Abstract
Existing supply chain information security methods all suffer from the difficulty of balancing information sharing efficiency and information privacy protection. Data is prone to leakage, resulting in an increase in the overall risk of the supply chain. In response to this situation, the study proposes a supply chain information security sharing method based on a blockchain consensus algorithm and federated learning. The study designs a blockchain formula algorithm based on a verifiable mechanism and combines this algorithm with federated learning to construct an encryption model. To address the issue of privacy leakage that is prone to occur in federated learning, this study introduces casual pseudo-random functions and cuckoo hashing to process data and reduce communication complexity, thereby avoiding hash conflicts. Finally, the encryption model is applied to the data transmission system, and combined with multi-factor authentication, the secure sharing of supply chain information is achieved. The experiment results indicated that the average latency of the consensus algorithm during node election was 115.20ms, and during node replacement, the average latency was 8.56ms. The information security sharing method proposed in the study achieved an accuracy rate of 92.48% in generating data after processing, with a data tampering detection rate of 98.87%. The frequency of privacy breaches during the experiment was only 0.1%, and the average response time was 1.0 s. The study proposes a new technical framework that takes into account both privacy protection and efficient sharing, effectively balancing the trust establishment and data security requirements in multi-party collaboration in the supply chain, and providing a verifiable and traceable technical path for information sharing in complex network environments. At the same time, by optimizing the integration mode of the consensus mechanism and federated learning, the system communication overhead and response delay have been significantly reduced, and the overall operational efficiency has been improved.
Keywords: Blockchain, Consensus algorithm, Federated learning, Supply chain, Information security, Data sharing
Subject terms: Computer science, Information technology
Introduction
As an important component of global trade, the security and efficiency of the supply chain (SC) are directly related to the stability and development of the economy. As information technology rapidly develops, SC management is gradually transitioning towards digitization and networking. However, information security issues in the SC are becoming increasingly prominent and have become one of the key factors restricting its development1. If data in the SC is maliciously tampered with or illegally accessed, it will not only cause losses to the enterprise but may also have a chain reaction on the entire SC, affecting the interests of other participants. Therefore, ensuring the implementation of SC information security sharing technology is critical for maintaining the stable operation of the SC2. With the continuous development of computer technology, there is an increasing amount of research on SC information security sharing technology. However, the current information security sharing methods still find it difficult to balance privacy protection and efficiency, especially in complex scenarios involving multiple parties, where the problems of data silos and trust barriers are prominent. Most of the existing technologies rely on centralized architectures, which are prone to single points of failure and are difficult to resist internal attacks and external infiltration. In addition, although traditional encryption methods can ensure transmission security, they sacrifice data availability and restrict the efficiency of information sharing3. Blockchain technology, with its decentralized, immutable, and traceable features, provides a new idea for building a trusted shared environment. Federated learning (FL), on the other hand, implements a collaborative training model where “data remains stationary while the model moves”, enabling knowledge extraction without sharing the original data4. In response to this situation, this study proposes a technical framework for SC information security sharing based on blockchain consensus algorithms and FL, aiming to achieve secure sharing of SC data. While ensuring the efficiency of data sharing, it will further enhance data privacy protection and ensure that information at all stages of the SC is transparent and tamper-proof. The innovation of the research lies in the introduction of a sample alignment scheme and the combination of consensus algorithms to select consensus nodes to assist in participating in FL. This solves the third-party dependency problem in FL, thereby improving the efficiency and security of data sharing. In addition, the study introduces a verifiable mechanism to design the blockchain consensus algorithm, achieving a decentralized trust mechanism to ensure the authenticity and consistency of data. The contributions of this study are: (1) Establishing an efficient and secure SC information sharing environment, reducing the risk of data leakage, enhancing the overall performance of the system, and effectively addressing trust issues in multi-party collaboration. (2) Through the deep integration of blockchain and FL, the dual goals of data privacy protection and efficient sharing have been achieved, providing verifiable, traceable, and undeniable information interaction mechanisms for enterprises in the upstream and downstream of the SC, and further promoting the trusted process of digital transformation in the SC.
Related works
Li et al. conducted a retrospective analysis of the existing defense framework to protect the information security in the shared system. It was found that the approach combining FL only shares model updates for aggregation and compares the practicality of different defense strategies. The analysis results show that distributed machine learning (ML) in FL can effectively achieve privacy protection5. Ramakrishna et al. conducted an empirical analysis on the applicability of mainstream algorithms such as DES, 3DES, AES, RSA, and ECC in supply chain data transmission and storage scenarios to facilitate a detailed discussion on encryption technologies. The results indicate that the combination of AES-256 and ECC demonstrates optimal performance in balancing efficiency and security strength6. Mou et al. proposed a lightweight encryption framework to address the widespread issues of high cost and low time efficiency in wireless body area network (WBAN) encryption schemes, integrating semi-tensor product compressed sensing (STP-CS) with a dynamic session key generation mechanism. The results show that this framework can reduce the key update cycle to under 30 s and lower energy consumption by 42% in 256 × 256 image encryption tasks7. Lian et al. proposed a two-stage FL method based on blockchain to effectively protect patients’ private data in the medical Internet of Things. This method enables devices to collaboratively train the global model and addresses the issue of poor training on non-independent and uniformly distributed data. The results show that this method can improve the model accuracy without threatening user privacy8.
FL is an emerging distributed ML technology that allows multiple participants to jointly train a model without sharing raw data, thereby achieving collaborative optimization of the model while protecting data privacy9. In recent years, the application of FL in privacy protection has been increasing. E. Rabieinejad et al. proposed a hybrid privacy protection scheme based on federated learning and partial homomorphic encryption to prevent unauthorized access and potential data leakage. This scheme was validated for its resistance to model inversion attacks in medical imaging diagnosis scenarios10. B. Rao et al. focused on the issue of enhancing the stealthiness of model inversion attacks in federated learning, providing a comprehensive review of privacy attacks in machine learning. They pointed out that attackers can reverse-engineer the statistical characteristics of client-side local data through gradient reconstruction techniques, even restoring pixel-level information close to the original images. However, existing defense mechanisms such as differential privacy and gradient clipping still face an irreconcilable tension between ensuring model utility and privacy strength11. Zhu J et al. found compatibility issues between blockchain and FL in establishing shared systems, proposed a series of solutions, and explored future development directions. Emphasis was placed on the role of blockchain technology in FL, particularly in ensuring data immutability and transparency12. Zhu M et al. aimed to raise the security and privacy protection of collaborative ML in FL. A new approach was proposed to reduce the dependence on centralized servers and mitigate potential security risks by improving FL protocols. The results show that the data leakage rate is reduced by 15.4% after the application of this method13.
A. Khan et al. proposed an ML framework based on support vector machine to solve the problems of management, device interconnection and resource optimization existing in current communication technologies. This framework exchanged the growth of the distributed alliance network and the information among the connected nodes. In addition, interoperable technologies were introduced to ensure data integrity, smooth communication and infrastructure security. The results show that this method exhibits significant advantages in improving network performance14. A. Khan et al. proposed a system review method based on fog and edge computing to optimize the service quality of blockchain. This method proposed a serverless edge distributed outsourcing computing architecture and realized data integrity and privacy protection through Hyperledger technology. The results show that the application of this method significantly improves the security and performance of the system15.
Based on the above content, although the existing SC information sharing technologies have improved efficiency to a certain extent, there are still challenges in data security and privacy protection. The existing SC information protection methods are difficult to fully adapt to the complex and changeable SC environment, and it is urgent to combine emerging technologies to solve this problem. Among them, blockchain and FL, with their unique technological advantages, have become powerful tools for solving the problems of information security and privacy protection in the SC. However, the existing blockchain technology still has problems such as insufficient scalability and slow transaction speed. FL, on the other hand, relies on third parties, making it difficult to balance data privacy protection and sharing efficiency. Therefore, the research considers combining FL with blockchain technology, which can leverage the advantages of both and overcome their respective shortcomings. The research proposes a distributed collaboration framework that integrates blockchain and FL. It achieves trusted verification of model updates through smart contracts, reduces reliance on central servers by leveraging the decentralized nature of blockchain, and ensures local data privacy with the aid of FL16,17.
The summary and analysis content of the literature review is shown in Table 1.
Table 1.
The summary and analysis content of the literature review.
| Technology | Main content | Advantages | Disadvantages | References |
|---|---|---|---|---|
| Blockchain, knowledge graph, federated learning | Supply chain collaborative management | Improving the knowledge sharing structure | Insufficient data security. Limited privacy protection measures | 3 |
| Asymmetric encryption, blockchain | Supply chain information sharing | Enhancing information transparency | Data is prone to tampering | 4 |
| AES-256, ECC | Discussion on Encryption Technology | Discovering Excellent Encryption Algorithms | Partial Analysis and Comparison | 6 |
| Wireless body area network, semi tensor product compressive sensing, dynamic session key | Propose a lightweight encryption framework. | Deeply integrate semi-tensor product compressive sensing (STP-CS) with the dynamic session key generation mechanism | The balance between encryption effect and efficiency is insufficient. | 7 |
| Blockchain, Federated Learning | Federated learning in healthcare applications | Protecting data privacy | Requiring further architecture optimization | 8 |
| Federated Learning | Mobile crowdsourcing privacy protection | Reducing data leakage rate | Relying on edge computing. Limited application scope | 9 |
| Federated learning, partially homomorphic encryption | Hybrid privacy protection scheme | Prevent unauthorized access and potential data leakage | High computing cost | 10 |
| Differential privacy, gradient shearing | Focus on the issue of enhanced concealment of model inversion attacks in federated learning | Derive the statistical characteristics of the client’s local data | It is difficult to reconcile the model utility and the privacy intensity | 11 |
| Blockchain, Federated Learning | Blockchain service quality optimization | Enhancing data integrity | High complexity | 12 |
| Federated Learning | Propose a review method for fog and edge computing systems | Enhance system security and performance | High complexity | 13 |
Methods and materials
To enhance the data security of the SC information sharing system, an SC information security sharing technology based on blockchain consensus algorithm and FL is proposed. The study first introduces a verifiable mechanism to design a blockchain consensus algorithm and establishes an FL model based on it. This achieves information encryption and enhances the security of information sharing systems.
Design of blockchain consensus algorithm based on verifiable mechanism
The consensus algorithm, as the core mechanism for coordinating participants to reach consensus in blockchain networks, directly affects the efficiency and security of the entire system. In the context of SC information security sharing, it is crucial to design an efficient consensus algorithm18,19. A decentralized consensus algorithm based on verifiable mechanisms is proposed to address the performance bottleneck problem of traditional blockchain technology in processing large-scale data. By introducing a verifiable consensus mechanism, this algorithm can ensure improved transaction processing speed and network throughput without sacrificing data security. Alliance chain is a blockchain network structure suitable for SC information sharing, as it allows limited, pre-selected nodes to participate in the consensus process of the network. In such a network, trust relationships can be established between nodes, thereby reducing the likelihood of malicious behavior. Therefore, the study chooses alliance chain groups as the foundational architecture for consensus algorithms. The blockchain architecture utilized in the research is shown in Fig. 1.
Fig. 1.
The blockchain architecture used in the study.
As shown in Fig. 1, the blockchain architecture used in the research includes the client side, the master node, the accounting node, and the authentication center. Among them, the client side is used to submit user operations. To use the consortium chain service, users first need to undergo identity verification at the authentication center. Clients that pass the verification will obtain certificates and be able to issue transaction proposals to the consortium chain. Nodes include the master node, endorsement node, and accounting node. The master node is responsible for maintaining the stable operation of the network and managing the consensus process of the network. The endorsement node is responsible for verifying and endorsing the transaction proposal to ensure the validity of the transaction. The accounting node is responsible for packaging the verified transactions into blocks and adding the blocks to the blockchain. The authentication center is the trust core of the entire architecture, responsible for managing the identity authentication and permission allocation of nodes, ensuring that only authorized nodes can participate in the consensus process. The Reliable-Replicated-Redundant and Fault-Tolerant (Raft) algorithm is a classic consensus algorithm in consortium chains. Nodes in the cluster select the master node through voting, which processes client requests uniformly and forwards them to child nodes as logs, ensuring data consistency among distributed nodes in the cluster. When Raft votes for the master, a random timer is set for each node. The node that finishes the timer first can send a voting request message to other nodes, and the node that receives the majority of votes is elected as the master node. However, this selection method is easily affected by network partitioning and the number of cluster nodes, and there are also issues with low node replacement efficiency and slow consensus speed in Raft.
A verifiable mechanism is proposed to address this issue. This mechanism utilizes verifiable random functions to improve the node election method of Raft algorithm. Verifiable random functions mainly include hash functions, verify functions, and a key generation process. Specifically, it can be expressed as Eq. (1).
 |
1 |
In Eq. (1), 
is the hash value, 
is the proof parameter, which can be utilized to verify the correctness of random values, 
is the hash function, 
is the private key, m is any string input, T and F are the verification correct and verification incorrect, 
is the verify function, and 
is the public key. Verifiable random functions generate random hash values by the public key owner using the private key as input. Others verify whether the hash value is generated by the corresponding private key using the public key and proof value. The study uses a private key and any value as inputs to a hash function to generate a random timer value. In the Raft algorithm, the timer value of each node will be determined by its private key and a random number. Using RSA public key, arbitrary string input value, hash function, and generated timer value as parameters, the node broadcasts these information during the election. After receiving the information, other nodes use the corresponding public key and input value to verify the legitimacy of the timer value. In this way, malicious nodes are prevented from influencing the election results by forging timer values. In addition, the study adds the role of consensus committee nodes and selects the main node and consensus committee nodes through a verifiable random function lottery algorithm20–22. On this basis, the study sets up four transition methods for the states of slave nodes, candidate nodes, consensus committee nodes, and master nodes, as shown in Fig. 2.
Fig. 2.
Node state transition process in the consensus algorithm.
As shown in Fig. 2, during the node state transition process, initially, all nodes are slave nodes. When receiving an FL task request sent by the training participation method, the slave nodes execute the algorithm to draw lots. If the draw result is positive, the node becomes a candidate node and sends a join request to the consensus committee node. After receiving a request, the consensus committee node will evaluate its performance and historical performance to decide whether to accept the node as a new member of the consensus committee. Once a member of the consensus committee, the node will participate in the subsequent consensus process and in the election of the master node if necessary23,24. If the draw result is negative, the node will remain in a slave state and continue to wait for the next draw opportunity. In addition, to improve the fault tolerance and security of the system, the study also introduces a node health monitoring mechanism. This mechanism regularly checks the operational status of nodes. Once a node is found to have a malfunction or performance degradation, it will automatically transition from its current state to a slave node, ensuring the stable operation of the entire blockchain network. Due to the need for multiple rounds of communication for consensus negotiation in blockchain systems, to avoid consuming a large amount of energy, research is being conducted on sharding to restrict nodes to shards and prevent them from participating in global consensus, thereby reducing the energy consumption of information sharing blockchain. The consensus energy consumption calculation of the Raft algorithm is shown in Eq. (2).
 |
2 |
In Eq. (2), 
and 
are the downlink and uplink delays, 
denotes the total amount of nodes, 
denotes the total transmission power, and 
indicates the transmission energy consumption. To guarantee the success of the committee consensus, the study classifies the infinite blockchain network nodes into multiple zones according to their geographical locations. Each zone is assigned a consensus committee node. The energy consumption after fragmentation is shown in Eq. (3).
 |
3 |
In Eq. (3), 
is the delay of the downlink in each shard, a means the path loss index, y is the number of shards, and 
is the energy consumption after sharding. To ensure the correct operation of the consensus process, four constraint conditions are proposed, as shown in Fig. 3.
Fig. 3.
Four constraints of consensus algorithm.
As shown in Fig. 3, in the selection of nodes for each FL task, it is ensured that there are only 5 consensus committee nodes and 1 master node. In addition, all log entries will only be written from the primary node to the child nodes, and the logs on the primary node will only increase, not be deleted or overwritten. If there is a conflict between the log entries of the child node and the master node, the master node shall prevail, and the child node shall copy the log entries of the master node. When the primary node fails, a replacement primary node needs to be selected from the consensus committee nodes, and the log number of the selected primary node needs to be consistent with the log number before the primary node failure. Based on the above content, the study takes consortium chain as the basic framework of consensus algorithm, and introduces verifiable mechanism to improve the conventional Raft consensus algorithm, ultimately obtaining a suitable blockchain consensus algorithm. The proposed blockchain consensus algorithm provides a foundation for establishing FL models in the future, thereby enhancing data security in the process of SC information sharing. Compared with the existing blockchain technology, the research introduces a verifiable mechanism into its consensus algorithm to ensure the transparency and traceability of data exchange between nodes. Through this mechanism, data tampering and forgery are effectively prevented, enhancing the trustworthiness of the system. Meanwhile, the research introduces the health monitoring mechanism of nodes to monitor the status of nodes in real time and ensure the stable operation of the system. The research also optimizes the communication efficiency of the consensus algorithm, reduces the communication overhead between nodes, and improves the overall processing speed.
Construction of encryption model based on blockchain consensus algorithm and federated learning
After introducing a verifiable mechanism and designing a blockchain consensus algorithm, this study aims to establish a vertical FL model based on this and train an encryption model to enhance the data security of SC information systems25,26. Vertical FL is a technique that allows multiple institutions to jointly train ML models without sharing raw data. In this way, different enterprises in the SC can jointly raise the accuracy and generalization ability of the model while protecting their sensitive data27,28. The architecture of the FL model based on the alliance chain is denoted in Fig. 4.
Fig. 4.
The architecture of the federated learning model based on alliance chain.
In Fig. 4, the study adopts the proposed blockchain formula algorithm for node selection, and the selected consensus committee nodes and master nodes assist in training participants for vertical FL. The master node acts as a third party to assist the two participants in completing the joint training of the sample and the encryption model. To achieve sample alignment among participants, an improved method is proposed based on conventional privacy set intersection techniques. The conventional privacy set intersection technique requires Participant 1 and Participant 2 to calculate the hash values of their respective datasets and send these hash values to each other. Then, both parties determine the shared data by comparing the hash values, thereby achieving data alignment. However, this method carries the risk of privacy breaches as hash values may be used to infer the original data. In response to this situation, research introduces unintentional pseudo-random functions and cuckoo hashing to process data and reduce communication complexity. The sample alignment scheme proposed by the study is shown in Fig. 5.
Fig. 5.
The specific flow of the sample alignment scheme designed by the study.
In Fig. 5, the study first uses an unintentional pseudo-random function to find the intersection of privacy sets. Next, cuckoo hashing is utilized to improve its communication complexity. In this way, Participant 1 and Participant 2 can effectively identify common elements in each other’s datasets without directly exchanging raw data. The improved scheme not only reduces the number of communications but also enhances the strength of privacy protection. The unintentional pseudo-random function first generates pseudo-random numbers through unintentional transmission, as shown in Eq. (4).
 |
4 |
In Eq. (4), 
represents the sender sending the two data 
and 
held by the sender to the receiver. The receiver holds a bit b, selects a data sent by the sender, and formalizes it. The final processed data is 
. Pseudo random functions use a deterministic algorithm to generate pseudo-random numbers, which relies on a key. During unintentional transmission, the sender cannot know which data the receiver ultimately chose, and the receiver cannot determine what specific data the sender holds. Moreover, cuckoo hashing maps data items to a hash table by selecting appropriate hash functions and parameters. When a hash conflict occurs, cuckoo hashing uses its unique mechanism to relocate the conflicting items to other positions in the table, thereby avoiding the expensive linear probing or linked list operations required in traditional hash tables. After aligning the samples, the data is encrypted separately to ensure the security and privacy of the data during transmission. The joint training process of the encryption model is shown in Fig. 6.
Fig. 6.
Joint training process of encryption model.
In Fig. 6, after completing sample alignment and obtaining the intersection of samples from both parties, joint training of the model is required. The study first generates a public-private key pair through the master node and sends the public key to Participant 1 and Participant 2 for encrypting intermediate parameters during the joint model training process. To avoid the failure of the master node during the training process and the inability to decrypt the gradient, the study sets the master node to send the private key to the consensus committee node in a secret sharing manner. In the event of a master node failure, the designated replacement master node is capable of collecting sub keys and subsequently acquiring the private key to maintain the continuity of the joint model training process. Upon completion of the training, the main node will then deploy the training gradient and loss to the chain. In the study, the public key was transmitted to the training participants as the encryption key for homomorphic encryption. In the research, the Paillier homomorphic encryption scheme is chosen because homomorphic encryption allows addition and multiplication operations on encrypted data without decryption. The master node uses key sharing to divide the private key into 4 parts and send them to the other 4 nodes of the consensus committee. In the process of key distribution, the master node first needs to randomly select 3 random numbers and construct a polynomial as shown in Eq. (5).
 |
5 |
In Eq. (5), 
, 
, and 
are three random numbers, 
is a constant term, and x is the input data. The study takes any four number inputs and substitutes them into the polynomial to get 
~
. Finally, the sub-secret is sent to each of the 4 consensus committee nodes. The private key reconstruction process involves first constructing a third-order linear equation of the sub secret and solving the reconstruction. The encryption training of the model is denoted in Fig. 7.
Fig. 7.
The encryption training process of the model.
As shown in Fig. 7, Participant 1 uses the public key sent by the master node as the homomorphic encryption key to calculate the encrypted data, and sends the encrypted data to Participant 2. Participant 2 can calculate the encrypted intermediate parameters and loss values, and send the intermediate parameters to Participant 1 and the loss values to Participant 3. Participant 1 and Participant 2 calculate their respective gradients and use a pseudo-random function to obtain random numbers, which are then sent to the master node. The main node is used to decrypt the sum of the gradient and random number of Participant 1, and send the loss value to Participants 1 and 2. This generates the training log for this round to the consensus committee nodes. Participant 1 and Participant 2 obtain their respective gradient results based on the calculation results, and update the model parameters until the model training is completed. After receiving the training log, the consensus committee node sends a response message to the master node. After receiving the response message, the master node sends a response message to the consensus committee node and stores this training log in the local state. Thus, the training of the encryption model is realized. The loss function during the encryption process is shown in Eq. (6).
 |
6 |
In Eq. (6), 
and 
are the model parameters of Participant 1 and Participant 2, 
is the regularization parameter, 
and 
are the feature spaces of each participant, 
is the label space owned by participant 2, and 
is the loss function of the encrypted model training process. This method enables data processing and analysis without exposing the original data. The above training process is repeated until the loss function converges or a sufficient number of training rounds are reached to complete the federation training. The main node will package and block the training logs of each round, and the remaining nodes on the consortium chain will synchronize the blockchain to complete consensus. After the training is completed, participants can use the trained model for prediction with the assistance of the master node. During the prediction process, the local data involved in the method will not be exposed to other places. Based on the above content, a consensus algorithm is designed for alliance chains in blockchain, and an FL model is established on this basis to achieve joint training of encryption models and secure sharing of SC information data. Compared with the existing FL security models, the method proposed in the study introduces the casual pseudo-random function and cuckoo hash to process the data and reduce the communication complexity. Furthermore, the research sets that the master node sends the private key to the consensus committee nodes in a secret sharing manner. If the master node fails, the replaced master node can collect sub-keys to recover the private key and continue the joint training of the model. Through this mechanism, the dual guarantees of the robustness of the system and data privacy are ensured, further enhancing the security and efficiency of FL.
In the face of attacks similar to Sybil, the system identifies abnormal nodes through multiple authentication and reputation-scoring mechanisms and restricts their participation rights. Participants in adversarial FL are screened through anomaly detection and denoising algorithms during model aggregation to ensure the training data’s purity and the model’s robustness. In multi-factor authentication, the combination of biometric recognition and digital signature is adopted to enhance security. The reputation scoring mechanism is dynamically adjusted based on historical behavior and contribution to ensure fairness. The anomaly detection algorithm uses statistical features to identify abnormal model updates, while the denoising algorithm eliminates noisy data through iterative optimization to ensure the overall performance and security of FL.
Attackers possess the ability to forge identities and tamper with data, and may obtain partial model information. If their goal is to disrupt model convergence or steal sensitive data, this threat model encompasses three dimensions: identity forgery, data tampering, and model information leakage, aiming to comprehensively assess the risks faced by the system and provide a basis for defense strategies.
To protect the label information during subsequent training, the research introduces the differential privacy mechanism. During the stage of label upload and aggregation, controlled noise is injected to prevent attackers from inferring the original labels through gradient analysis. At the same time, in combination with secure multi-party computation, it ensures that each participant only shares the encrypted gradient updates, and the labels are always locally retained.
Results
To assess the effect of the proposed SC information security sharing technology based on blockchain consensus algorithm and FL, different experiments were designed to analyze and discuss it. The performance of the research-designed method was compared with other methods to further confirm its superior performance.
Experimental analysis of blockchain consensus algorithm based on verifiable mechanism
The experimental environment was Windows 10, using Python 3.8 as the development language and PyTorch framework to build a deep learning model. In the experiment, the standard MNIST dataset was used for training and testing to validate the effect of the algorithm. To verify the rationality of the improved Raft consensus algorithm, simulation experiments were conducted with traditional Raft algorithm and KRaft algorithm to compare the real-time and scalability of the three consensus algorithms. The research used three consensus algorithms for node election and node replacement. The experimental platform built a blockchain network consisting of 50 nodes, including 20 master nodes and 30 slave nodes. The master nodes were responsible for block generation and verification, while the slave nodes participated in the consensus process and stored the complete ledger. The experiment set the network latency at 50ms and the bandwidth at 100Mbps to simulate the communication environment in a real SC scenario. The latency of each algorithm under the two tasks is shown in Fig. 8.
Fig. 8.
Node election and replacement delay of three consensus algorithms.
In Fig. 8a, during node election, the improved Raft algorithm exhibited lower latency compared to traditional Raft and KRaft algorithms, with an average latency of only 115.20ms. This indicated that the improved algorithm could reach consensus more quickly during the node election, thereby increasing the system’s response speed. From Fig. 8b, in the scenario of node replacement, the improved Raft algorithm also had a shorter latency, with an average latency of 8.56ms. This indicated that it also had advantages in handling node failures and maintaining system stability.
To further test the real-time and scalability of the algorithm, the delay, throughput, and consensus achievement time of the three algorithms were compared at four different scale nodes of 50, 100, 150, 200, and 250. The specific outcomes are shown in Table 2.
Table 2.
Performance comparison of three consensus algorithms under different node sizes.
| Project | Delay (ms) | Throughput(TPS) | Consensus time(s) | |
|---|---|---|---|---|
| 50-node | Raft | 150.30 | 1200 | 1.25 |
| KRaft | 140.52 | 1300 | 1.12 | |
| The improved Raft | 115.23 | 1400 | 0.91 | |
| 100-node | Raft | 200.42 | 1100 | 1.53 |
| KRaft | 190.62 | 1250 | 1.46 | |
| The improved Raft | 158.26 | 1350 | 1.18 | |
| 150-node | Raft | 250.54 | 1000 | 1.83 |
| KRaft | 240.71 | 1150 | 1.74 | |
| The improved Raft | 210.00 | 1250 | 1.48 | |
| 200-node | Raft | 300.62 | 900 | 2.11 |
| KRaft | 290.87 | 1025 | 2.03 | |
| The improved Raft | 260.11 | 1150 | 1.72 | |
| 250-node | Raft | 350.26 | 800 | 2.43 |
| KRaft | 340.89 | 950 | 2.36 | |
| The improved Raft | 308.46 | 1050 | 2.09 | |
In Table 2, the improved Raft algorithm outperformed traditional Raft algorithm and KRaft algorithm in terms of latency and throughput at different scale nodes. Especially when the node size increased, the delay growth of the improved algorithm was relatively small, demonstrating good scalability. At a scale of 250 nodes, the improved Raft algorithm had a latency of 308.46ms, while the traditional Raft algorithm had a latency of 350.26ms and the KRaft algorithm had a latency of 340.89ms. In terms of throughput, the improved Raft algorithm achieved 1050 TPS at a scale of 250 nodes, which was higher than the traditional Raft algorithm’s 800 TPS and KRaft algorithm’s 950 TPS. In addition, the consensus reaching time also showed a similar trend, and the improved Raft algorithm exhibited faster consensus speed at all node sizes.
Performance analysis of privacy set intersection scheme based on blockchain improvement
To test the superiority of the proposed blockchain improved privacy set intersection scheme (Option 1), the study compared it with the privacy set intersection schemes based on homomorphic confidentiality (Option 2), naive hashing (Option 3), secure multi-party computation (Option 4), secure aggregation protocol (Option 5), and differential privacy (Option 6). The comparison results of protocol functions and protocol complexity among the four schemes are denoted in Table 3.
Table 3.
Comparison results of protocol functions and protocol complexity of the four schemes.
| Privacy set intersection scheme | Protocol function or not | Computational complexity | Communication complexity | |||
|---|---|---|---|---|---|---|
| Security | No multiple interactions required | Multi-party support | No third party required | |||
| Option 1 | Y | Y | Y | Y | 2 t asym | (3 + s)Ny |
| Option 2 | Y | Y | N | Y | 2 t asym | Nylog2Nx |
| Option 3 | N | Y | Y | Y | m asym | Ny |
| Option 4 | Y | N | Y | N | m asym | t + Nx |
| Option 5 | Y | N | Y | N | 2 t asym | Ny + Nx |
| Option 6 | N | N | Y | N | m asym | t + Nylog2Nx |
In Table 3, Ny and Nx are the data sets of the participants, t is the total number of data of the participants, and m is the maximum value of the data sets of the two participants.
As shown in Table 3, Option 1 was comprehensive in protocol functionality, but had relatively high computational and communication complexity. Option 2 did not require a third party, but the communication complexity increased with the volume of data. Option 3 had a single function and relatively low complexity. Options 4 and 5 were somewhat lacking in terms of security and multi-party support, but their complexity was moderate. Option 6, however, performed relatively weakly in both functionality and complexity. Taking all factors into account, Option 1 has more advantages in terms of comprehensive functionality and security.
To prove the computational efficiency of the sample alignment scheme proposed in the study, the sample encryption delay and sample intersection delay of four schemes were compared. The experiment findings are denoted in Fig. 9.
Fig. 9.
Time delay comparison results of four PSI schemes.
As shown in Fig. 9a, under different set sizes, the sample encryption latency of Scheme 1 was consistently lower than the other three schemes. When the number of nodes was 216, the latency was only 125ms, which was significantly lower compared to other schemes. In Fig. 9b, Scheme 1 also performed well in terms of sample intersection delay calculation. This indicated that Scheme 1 could effectively reduce the latency of sample intersection and improve data processing speed while ensuring privacy protection. In addition, the scalability of Scheme 1 in a multi node environment was also verified. As the number of nodes increased, the delay growth of Scheme 1 was relatively small, demonstrating good scalability.
Performance analysis of federated learning model based on blockchain consensus algorithm
The study implemented node selection through consensus algorithm in FL models and improved the sample alignment process by introducing unintentional pseudo-random functions and cuckoo hashing. To prove the effect of the raised FL model, a comparative experiment was conducted with traditional FL models. The comparison findings of the learning accuracy and running time of the two models are shown in Fig. 10.
Fig. 10.
Comparison of learning accuracy and running time of the two models.
In Fig. 10a, when the global iteration number reached 12, the learning accuracy of the proposed model reached a convergence value of 0.98, while the accuracy of the traditional FL model was only 0.82. In Fig. 10b, as the number of data providers increased, the running time of both models gradually increased. But the running time of the proposed method has maintained a relatively small growth rate. Its average running time was only 2145ms, while the average running time of traditional models was 8462ms.
To further test the scalability of the proposed method, the changes in various indicators of each method as the number of nodes increases were studied and compared, as shown in Table 4.
Table 4.
Analysis results of the scalability of blockchain consensus algorithms.
| Number of nodes (in) | Method | Learning accuracy(%) | Running time(s) |
|---|---|---|---|
| 50 | Traditional model | 91.47 | 12.35 |
| Improved model | 96.88 | 8.11 | |
| 120 | Traditional model | 91.06 | 15.84 |
| Improved model | 96.57 | 9.00 | |
| 240 | Traditional model | 89.88 | 17.26 |
| Improved model | 96.54 | 9.34 | |
| 360 | Traditional model | 82.34 | 22.51 |
| Improved model | 96.20 | 9.68 |
As shown in Table 4, as the node scale expanded from 50 to 360, the learning accuracy of the improved model always remained above 96%, with a fluctuation range of less than 0.9%, demonstrating extremely strong stability. In contrast, the accuracy of the traditional model showed a significant downward trend as the number of nodes increased, dropping from 91.47% to 82.34%, a decrease of 9.13%. In terms of operational efficiency, the running time of the improved model only increased from 8.11 s to 9.68 s, representing an increase of 19.4%. The running time of the traditional model rose from 12.35 s to 22.51 s, an increase of 82.3%. Especially at a scale of 360 nodes, the running time of the improved model was only 43% of that of the traditional model, and its learning accuracy was 96.20%, which was 13.86% points higher than that of the traditional model. These data fully verify the role of the node selection mechanism based on the blockchain consensus algorithm and the improved sample alignment technology in enhancing the scalability of the model: while ensuring high precision, they effectively suppress the exponential growth of system overhead, providing technical feasibility for the large-scale deployment of SC nodes.
Performance analysis of information security sharing method based on consensus algorithm and federated learning
To test the data protection effectiveness of the proposed information security sharing method (Method 1) based on the blockchain consensus algorithm and FL, it was compared and analyzed with the more advanced methods in the existing research. The comparison methods included the information sharing and privacy protection methods (Method 2) in reference29, the information sharing and privacy protection methods (Method 3) in reference30, the information sharing and privacy protection methods (Method 4) in reference31, and the information sharing and privacy protection methods (Method 5) in reference32. The study first compared the time consumption of five methods in the encryption and decryption processes, as denoted in Fig. 11.
Fig. 11.
Comparison of the time consuming results of five methods in the process of encryption and decryption.
In Fig. 11a, during the encryption process, the encryption running time of Method 1 among the five methods was consistently lower than that of the other methods. In Fig. 11b, during the decryption process, the running time of all five methods gradually increased with the size of the plaintext. The decryption time curve of Method 1 had always been lower than the other curves. Overall, the average time for encryption and decryption in Method 1 was only 52.48ms and 26.33ms, respectively. Compared to other methods, it was significantly lower.
To validate the usability of the data after applying this method, five methods were used to generate and transmit shared data from the SC shared information system of two SC enterprises in location A. 43,433 pieces of data were collected from each company, and the study divided them into Dataset 1, Dataset 2, Dataset 3, Dataset 4, and Dataset 5 according to their types. The environment for the testing process was the actual SC environment, where real-time monitoring and data transmission tests were conducted on the information sharing systems of two SC enterprises in Location A. The study compared the generated data and raw data of each of the five methods for accuracy comparison. The comparison outcomes are indicated in Fig. 12.
Fig. 12.
Results of data availability analysis in information sharing system after the application of the five methods.
In Fig. 12a and b, Method 1 consistently outperformed other methods in five datasets, with an average accuracy of 92.48% in generating data. Although Method 2 performed slightly worse than Method 1 in Dataset 1 and Dataset 2, it showed good stability and accuracy in Dataset 3, Dataset 4, and Dataset 5. Method 3 performed outstandingly in dataset 1, but gradually decreased in other datasets, showing certain limitations. The performance of Methods 4 and 5 was not as good as the first three in all datasets, especially in dataset 5 where the accuracy was significantly lower than other methods. Through these comparisons, it can be concluded that Method 1 had significant advantages in ensuring data availability and was suitable for use in SC information security sharing systems.
To comprehensively compare the application effects of several methods in information security sharing, the study compared the performance of several methods in SC enterprise information data. Comparative indicators included data tampering detection rate, frequency of privacy breaches, and average response time. The experiment lasted for 8 months, with 4 months as a cycle, to compare the performance of five methods within two cycles. The comparison outcomes are denoted in Table 5.
Table 5.
Comparison of application effects of five methods in data information sharing.
| Project | Period 1 | Period 2 | ||||
|---|---|---|---|---|---|---|
| Data tampering detection rate(%) | Frequency of privacy breach events(%) | Mean response time(s) | Data tampering detection rate | Frequency of privacy breach events(%) | Mean response time(s) | |
| Method 1 | 98.50 | 0.12 | 1.05 | 99.24 | 0.08 | 0.93 |
| Method 2 | 91.11 | 0.25 | 1.92 | 93.45 | 0.19 | 1.88 |
| Method 3 | 80.42 | 0.32 | 2.56 | 82.42 | 0.28 | 2.43 |
| Method 4 | 85.32 | 0.22 | 2.22 | 87.07 | 0.20 | 2.05 |
| Method 5 | 89.03 | 0.26 | 2.14 | 91.51 | 0.24 | 1.97 |
In Table 5, each method outperformed cycle 1 in all indicators during cycle 2. This is because during the continuous experimental process, each method continuously improved its performance through learning and training. Method 1 outperformed other methods in all indicators, with an average data tampering detection rate of 98.87% and an average leakage event frequency of 0.1%. Based on the above content, the proposed SC information security sharing method could effectively raise the level of data protection and reduce the risk of privacy leakage. In addition, the average response time of this method was 1.0s, which was much lower than other methods.
Conclusion
To enhance the data security of the SC information sharing system, an SC information security sharing method based on blockchain consensus algorithm and FL was proposed. This method first designed and improved the consensus algorithm, and on this basis, established an FL model to realize the construction of the encryption model. In the performance analysis experiment of the public recognition algorithm, the average delay of the proposed consensus during the node election process was only 115.20ms, and during the node replacement process, the average delay was only 8.56ms. In the verification and analysis of the proposed privacy set intersection scheme, the sample encryption delay has always been lower than the other three schemes under different set sizes. When the number of nodes was 216, its delay was only 125ms, which is significantly lower than that of the other schemes. When the number of global iterations of the FL model reached 12 times, the learning accuracy reached the convergence value of 0.98, which converged faster than other models. In practical experiments, the average accuracy rate of the data generated by the proposed method was 92.48%. It was significantly higher than other methods. Moreover, the average value of the data tampering detection rate of the research-designed method reached 98.87%, the average frequency of leakage events was 0.1%, and the average response time was only 1.0 s. On the basis of the above content, the proposed information security sharing method based on blockchain consensus algorithm and FL could effectively ensure data security and improve the sample processing efficiency of data sharing. The study only used homomorphic encryption methods in the training of the encryption model. In future research, more efficient encryption methods can be considered to further enhance the performance of the model.
The proposed method has demonstrated high stability and reliability in practical applications, effectively addressing data security threats and reducing the risk of information leakage.
The proposed method not only enhances data security, but also optimizes the sharing process and reduces the operating costs of enterprises. In actual management, the application of this method simplifies the data review process, reduces manual intervention, and improves decision-making efficiency. Through this method, enterprises have achieved efficient data circulation and security protection, enhanced the collaborative ability of each link in the SC, and further consolidated their market competitiveness. Meanwhile, the application of this method has also promoted the trust and cooperation among enterprises in the upstream and downstream of the SC, enhanced the overall operational transparency, and laid a solid foundation for the sustainable development of enterprises.
Discussion
The research proposes an information security sharing method based on blockchain formula algorithms and FL. This method enhances the efficiency of node election and replacement by improving the consensus algorithm and realizes data encryption and sharing in combination with FL. This train of thought is clear and forward-looking. The current SC systems are generally constrained by a large number of scalability and security issues. In reference33, scholars proposed a scalable blockchain-based SC model assisted by home matcha pastries in response to this situation. This model features custom transaction generation rules and a dynamic node management mechanism, which can effectively alleviate the scalability bottleneck of blockchain in SC scenarios. This is consistent with the idea mentioned in the research of improving the consensus algorithm to enhance the efficiency of node management. Both aim to enhance the scalability of the system by optimizing the dynamic adjustment mechanism of nodes. In reference34, scholars found that the supply of drugs is prone to record forgery and drug counterfeiting. For this purpose, it proposes a blockchain-based tamper-proof traceability system, which uses smart contracts to automatically record and verify the entire process of drug circulation. The system copies the records into multiple blocks and achieves efficient traceability through parallel search. This research is highly consistent with the goals of this paper in ensuring data integrity and traceability, both of which prevent data tampering through a distributed architecture. In addition, the automated verification mechanism of smart contracts emphasized in reference34 can provide a reference for further integrating the rule engine in this study to enhance the autonomous decision-making ability in the sharing process.
Based on the above content, the exploration of this study in enhancing data security and sharing efficiency has strong practical value and expansion potential. By integrating the decentralized feature of blockchain with the privacy protection capability of FL, this method not only addresses the dual challenges of data silos and security risks in the current SC but also provides a reusable technical framework for multi-agent collaborative scenarios.
The limitation of the research lies in the fact that the computational cost problem of FL in high-dimensional data has not been completely solved, which may affect the deployment effect of the system in resource-constrained nodes. In addition, the rule design of smart contracts relies on manual presets. When facing complex and dynamic SC demands, their autonomous evolution capabilities still need to be improved. Future work will focus on optimizing the lightweight design of algorithms and exploring the combination of dynamic game models to enhance the adaptive ability of rule engines, to expand the application boundaries of methods in diverse scenarios.
List of notations
The relevant annotations of the parameter symbols proposed in the research are shown in Table 6.
Table 6.
The relevant annotations of the proposed parameter symbols.
| Symbol | Classification | Specified value/decision variable | Explanation |
|---|---|---|---|
| hash | Hash value | Hash function calculation result | Generated by cryptography and of fixed length |
| proof | Proof parameter | Non-interactive proof in zero-knowledge proof | Rely on random numbers for generation |
| sk | Private key | A randomly generated 256-bit string | Used for signature or decryption |
| m | Numerical data | Any string | Plaintext or encoded data |
| pk | Public key | Derived from the private key | Used for verifying signatures or decrypting |
|
Input data pairs | The plaintext/key pair selected by the sender | Select based on the decision variables |
|
Random number | Cryptography is randomly generated securely | Used for constructing proof or mask data |
|
Constant term | Predefined standard quantity, =1 | Fixed offset |
| x | Input variable | User input data | Used for polynomial calculation |
|
Polynomial output | - | - |
|
The model parameters of Participant 1 | Obtained through iterative update of gradient descent | It is a decision variable and will affect the convergence of the model |
|
The model parameters of Participant 2 | Obtained through iterative update of gradient descent | It is a decision variable and will affect the convergence of the model |
|
Regularization parameter | Preset value 0.01 | Control the intensity of regularization |
| Loss | Loss function | Cross-entropy and mean square error | Calculate the gradient in the encrypted state |
Author contributions
D.X. processed the numerical attribute linear programming of communication big data, and the mutual information feature quantity of communication big data numerical attribute was extracted by the cloud extended distributed feature fitting method. J.L. Combined with fuzzy C-means clustering and linear regression analysis, the statistical analysis of big data numerical attribute feature information was carried out, and the associated attribute sample set of communication big data numerical attribute cloud grid distribution was constructed. D.X. and Z.W.R. did the experiments, recorded data, and created manuscripts. All authors read and approved the final manuscript.
Funding
The research is supported by: Scientific research Foundation of the Education Department of Hubei Province, under Grant NO. B2020301; the major project Foundation of philosophy and social sciences research in Hubei Province under Grant NO. 20ZD107; And the Fundation of excellent young and middle-aged scientific and Technological Innovation Team project in Hubei Province under Grant NO. T201943.
Data availability
The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Tang, Y. M., Chau, K. Y., Kuo, W. T. & Liu, X. X. IoT-based information system on cold-chain logistics service quality (ICCLSQ) management in logistics 4.0. Inform. Syst. Front.26(2), 689–708. 10.1007/s10796-023-10393-7 (2024). [Google Scholar]
- 2.Prajapati, P. & Shah, P. A review on secure data deduplication: Cloud storage security issue. J. King Saud Univ. Comput. Inform. Sci.34(7), 3996–4007. 10.1016/j.jksuci.2020.10.021 (2022). [Google Scholar]
- 3.Philsoophian, M., Akhavan, P. & Namvar, M. The mediating role of blockchain technology in improvement of knowledge sharing for supply chain management. Manag. Decis.60(3), 784–805. 10.1108/md-08-2020-1122 (2022). [Google Scholar]
- 4.Bechtsis, D., Tsolakis, N. & Iakovou, E. Data-driven secure, resilient and sustainable supply chains: gaps, opportunities, and a new generalised data sharing and data monetisation framework. Int. J. Prod. Res. 60(14), 4397–4417. 10.1080/00207543.2021.1957506 (2022). [Google Scholar]
- 5.Li, Y. et al. Threats and defenses in the federated learning life cycle: a comprehensive survey and challenges. IEEE Trans. Neural Netw. Learn. Syst. 5(15), 1–21. 10.1109/TNNLS.2025.3563537 (2025). [DOI] [PubMed] [Google Scholar]
- 6.Ramakrishna & Ali Shaik, M. A. Comprehensive analysis of cryptographic algorithms: evaluating security, efficiency, and future challenges. IEEE Access.13 (1), 11576–11593. 10.1109/ACCESS.2024.3518533 (2025). [Google Scholar]
- 7.Mou, J., Zhang, Z., Banerjee, S. & Zhang, Y. Combining semi-tensor product compressed sensing and session keys for low-cost encryption of batch information in WBANs. IEEE Internet Things J.11(20), 33565–33576. 10.1109/JIOT.2024.3429349 (2024). [Google Scholar]
- 8.Lian, Z., Zeng, Q., Wang, W., Gadekallu, T. R. & Su, C. Blockchain-based two-stage federated learning with non-IID data in IoMT system. IEEE Trans. Comput. Soc. Syst.10(4), 1701–1710. 10.1109/TCSS.2022.3216802 (2023). [Google Scholar]
- 9.Azbeg, K., Ouchetto, O. & Andaloussi, S. J. BlockMedCare: A healthcare system based on IoT, Blockchain and IPFS for data management security. Egypt. Inform. J.23(2), 329–343. 10.1016/j.eij.2022.02.004 (2022). [Google Scholar]
- 10.Rabieinejad, E., Yazdinejad, A., Dehghantanha, A. & Srivastava, G. Two-level privacy-preserving framework: federated learning for attack detection in the consumer Internet of Things. IEEE Trans. Consum. Electron.70(1), 4258–4265. 10.1109/TCE.2024.3349490 (2024). [Google Scholar]
- 11.Rao, B. et al. Privacy inference attack and defense in centralized and federated learning: A comprehensive survey. IEEE Trans. Artif. Intell.6(2), 333–353. 10.1109/TAI.2024.3363670 (2025). [Google Scholar]
- 12.Zhu, J., Cao, J., Saxena, D., Jiang, S. & Ferradi, H. Blockchain-empowered federated learning: Challenges, solutions, and future directions. ACM Comput. Surv.55(11), 1–31. 10.1145/3570953 (2023). [Google Scholar]
- 13.Zhu, M., Yuan, J., Wang, G., Xu, Z. & Wei, K. Enhancing collaborative machine learning for security and privacy in federated learning. JTPES44(2), 74–82. 10.53469/jtpes.2024.04(02).11 (2024). [Google Scholar]
- 14.Khan, A. A. et al. BDLT-IoMT-a novel architecture: SVM machine learning for robust and secure data processing in Internet of Medical Things with blockchain cybersecurity. J. Supercomput.81(1), 1–22. 10.1007/s11227-024-06782-7 (2025). [Google Scholar]
- 15.Khan, A. A. et al. Blockchain-enabled infrastructural security solution for serverless consortium fog and edge computing. PeerJ Comput. Sci. 3 (26), 1933–1943. 10.7717/peerj-c.s.1933 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lotfi, R., Rajabzadeh, M., Zamani, A. & Rajabi, M. S. Viable supply chain with vendor-managed inventory approach by considering blockchain, risk and robustness. Ann. Oper. Res.344(2), 575–594. 10.1007/s10479-022-05119-y (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lotfi, R. et al. Resource-constrained time–cost-quality-energy-environment tradeoff problem by considering blockchain technology, risk and robustness: a case study of healthcare project. Environ. Sci. Pollut. Res.29(42), 63560–63576. 10.1007/s11356-022-20334-7 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Lotfi, R. et al. A data-driven robust optimization in viable supply chain network design by considering open innovation and blockchain technology. J. Clean. Prod.436(1), 140369–140381. 10.1016/j.jclepro.2023.140369 (2024). [Google Scholar]
- 19.Hebbi, C. & Mamatha, H. Comprehensive dataset building and recognition of isolated handwritten kannada characters using machine learning models. Artif. Intell.1(3), 179–190. 10.47852/bonviewAIA3202624 (2023). [Google Scholar]
- 20.Ali, A. et al. A novel homomorphic encryption and consortium blockchain-based hybrid deep learning model for industrial internet of medical things. IEEE Trans. Netw. Sci. Eng.10(5), 2402–2418. 10.1109/TNSE.2023.3285070 (2023). [Google Scholar]
- 21.Moore, S. K. Chips to compute with encrypted data are coming: fully homomorphic encryption could make data unhackable. IEEE Spectr.61(1), 38–40. 10.1109/MSPEC.2024.10380468 (2024). [Google Scholar]
- 22.Kumari, K. A., Sharma, A., Chakraborty, C. & Ananya, M. Preserving health care data security and privacy using Carmichael’s theorem-based homomorphic encryption and modified enhanced homomorphic encryption schemes in edge computing systems. Big Data-US10(1), 1–17. 10.1089/big.2021.0012 (2022). [DOI] [PubMed] [Google Scholar]
- 23.Razikin, K. & Soewito, B. Cybersecurity decision support model to designing information technology security system based on risk analysis and cybersecurity framework. Egypt. Inform. J.23(3), 383–404. 10.1016/j.eij.2022.03.001 (2022). [Google Scholar]
- 24.Wang, Y., Zhang, F. & Li, X. Methodology for mapping the ecological security pattern and ecological network in the arid region of Xinjiang, China. Remote Sens. 15(11), 2836–2848. 10.3390/rs15112836 (2023). [Google Scholar]
- 25.Liang, W. et al. PDPChain: A consortium blockchain-based privacy protection scheme for personal data. IEEE Trans. Reliab.72(2), 586–598. 10.1109/TR.2022.3190932 (2023). [Google Scholar]
- 26.Jayabalan, J. & Jeyanthi, N. Scalable blockchain model using off-chain IPFS storage for healthcare data security and privacy. J. Parallel. Distr. Com.164(1), 152–167. 10.1016/j.jpdc.2022.03.009 (2022). [Google Scholar]
- 27.Zhang, Y., Wang, T. & Yuen, K. V. Construction site information decentralized management using blockchain and smart contracts. Comput.-Aided Civ. Infrastruct. Eng.37(11), 1450–1467. 10.1111/mice.12804 (2022). [Google Scholar]
- 28.Alsamhi, S. H., Shvetsov, A. V. & Shvetsova, S. V. Blockchain-empowered security and energy efficiency of drone swarm consensus for environment exploration. IEEE T. Green. Commun.7(1), 328–338. 10.1109/TGCN.2022.3195479 (2023). [Google Scholar]
- 29.Liu, Y., Hao, X. & Ren, W. A blockchain-based decentralized, fair and authenticated information sharing scheme in zero trust internet-of-things. IEEE Trans. Comput.72(2), 501–512. 10.1109/TC.2022.3157996 (2023). [Google Scholar]
- 30.Xu, X. & He, Y. Blockchain application in modern logistics information sharing: A review and case study analysis. Prod. Plan. Control35(9), 886–900. 10.1080/09537287.2022.2058997 (2024). [Google Scholar]
- 31.Al-Shaarani, F. & Gutub, A. Securing matrix counting-based secret-sharing involving crypto steganography. J. King Saud Univ. Comput. Inform. Sci.34(9), 6909–6924. 10.1016/j.jksuci.2021.09.009 (2022). [Google Scholar]
- 32.Zhang, Q., Sun, H. & Gao, X. Time-division ISAC enabled connected automated vehicles cooperation algorithm design and performance evaluation. IEEE J. Sel. Areas Commun.40(7), 2206–2218. 10.1109/JSAC.2022.3155506 (2022). [Google Scholar]
- 33.Mishra, R., Ramesh, D., Edla, D. R. & Qi, L. VaccineChain: A checkpoint assisted scalable blockchain based secure vaccine supply chain with selective revocation. J. Ind. Inf. Integr.34(1), 100485–100500. 10.1016/j.jii.2023.100485 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Mishra, R., Ramesh, D., Mohammad, N. & Mondal, B. Blockchain enabled secure pharmaceutical supply chain framework with traceability: An efficient searchable pharmachain approach. Cluster Comput.27(10), 13621–13641. 10.1007/s10586-024-04626-w (2024). [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.