Development of Blockchain Model with BFT Algorithm for Halal Rating and Review System

The research landscape on blockchain technology and its applications has expanded significantly, covering areas such as food traceability, security, privacy, consensus algorithms, and industry integration. For example, Lin et al. [12] emphasized the importance of consumers' intention to adopt blockchain food traceability technology for enhancing food safety. Similarly, Zhang et al. [13] discussed the significance of security and privacy in blockchain applications for establishing trust in open environments. These studies illustrate the diverse applications of blockchain technology, yet they also highlight the need for further exploration in specific contexts, particularly in the realm of halal certification.

While existing studies have explored blockchain applications in various domains, there is a notable research gap in developing blockchain models that incorporate BFT algorithms for halal rating and review systems. For instance, a study focused on a blockchain-based traceability system for the Indonesian halal supply chain did not specifically address the integration of BFT algorithms for halal rating and review systems [5]. This gap in the literature motivates the proposed study, which aims to enhance the credibility and transparency of halal certification processes through a tailored solution that integrates blockchain technology with BFT algorithms.

In contrast to previous research on blockchain applications in halal traceability, the proposed study uniquely focuses on developing a blockchain model that incorporates BFT algorithms specifically for halal rating and review systems. While prior studies have discussed blockchain for the traceability of Indonesian halal food, the emphasis on BFT algorithms for security and fault tolerance distinctly sets the proposed research apart [1]. Furthermore, a recent study on an optimized BFT algorithm for medical data security underscores the importance of such algorithms in ensuring secure data sharing, thereby reinforcing the relevance of BFT algorithms in halal certification systems [14].

Moreover, the proposed research stands out by concentrating on the integration of BFT algorithms tailored for halal rating and review systems. Although the study [15] has surveyed consensus algorithms in blockchain applications, they have not specifically addressed the development of a model for halal rating systems. By focusing on the unique needs of halal certification processes and leveraging BFT algorithms, the proposed study aims to fill this critical gap in the literature and advance the application of blockchain technology in ensuring halal product integrity.

In summary, the literature review showcases extensive research on blockchain technology across various domains, while simultaneously highlighting a significant gap in the development of blockchain models with BFT algorithms for halal rating and review systems. The proposed study seeks to address this gap by providing a specialized solution that enhances the credibility and transparency of halal certification processes through the innovative integration of blockchain technology and BFT algorithms.

This section presents research results on the development of a blockchain model with BFT algorithm for halal rating and review system. The use of the BFT algorithm aims to ensure data integrity and system resilience.

4.1 Initiation

Based on the literature review, the application of the BFT algorithm in this research consists of a set of nodes that form a blockchain network. Each node has a database of reviews and a set of peer nodes. When a rater submits a new rating and review, it will be proposed to the network, and each node will verify the proposal using the BFT algorithm. If the proposal is valid, it will be added to the review database. The following is the component system for a halal rating and review system that uses the BFT consensus algorithm:

Nodes: A set of nodes that make up the blockchain network, responsible for storing and validating ratings and reviews.

Raters: Users who submit ratings and reviews for halal products/services.

Review database: A decentralized database that stores all ratings and reviews.

In the halal rating and review system, the interaction between nodes, raters, and the review database is crucial for maintaining the integrity and reliability of the ratings through the BFT algorithm. Raters initiate the process by submitting their ratings and reviews for halal products and services, which are then broadcast to the network. Upon receiving these submissions, the nodes—each maintaining a copy of the blockchain—engage in a consensus process governed by the BFT algorithm. This process involves validating the submitted reviews to ensure they meet the system's criteria and are not fraudulent or erroneous. Once a sufficient number of nodes agree on the validity of a review, it is added to the decentralized review database, which is distributed across all nodes. This architecture ensures that all ratings are securely stored and immutable, while the BFT algorithm allows the system to function correctly even if some nodes fail or act maliciously. Thus, the seamless interaction between raters, nodes, and the review database not only enhances the credibility of the halal rating system but also fosters trust among users.

4.2 Conceptual framework development

This section provides information on the development of the system model concept, rating, review and BFT algorithm. System model with n nodes, denoted by N = {1, 2, ..., n}. Each node i ∈ N has a local review database, denoted by D_i. The system has a number of raters, denoted by R = {1, 2, ..., m}, who submit ratings and reviews for halal products. Furthermore, the rating and review model assumes that each rater j ∈ R sends a rating r_j ∈ {1, 2, ..., k} and a review v_j ∈ {0, 1}^l, where k is the maximum rating and l is the review length. Rating r_j represents the halal rating of the product, and review v_j represents the textual review of the product.

The BFT algorithm plays an important role in ensuring the integrity and consistency of the review database. The following is the working principle of the BFT algorithm:

Proposal: Rater j ∈ R sends rating r_j and review v_j to node i ∈ N. Node i creates proposal p = (r_j, v_j, i) and sends it to all other nodes in N.

Verification: Each node k ∈ N verifies the proposal p using the verification function V(p). The verification function checks the validity of the rating and review, and ensures that the proposal is not malicious.

Vote: If proposal p is valid, each node k ∈ N votes v_k ∈ {0, 1} on proposal p. The vote function returns a vote v_k, where v_k = 1 indicates that the node accepts the proposal, and v_k = 0 indicates that the node rejects the proposal.

Consensus: Nodes in N reach consensus on proposal p using the consensus function C(p). The consensus function returns a consensus value c ∈ {0, 1}, where c = 1 indicates that the proposal is accepted, and c = 0 indicates that the proposal is rejected.

Based on the working principle, the modelling of the BFT algorithm can use variables such as in Table 1.

Table 1. Variables of BFT algorithm

The mathematical model of the BFT algorithm goes through five equations. The first equation defines the verification function and checks the validity of the proposal as in Eq. (1).

$V(p)=\{1$ if $p$ is valid, 0 otherwise$\}$                      (1)

The Eq. (1) defines the V function, which takes a proposal p as input and returns 1 if p is valid, and 0 otherwise. In other words, V(p) is a binary indicator of whether p is valid or not. The {0 otherwise} part is a shorthand way of saying "if the condition is not true, then the value is 0". It's a common notation in mathematics and computer science. The second equation defines the vote function, which returns the vote of each node in the proposal as in Eq. (2).

$F(p)=\{1$ if node i receives $p, 0$ otherwise$\}$                   (2)

This Eq. (2) defines the F function, which takes a proposal p as input and returns 1 if node i receives p, and 0 otherwise. The third equation is a consensus function that returns the consensus value of the proposal as in Eq. (3).

$C(p)=\left\{1\right.$ if $\sum[i \in N] F(p) \geq(2 n) / 3,0$ otherwise$\}$                  (3)

This Eq. (3) defines the C function, which takes a proposal p as input and returns 1 if the consensus condition is met, and 0 otherwise. The consensus condition is met when the sum of the F values for all nodes i in N is greater than or equal to (2n)/3, where n is the total number of nodes. The fourth equation affects the local review database, which is the locally collected review data such as Eq. (4).

$D_i=D_i \cup\{p\}$ if $C(p)=1$                        (4)

This Eq. (4) updates the set D_i of proposals accepted by node i. If the consensus condition C(p) is met (i.e., C(p) = 1), then p is added to the set D_i. The fifth equation affects the set of ratings and reviews, which is a collection of data in the form of scores and reviews given by users as Eq. (5).

$R_i=R_i \cup\left\{\left(r_{i_2} v_i\right)\right\}$ if $C(p)=1$                           (5)

This Eq. (5) updates the set R_i of responses received by node i. If the consensus condition C(p) is met, then the response (r_j, v_j) is added to the set R_i. These two equations will apply if the proposal is accepted according to the structure described in Eq. (4) and Eq. (5). As such, both equations have a direct impact on review data management and scoring.

4.3 Technical implementation

Here is a possible mathematical model in Solidity code based on the context provided as shown in Figure 2.

0302.png

Figure 2. Code of mathematical model

This contract defines a Consensus contract that implements the BFT algorithm. The contract has several functions:

Verify: simple verification function that returns true with probability 0.8.

Vote: A simple voting function that returns true with probability 0.7.

Consensus: A function that takes an array of proposals and returns the consensus value (0 or 1).

Simulate: A function that generates proposals from each rater, verifies and votes on each proposal, and returns the consensus value and the number of accepted proposals.

Based on Table 2 the simulation results provide insight into the effectiveness and robustness of the BFT algorithm in achieving consensus among distributed nodes. The primary metrics observed include the consensus value, the acceptance rate of proposals, the occurrence of consensus failures, and the diversity of ratings and reviews provided by raters.

Consensus Achievement, the consensus value reached 1 in 40 out of 50 simulations. This high success rate of 80% demonstrates the BFT algorithm's robustness in achieving agreement among nodes, even in the presence of potential faults or malicious actors. The remaining 10 simulations (20%) where the consensus value was 0 indicate instances where consensus was not achieved. These instances warrant further investigation to understand the underlying causes, which could include node failures, network partitions, or deliberate malicious behavior.

Proposal Acceptance, the majority of accepted proposals were valued at 2, suggesting that the system effectively incorporates the proposals from both raters. This indicates that the BFT algorithm is capable of fairly considering multiple inputs and arriving at a unified decision that reflects the contributions of both raters.

Consensus Failures, in the simulations where the consensus value was 0, the failure to achieve consensus may be attributed to several factors:

Node failures: Unexpected failures of nodes can disrupt the consensus process by reducing the number of participating nodes below the threshold required for agreement.

Malicious behavior: Nodes acting maliciously can send incorrect or misleading information, preventing the system from reaching a correct consensus.

Network issues: Communication delays or partitions can lead to inconsistencies in the information received by different nodes, hindering the consensus process.

Diversity of Ratings and Reviews, the diversity in the ratings and reviews provided by the raters highlights the system's ability to handle a wide range of opinions and assessments. This is crucial for applications that rely on subjective inputs, such as peer reviews, ratings of products or services, and other evaluative scenarios. The system's capacity to integrate these diverse inputs into a coherent consensus demonstrates its flexibility and adaptability in varied environments. Here is a line graph showing the consensus value and accepted proposals for each of the 50 simulations as shown in Figure 3.

0303.png

Figure 3. Graph of consensus value and proposals

The Figure 3 shows the consensus value and accepted proposals over 50 simulations. We can see that the consensus value is always 1, while the accepted proposals fluctuate significantly. This suggests that the system is not able to reach a consensus on the value of the proposal. The fluctuations in the accepted proposals are likely due to the fact that the system is based on a decentralized consensus mechanism. This means that there is no single authority that decides on the value of the proposal. Instead, the value is determined by the agreement of the participants in the system. Because the participants are not always in agreement, the value of the proposal can fluctuate significantly.

4.5 Evaluation

This section elucidates the evaluation and validation of the proposed BFT algorithm designed for rating and reviewing halal products. The effectiveness and reliability of the algorithm were assessed through a series of 50 simulations, focusing on key performance metrics such as fault tolerance and consensus accuracy.

4.5.1 Simulation data and metrics

The evaluation of the proposed BFT algorithm involved analyzing key metrics across 50 simulations as shown in Figure 4.

0304.png

Figure 4. Faulty node of simulations

As illustrated in Figure 4, the graph demonstrates the detection of faulty nodes across simulations, supported by detailed performance metrics from the Python shell output. Despite fluctuations, the system achieves an average detection rate of 49.60% and a 100% success rate across all 50 simulations, reliably identifying at least one faulty node in each run as shown in Figure 5.

0305.png

Figure 5. The evaluation metrics of consensus values

Figure 5 further emphasizes the algorithm's robustness, with the average consensus value consistently reaching 42.00, signifying uniform agreement among nodes. The fault tolerance of 49.60% demonstrates the system's ability to maintain functionality even with nearly half of the nodes compromised. The flawless 100% success rate across all simulations reinforces the algorithm's efficacy in achieving unanimous agreement, making it well-suited for applications demanding high reliability and fault tolerance.

The following is a comparison of the evaluation of other algorithms in several research cases as in Table 3.

Table 3. The performance comparison across several studies

This research shows competitive throughput results of 883.59 tx/s, far exceeding the research of Wei, Runze (7 tx/s) and Wu, Jiangyao et al. (290 tx/s), although still below the research of Oh, Haneul, and Park, Chanik (1997 tx/s). However, the advantage of this research lies in its specific focus on the halal review system, which has unique technical challenges and complexities compared to general blockchain or energy blockchain. With high efficiency in throughput, this research proves that solutions designed for specialized needs can remain competitive without sacrificing reliability, making it a significant contribution in the development of application-specific blockchain technology.

4.5.2 Consensus Value Accuracy calculation

To further substantiate the algorithm's accuracy, we calculated the Consensus Value Accuracy (CA). This metric assesses the alignment between the consensus values generated by the algorithm and the actual values observed in the simulations.

Let C denote the set of consensus values, C={c₁, c₂, …, c_n}, where $n$ represents the number of simulations. Correspondingly, A represents the set of actual values, A ={a₁, a₂, …, a_n}. The set M includes the indices of simulations where the consensus values matched the actual values, M = $\left\{i \mid c_{\mathrm{i}}=a_{\mathrm{i}}, i=1,-2, \ldots, n\right\}$. The Consensus Value Accuracy (CA) is calculated using the Eq. (6).

$C A=\left(\frac{|M|}{n}\right) \times 100$                    (6)

where the Eq. (6), $|M|$ is the number of simulations with matching consensus and actual values, providing a measure of the algorithm's precision in consensus decision-making.

4.5.3 Analysis

The evaluation results indicate that the proposed BFT algorithm performs robustly in scenarios involving a substantial proportion of faulty nodes, with an average faulty node percentage of 49.40%. This level of fault tolerance suggests the algorithm's potential applicability in real-world situations where system integrity may be compromised by unreliable participants. The perfect consensus success rate of 100.00% achieved across all simulations further demonstrates the algorithm's reliability and effectiveness in reaching agreement among nodes, even under adverse conditions. These findings validate the theoretical underpinnings and practical implementation of the BFT algorithm, confirming its suitability for applications requiring high reliability and fault tolerance.

To further bolster the algorithm's robustness, it is recommended to implement additional measures for detecting and mitigating node failures and malicious activities. The high rate of consensus achievement underscores the effectiveness of the BFT algorithm; however, incorporating these measures could enhance its reliability. Continuous system monitoring can play a crucial role in maintaining network health, allowing for the prompt identification and resolution of issues that could lead to consensus failures. Moreover, implementing redundancy and failover mechanisms would improve the overall resilience of the system.

The algorithm's ability to process diverse ratings and reviews is a significant strength, enabling it to handle a wide range of inputs. To sustain this capability, the system must ensure that the algorithms used for aggregating inputs are designed to accommodate variability without introducing bias. This aspect is critical for maintaining the integrity and trustworthiness of the consensus mechanism, especially in diverse and dynamic environments.

Further investigation into the simulations where consensus was not achieved, despite the high success rate, could provide valuable insights into specific failure modes. Analyzing these cases can guide improvements in the BFT algorithm and the overall system architecture, ensuring that it can handle a broader range of scenarios. This continuous improvement process is vital for adapting the algorithm to other domains where consensus mechanisms are critical for maintaining system integrity and trustworthiness.