fog node integration, enhancing system integrity and
security while removing third-party intermediaries.

5
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

4.2.1. The pseudocode of the node inclusion algorithm 4.3.1. The pseudocode of zero-knowledge proof cryptographic technique

4.3. Proposed data authorization algorithm The Data Authorization algorithm utilizes encryption, Zero-
Knowledge Proof, and digital signatures to ensure the confidentiality,
The proposed Data Authorization algorithm aims to enhance the integrity, and authenticity of the data. By encrypting the data and gen-
security and privacy of data in the system by utilizing the Zero- erating a Zero-Knowledge Proof, the algorithm provides a secure way to
Knowledge Proof cryptographic technique. This technique allows for authorize and verify data without revealing sensitive information. The
data verification without revealing the underlying information. use of digital signatures further enhances the trust and non-repudiation
of the data transactions within the system.

4.4. Proposed load balancing algorithm within fog-nodes

The proposed Load Balancing Within Fog-Nodes algorithm aims to

distribute the incoming encrypted data among the regional fog nodes
in order to achieve efficient utilization of resources and prevent single
node failures.

4.4.1. The pseudocode of load balancing algorithm

The Load Balancing Within Fog-Nodes algorithm aims to evenly
distribute the encrypted data among the fog nodes by selecting the fog
node with the minimum load at each iteration. This helps to prevent
any single fog node from becoming overloaded and ensures efficient
utilization of resources. By distributing the data across multiple fog
nodes, the algorithm also reduces the risk of downtime and disruption
of emergency services that may occur due to a single node failure.

6
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

3. Convergence Check:

• Check for convergence based on defined criteria.

4. Optimal Offloading Policy:

• Determine the optimal offloading policy based on the

learned Q-values.
• Select the node for data offloading according to the policy.

4.5.2. The pseudocode of multi-critic deep deterministic policy gradient

4.5. Proposed data offloading algorithm

The proposed Data Offloading algorithm aims to determine the

optimal offloading policy and select the appropriate node for data of-
floading within the fog nodes or fog-to-cloud scenario. This is achieved
through the use of the Markov Decision Process (MDP) combined
with the Multi-critic Deep Deterministic Policy Gradient (MCDDPG)
algorithm.

4.5.1. Proposed data offloading algorithm steps

1. Initialization:

• Input encrypted data (ED), fog node list (F), number of fog
nodes (N), cloud node (C), system state (S), action space
(A), reward function (R), discount factor (𝛾), learning rate
(𝛼), exploration rate (𝜀).
• Initialize Q-values for each state–action pair.

2. Learning Process (Repeat until convergence):

• Observe the current system state (S).

• Select an action (A) using the MCDDPG algorithm based
on the current state and Q-values.
• Execute the selected action (A) in the system.
• Observe the new system state (S’) and associated reward
(R).
• Update the Q-value for the state–action pair using the
formula:
Q(S, A) = (1 - 𝛼) * Q(S, A) +𝛼 * (R +𝛾 * max(Q(S’, A’)))
• Update the system state (S) to the new state (S’).

7
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

Table 1
Comparison between Quad Merkle hash tree & binary Merkle tree.
Aspect Traditional binary Merkle tree [5,20] Quad Merkle tree [8,46]
Structure Each non-leaf node has exactly two children, resulting in a binary Each non-leaf node has four children, forming a quad branching
branching structure structure
Efficiency Handle larger datasets Suitable for smaller datasets. Overall efficiency is less.
Processing speed Not very satisfactory due to sequential dependent processing Faster processing speed due to parallel independent processing
Data reduction and storage Limited reduction capability Reduces storage requirements
Computational overhead Higher computational overhead Lower computational overhead
Search complexity (log N) (m log mN for m-fork tree)
Memory overhead 6 layers, 31 hash operations, and 47 nodes. Moderate 4 layers, 21 hash operations, and 37 nodes. Lower
Flexibility Widely adopted well-established and well-documented structure. Newer variation, comparatively limited acceptance
Degree of parallelism Limited due to sequential processing of hashing of sibling nodes. Higher degree due to hashing of children nodes can be done
independently in parallel.
Depth and efficiency O(log N) Fixed depth of 4 layers, regardless of the number of data blocks or
leaf nodes in the tree

The Data Offloading algorithm utilizes the MDP framework and • Robustness and Stability: Incorporating multiple-critic networks
MCDDPG algorithm to learn the optimal offloading policy based on the boosts algorithm stability and robustness, addressing value func-
observed system states and associated rewards. The algorithm learns tion overestimation or underestimation issues. This results in
through iterations and updates the Q-values for each state–action pair, more reliable and consistent decision-making in task offloading
which represent the expected cumulative rewards for taking a specific scenarios.
action in a particular system state. By selecting actions based on the • Learning from Experience: DDPG uses experience replay, storing
MCDDPG algorithm and updating the Q-values, the algorithm learns to past interactions in a buffer to enhance efficiency and stabi-
make decisions that maximize the long-term rewards and achieve the lize learning. Learning from prior experiences improves decision-
optimal offloading policy. The mathematical expression for the Data making, adapts to change, and boosts performance.
Offloading algorithm can be represented as follows:
5. Result analysis with state-of-art

5.1. Comparative analysis of Quad Merkle hash tree & binary Merkle tree

The Quad Merkle hash tree and the traditional binary Merkle tree
are both data structures used to efficiently verify the integrity and
authenticity of large data sets. While they serve the same purpose,
there are notable differences between the two. Here’s in Table 1 a
comparative analysis has been depicted between Quad Merkle hash tree
and traditional binary Merkle tree:

5.2. Comparative analysis of multi-critic deep deterministic policy gradient

vs traditional actor critic algorithm

The Multi-critic DDPG and Traditional Actor Critic are reinforce-

ment learning methods for policy optimization in decision-making.
Table 2 compares their parameters. Both aim to optimize policies but
differ in architecture and approach. Traditional Actor Critic [47] has
By utilizing the Data Offloading algorithm, the system can dynami-
cally determine the optimal offloading policy and select the appropriate an actor selecting actions based on state and a critic evaluating actions
fog node or fog-to-cloud offloading based on the observed system states via a value function. The actor updates its policy based on the critic’s
and expected rewards. This allows for efficient utilization of resources feedback to maximize returns. In contrast, DDPG [48,49] extends DPG
and improved performance in terms of data transmission and network with deep neural networks. It adds target actor and critic networks
security within the IoMT architecture. for action-value and value functions. DDPG uses a replay buffer and
soft target updates to stabilize learning, making it effective in high-
4.5.3. Advantages of multi-critic deep deterministic policy gradient (DDPG) dimensional spaces but computationally more complex than Traditional
with Markov decision process (MDP) Actor Critic algorithm:
• Optimal Task Offloading: The Multi-critic DDPG algorithm, in
conjunction with MDP, offers an optimal approach for complex 5.3. Performance analysis
task offloading decisions. It merges deep reinforcement learn-
ing with MDPs to optimize task allocation, maximizing system To evaluate the effectiveness of the proposed model, we present a
performance. comprehensive analysis of its performance compared to an alternative
• Adaptive Learning: Using multiple-critic networks in DDPG en- approach in Table 3. The performance metrics are rated on a scale from
ables adaptive learning and precise value function estimation for 1 to 10, with 10 indicating the best performance. The following table
each task. This improves decision-making based on task-specific summarizes the performance metrics for the proposed model and the
requirements, enhancing resource allocation and task offloading alternative approach:
choices. Based on the analysis, the proposed model consistently outperforms
• Performance Maximization: The Multi-critic DDPG algorithm the alternative approach across multiple performance metrics. It ex-
maximizes system performance by considering multiple criteria hibits higher efficiency, faster processing speed, better data reduction
like execution time, energy consumption, and communication and storage capabilities, lower computational overhead, lower search
overhead. This enhances resource utilization and overall system complexity, and lower memory overhead. Additionally, it offers in-
efficiency. creased flexibility, a higher degree of parallelism, and improved depth.

8
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

Table 2
Comparison of performance metrics between multi-critic DDPG and actor critic algorithms.
Algorithm Learning rate Discount factor Average reward Convergence speed Learning stability
Multi-critic DDPG (Case1) 0.001 0.9 100 Fast Stable
Multi-critic DDPG (Case2) 0.01 0.9 120 Moderate Stable
Multi-critic DDPG (Case3) 0.1 0.9 80 Slow Unstable
Actor Critic (Case1) 0.001 0.9 90 Moderate Stable
Actor Critic (Case2) 0.01 0.9 110 Fast Stable
Actor Critic (Case3) 0.1 0.9 70 Slow Unstable

Table 3 can lead to the widespread deployment and adoption of the pro-
Performance metrics comparison between the proposed model and the alternative posed system, ultimately improving data integrity, privacy, and task
approach.
offloading decision-making in real-world healthcare environments.
Metric Proposed model Alternative approach
Efficiency 9 7 7. Simulation setup
Processing speed 9 6
Data reduction and storage 10 7
Computational overhead 8 6
To evaluate the effectiveness of the proposed solution, a simulation-
Search complexity 9 6 based methodology will be employed. This involves creating a simu-
Memory overhead 8 7 lation model that accurately represents the components of the IoMT
Flexibility 9 6 network, including fog nodes, cloud servers, encryption, and commu-
Degree of parallelism 9 6
nication links. The simulation model will incorporate realistic request
Depth 8 7
Overall efficiency 9 6
patterns and consider various factors such as efficiency, processing
speed, data reduction and storage, computational overhead, search
complexity, memory overhead, flexibility, degree of parallelism, depth,
and efficiency. To assess the system’s performance, which utilizes the
Quad Merkle tree algorithm, Zero-knowledge proof encryption through
smart contracts on the blockchain, and the Multi Critic DDPG algo-
rithm with MDP, Python will be used for implementing blockchain
algorithms, reinforcement learning algorithms, and simulations due to
its rich ecosystem of libraries and frameworks. Additionally, a suitable
blockchain framework (e.g., Ethereum, Hyperledger Fabric) will be uti-
lized, along with machine learning libraries (e.g., TensorFlow, PyTorch,
Keras) for implementing the DDPG algorithm with MDP. A simulation
framework such as SimPy, AnyLogic, or NS-3 will be employed for mod-
eling and simulating the system’s behavior, enabling the definition of
components, simulation of interactions, and collection of performance
metrics. The hardware requirements will depend on the scale and com-
plexity of the simulation, with smaller-scale simulations manageable
on standard computers, while larger-scale or real-time data processing
Fig. 4. Graphical representation of performance comparison between proposed model
and alternative approach.
simulations may require more powerful hardware configurations or
cloud-based computing resources.

8. Case study
Notably, the overall efficiency score for the proposed model is 9,
while the alternative approach achieves a score of 6. This significant In this groundbreaking case study, we introduce an innovative
difference indicates that the proposed model is superior in terms of per- blockchain-based privacy-preserving system for healthcare data shar-
formance, making it the preferred choice for addressing the challenges ing. Our system leverages cutting-edge technologies, including the
in the IoMT environment. Fig. 4 shows graphical representation of Quad Merkle tree algorithm for impeccable data integrity, Zero-
comparison of various performance metrices between proposed model Knowledge Proof (ZKP) encryption via smart contracts to ensure pri-
and alternative approach. vacy, and the Multi Critic DDPG algorithm alongside MDP for opti-
mizing task offloading decisions. This work is especially crucial in the
6. Challenges related to deployment and adoption in real-world context of a regional healthcare network comprising diverse institutions
setting needing to securely share patient data while upholding privacy stan-
dards. Our decentralized system architecture, powered by blockchain,
The deployment and adoption of a system utilizing a Quad Merkle establishes unique blockchain addresses for participants, guaranteeing
tree algorithm, Zero-knowledge proof encryption through smart con- secure and private data sharing. The implementation incorporates ad-
tracts on the blockchain, and the Multi Critic DDPG algorithm with vanced features such as individual blockchain addresses, Quad Merkle
MDP may face several challenges in real-world settings. These chal- tree verification, and Zero-Knowledge Proof encryption to conceal
lenges include technical complexity in integrating advanced technolo- sensitive patient data during transmission and storage. The integration
gies, scalability to handle large-scale deployments and increasing data of the Multi Critic DDPG algorithm with MDP in the fog-cloud layer
volumes, security and privacy concerns in protecting sensitive med- optimizes task offloading decisions, considering key factors like com-
ical data, regulatory and legal compliance with healthcare industry putational resources, network latency, and energy consumption. This
standards and data protection laws, acceptance and adoption by stake- optimization ensures efficient resource utilization and alleviates the
holders, and infrastructure and resource requirements for computa- strain on local devices. This case study not only showcases the success-
tional demands. Overcoming these challenges requires a comprehensive ful implementation of our system but also underscores its paramount
approach involving technological advancements, stakeholder collab- significance in enhancing privacy, maintaining data integrity, and
oration, regulatory adherence, security best practices, and ongoing enabling intelligent task offloading decisions within healthcare data
monitoring and improvement. Successfully addressing these challenges sharing — a pivotal advancement in the field.

9
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

9. Discussion References

[1] P. Dhiman, S.K. Henge, S. Singh, A. Kaur, P. Singh, et al., Blockchain Merkle-tree
The Discussion section of our paper underscores the implementa- ethereum approach in enterprise multitenant cloud environment, Comput. Mater.
tion of an innovative blockchain-based privacy-preserving system for Contin. 74 (2) (2023) 3297–3313.
healthcare data sharing in the Internet of Medical Things (IoMT) en- [2] Z. Liu, L. Ren, Y. Feng, S. Wang, J. Wei, Data integrity audit scheme based
on quad Merkle tree and blockchain, IEEE Access 11 (2023) 59263–59273,
vironment. This system employs advanced technologies, including the http://dx.doi.org/10.1109/ACCESS.2023.3240066.
Quad Merkle tree algorithm for robust data integrity, Zero-Knowledge [3] C. Hebert, F.D. Cerbo, Secure blockchain in the enterprise: A methodology,
Proof (ZKP) encryption through smart contracts for privacy preserva- Pervasive Mob. Comput. 59 (2019) 101038.
[4] W. Rafique, M. Khan, S. Khan, J.S. Ally, SecureMed: A blockchain-based privacy-
tion, and the Multi Critic Deep Deterministic Policy Gradient (DDPG)
preserving framework for Internet of Medical Things, Wirel. Commun. Mob.
algorithm alongside Markov Decision Process (MDP) for optimizing Comput. 2023 (2023) http://dx.doi.org/10.1155/2023/2558469, 14 pages.
task offloading decisions. Our simulation-based methodology ensures [5] Z. Chai, G. Li, Quadtree-based geospatial indexing for large-scale spatial data,
a comprehensive evaluation, considering diverse factors such as effi- IEEE Trans. Big Data (2022).
[6] J.H. Ryu, P.K. Sharma, J.H. Jo, J.H. Park, A blockchain-based decentralized
ciency, processing speed, data reduction and storage, computational efficient investigation framework for IoT digital forensics, J. Supercomput. 75
overhead, search complexity, memory overhead, flexibility, degree of (2019) 4372–4387.
parallelism, and depth. The implementation leverages Python, [7] I. Ali, M. Gervais, E. Ahene, F. Li, A blockchain-based certificateless public key
signature scheme for vehicle-to-infrastructure communication in VANETs, J. Syst.
blockchain frameworks (e.g., Ethereum, Hyperledger Fabric), and ma-
Archit. 99 (2019) 101636.
chine learning libraries (e.g., TensorFlow, PyTorch, Keras). Perfor- [8] R. García-Fernández, L. Velasco, A scalable Merkle tree construction based on
mance analysis reveals that our model consistently outperforms alter- recursive bloom filters, 2023, arXiv.
natives across multiple metrics, showcasing higher efficiency, faster [9] A.A. Khan, Z.A. Shaikh, L. Baitenova, L. Mutaliyeva, N. Moiseev, A. Mikhaylov,
A.A. Laghari, S.A. Idris, H. Alshazly, QoS-Ledger: Smart contracts and meta-
processing speed, better data reduction and storage, lower computa- heuristic for secure Quality-of-Service and cost-efficient scheduling of medical-
tional and memory overhead, lower search complexity, increased flex- data processing, Electronics 10 (24) (2021) 3083, http://dx.doi.org/10.3390/
ibility, higher parallelism, and improved depth. The proposed model’s electronics10243083.
[10] A.A. Amponsah, A.F. Adekoya, B.A. Weyori, A novel fraud detection and
significantly higher overall efficiency score establishes its superior-
prevention method for healthcare claim processing using machine learning and
ity in addressing IoMT challenges. Our work’s objectives center on blockchain technology, Decis. Anal. J. 4 (2022) 100122, http://dx.doi.org/10.
enhancing data integrity verification, privacy preservation, and task 1016/j.dajour.2022.100122.
offloading decision-making. The conclusion emphasizes the system’s [11] R.K. Chakrabortty, M. Abdel-Basset, A.M. Ali, A multi-criteria decision analysis
model for selecting an optimum customer service chatbot under uncertainty,
potential while acknowledging challenges such as technical complexity,
Decis. Anal. J. 6 (2023) 100168, http://dx.doi.org/10.1016/j.dajour.2023.
scalability, security and privacy concerns, regulatory compliance, and 100168.
stakeholder acceptance and adoption. In essence, our work presents [12] A. Farouk, A. Alahmadi, S. Ghose, A. Mashatan, Blockchain platform for
a promising and impactful solution to elevate data integrity, privacy, industrial healthcare: Vision and future opportunities, Comput. Commun.
154 (2020) 223–235, https://www.researchgate.net/publication/376164702_
and task offloading decision-making within the IoMT environment, Routing_algorithms_in_internet_of_things_complex_network_with_the_role_of_
addressing critical challenges and advancing the state of the field. machine_learning.
[13] S. Sutradhar, S. Karforma, R. Bose, S. Roy, S. Djebali, D. Bhattacharyya,
Enhancing identity and access management using hyperledger fabric and OAuth
10. Conclusion and future scope 2.0: A block-chain-based approach for security and scalability for the healthcare
industry, Internet of Things Cyber-Phys. Syst. (2023) https://onlinelibrary.wiley.
com/doi/abs/10.1111/j.0953-816X.2004.03407.x.
Our integrated healthcare system stands as a promising solution [14] S. Sutradhar, S. Karforma, R. Bose, S. Roy, A dynamic step-wise tiny encryption
to IoMT challenges, offering robust security, scalability, and enhanced algorithm with fruit fly optimization for Quality-of-Service improvement in
healthcare, Healthc. Anal. 3 (2023) 100177, https://www.sciencedirect.com/
data integrity, privacy, and task offloading. Leveraging smart tech- science/article/pii/S2772442523000448.
nologies like smartphones, IoMT, and wireless sensor networks, it [15] H. Huang, X. Sun, F. Xiao, P. Zhu, W. Wang, Blockchain-based ehealth system
ensures comprehensive support care for overall wellness. The use of for auditable EHRs manipulation in cloud environments, J. Parallel Distrib.
Comput. 148 (2021) 46–57, https://www.sciencedirect.com/journal/journal-of-
a distributed blockchain-based fog-oriented privacy-conserving frame-
parallel-and-distributed-computing/special-issue/10XP13P39BM.
work facilitates controlled medical data transmission, fortified by Zero- [16] H.D. Zubaydi, Y.-W. Chong, K. Ko, S.M. Hanshi, S. Karuppayah, A review on the
Knowledge Proof encryption. The Multi Critic DDPG algorithm with role of blockchain technology in the healthcare domain, Electronics 8 (6) (2019)
MDP optimizes task-offloading in the fog-cloud layer, ensuring su- 679, http://dx.doi.org/10.3390/electronics8060679.
[17] S. Khezr, M. Moniruzzaman, A. Yassine, R. Benlamri, Blockchain technology in
perior decision-making and scalability. Real-world implementations healthcare: A comprehensive review and directions for future research, Appl. Sci.
provide deep insights into performance and feasibility. Future work 9 (9) (2019) 1736, http://dx.doi.org/10.3390/app9091736.
involves optimizing system efficiency, adopting advanced encryption [18] A. Sharma, R.Sarishma Tomar, N. Chilamkurti, B.-G. Kim, Blockchain based smart
contracts for Internet of Medical Things in e-healthcare, Electronics 9 (10) (2020)
and privacy mechanisms, and exploring edge computing, AI, and 5G
1609, http://dx.doi.org/10.3390/electronics9101609.
networks for further advancements. System refinement aims to meet [19] H.E. Michail, G.S. Athanasiou, G. Theodoridis, A. Gregoriades, C.E. Goutis, Design
industry requirements, comply with regulations, and address ethical and implementation of totally-self checking SHA-1 and SHA-256 hash functions’
considerations in healthcare data transmission and privacy. architectures, Microprocess. Microsyst. 45 (2016) 227–240.
[20] Z. Zhao, X. Xu, Y. Ge, Secure and efficient query scheme over encrypted
blockchain using Merkle Tree, J. Netw. Comput. Appl. 117 (2018) 62–70.
[21] I.G. Prokopenko, Nonparametric algorithms for detection of radar Markov signals
Declaration of competing interest
against the background of Markov noise, in: 2020 21st International Radar Sym-
posium, IRS, 2020, pp. 356–361, http://dx.doi.org/10.23919/IRS48640.2020.
The authors declare that they have no known competing finan- 9253776.
[22] Y. Su, Q. Li, X. Sun, X. Zhao, H. Zhang, Secure and efficient medical data
cial interests or personal relationships that could have appeared to sharing using blockchain-based attribute encryption scheme, IEEE Access 8
influence the work reported in this paper. (2020) 54606–54619.
[23] Y. Zhang, S. Li, Y. Ren, X.S. Shen, Blockchain-based privacy-preserving frame-
work for medical data sharing in IoMT, IEEE Internet Things J. 10 (4) (2023)
Data availability 3815–3825.
[24] J. Pan, J. Wang, A. Hester, I. Alqerm, Y. Liu, Y. Zhao, Edgechain: An edge-IoT
framework and prototype based on blockchain and smart contracts, IEEE Internet
No data was used for the research described in the article. Things J. 6 (3) (2019) 4719–4732.

10
S. Sutradhar, S. Majumder, R. Bose et al. Decision Analytics Journal 10 (2024) 100419

[25] S.N. Mohanty, K. Ramya, S.S. Rani, et al., An efficient lightweight integrated [36] Y. Liu, Z. Xu, Q. Zhang, H. Wang, Blockchain-enabled privacy-preserving
blockchain (elib) model for iot security and privacy, Future Gener. Comput. Syst. framework for health data sharing in IoMT, IEEE Trans. Netw. Sci. Eng. 10
102 (2020) 1027–1037. (4) (2023) 2781–2793.
[26] K.K. Karmakar, V. Varadharajan, U. Tupakula, S. Nepal, C. Thapa, Towards [37] Y. Zhou, X. Chen, G. Zhang, Z. Wang, Secure data sharing scheme in IoMT based
a security enhanced virtualised network infrastructure for Internet of Medical on blockchain and homomorphic encryption, J. Med. Internet Res. 25 (3) (2023)
Things (IoMT), in: 2020 6th IEEE Conference on Network Softwarization, NetSoft, e43567.
Ghent, Belgium, 2020, pp. 257–261. [38] K. Qian, R. Wang, X. Li, Blockchain-based privacy-preserving and secure sharing
[27] J. Wu, R. Wang, R. Li, H. Zhang, X. Hu, Multi-critic DDPG method and double of healthcare data for mobile-edge computing, Future Gener. Comput. Syst. 113
experience replay, in: 2018 IEEE International Conference on Systems, Man, (2020) 720–733.
and Cybernetics, SMC, 2018, pp. 165–171, http://dx.doi.org/10.1109/SMC.2018. [39] S. Kim, J. Park, S. Lee, Y. Chung, Privacy-preserving data aggregation in IoMT
00039. using blockchain, Sensors 23 (7) (2023) 2100.
[28] K.-H. Liu, Y.-H. Hsu, W.-N. Lin, W. Liao, Fine-grained offloading for multi-access [40] S. Yang, J. Wang, Y. Zhang, H. Zhang, Enhanced data integrity assurance in
edge computing with actor-critic federated learning, in: 2021 IEEE Wireless IoMT using blockchain and zero-knowledge proof, J. Biomed. Inform. 124 (2023)
Communications and Networking Conference, WCNC, 2021, http://dx.doi.org/ 103802.
10.1109/WCNC49053.2021.9417477. [41] Y. Chen, J. Zhang, S. Li, W. Chen, Blockchain-enabled secure framework for
[29] K.-H. Liu, W. Liao, Intelligent offloading for multi-access edge computing: A IoMT data analytics, Inf. Fusion 85 (2023) 261–273.
new actor-critic approach, in: ICC 2020-2020 IEEE International Conference on [42] H. Wu, X. Zhang, L. Wang, S. Li, Privacy-preserving IoMT data sharing via
Communications, ICC, 2020, pp. 1–6, http://dx.doi.org/10.1109/ICC40277.2020. blockchain-based attribute-based encryption, J. Parallel Distrib. Comput. 160
9149387. (2023) 127–138.
[30] T. Gao, Q. Tang, J. Li, Y. Zhang, Y. Li, J. Zhang, A particle swarm optimization [43] L. Zhu, Q. Liu, F. Gao, X. Cheng, Blockchain-driven secure authentication
with Lévy flight for service caching and task offloading in edge-cloud computing, protocol for IoMT devices, Comput. Secur. 116 (2023) 102672.
IEEE Access 10 (2022) 76636–76647, http://dx.doi.org/10.1109/ACCESS.2022. [44] H. Jiang, Y. Wang, J. Wu, X.S. Shen, Efficient smart contract design for IoMT
3192846. data transactions using blockchain, J. Syst. Archit. 116 (2023) 102092.
[31] H. Chen, Q. Wang, J. Li, K. Li, A secure and scalable blockchain-enabled IoMT [45] X. Guo, S. Zhang, J. Chen, H. Wang, Blockchain-enhanced identity management
system for patient-centric healthcare, Sensors 23 (1) (2023) 123. for IoMT devices, Future Gener. Comput. Syst. 128 (2023) 511–521.
[32] L. Jin, X. Li, Z. Li, F. Jiang, Enhancing security and privacy in IoMT with [46] D. Kwon, P. Viswanath, An efficient binary Merkle tree construction for
lightweight blockchain, J. Med. Syst. 47 (3) (2023) 57. blockchain systems, IEEE Access (2022).
[33] Y. Wang, L. Zhang, W. Chen, J. Wu, Efficient data transmission in IoMT [47] V.R. Konda, J.N. Tsitsiklis, Actor-critic algorithms, SIAM J. Control Optim. 42
using blockchain-based smart contracts, IEEE Trans. Ind. Inform. 19 (2) (2023) (4) (2000) 1143–1166.
1545–1555. [48] T.P. Lillicrap, J.J. Hunt, A. Pritzel, N. Heess, T. Erez, Y. Tassa, D. Silver, D.
[34] S. Xu, Q. Liu, L. Cui, H. Jiang, Blockchain-based secure authentication for IoMT Wierstra, Continuous control with deep reinforcement learning, 2015, arXiv
devices, Future Gener. Comput. Syst. 129 (2023) 176–187. preprint arXiv:1509.02971.
[35] J. Li, W. Li, X. Guo, K. Yang, Blockchain-enabled access control for IoMT data [49] D. Silver, G. Lever, N. Heess, T. Degris, D. Wierstra, M. Riedmiller, Deterministic
sharing, J. Ambient Intell. Humaniz. Comput. 14 (5) (2023) 1709–1721. policy gradient algorithms, ICML, in: Proceedings of the 31st International
Conference on Machine Learning, Vol. 32, 2014.

11