Blockchain-enabled supply chain finance risk intelligent assessment and trust mechanism construction Authors: Zeyu Chen, Fangfang Zhang Published: 2026 ABSTRACT AbstractSupply Chain Finance (SCF) enhances financial liquidity, optimizes operational efficiency, and strengthens collaboration among supply chain stakeholders. Traditional SCF systems are vulnerable to economic risks, fraud, and a lack of transparency due to centralized data control and limited real-time monitoring capabilities. Current systems rely on slow central financial bodies and static credit evaluations, lacking dynamic risk assessment and mechanisms to build stakeholder trust. This research proposes Bi-MATRA, a Blockchain-Integrated Multi-Agent Trust and Risk Assessment system. The goal is to create a decentralized, intelligent risk evaluation system for real-time trust computation and automated decision-making in SCF contexts. Bi-MATRA combines blockchain’s immutability and transparency with a multi-agent system that automatically monitors and assesses risk. Bi-MATRA uses an adapted Eigen Trust algorithm on the blockchain to calculate trust. Agents rate each other based on interaction history, and transitive trust linkages determine global trust ratings. Smart contracts for dynamic trust-based decision-making update these scores. Smart contracts enforce predefined financial agreements, automating processes and reducing the need for intermediaries. Agent-based modelling was employed to develop a blockchain-based simulation testbed for assessing Bi-MATRA’s responsiveness, accuracy, and resilience. The framework was more efficient at trust validation and risk identification than typical SCF systems. Key studies reveal that Bi-MATRA saves transaction clearance time by 35% and improves early risk detection by 28%. EigenTrust-based trust computation builds agent trust and accountability, enhancing financial interactions. In conclusion, the Bi-MATRA framework offers a scalable, decentralized approach to intelligent risk assessment and trust-building in blockchain-enabled supply chain finance systems.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-53135-w.Keywords: Blockchain, Supply chain finance, Risk assessment, Multi-agent system, Smart contracts, Trust mechanismSubject terms: Engineering, Mathematics and computing Introduction SMEs need supply chain finance in today’s competitive global market. Larger firms may lend financing to startups using SCF’s buyer-provider platform. Supply chain operational efficiency and cash flow improvements save waste and increase funds1,2. SCF allows faster, more transparent, and simpler supplier connections due to globalization and technology3. Pre-decentralization SCF efficacy was greatly diminished. Traditional banks outsource transaction processing and validation4. Centralized administration increases fraud, network issues, and loan approval delays. Business risk levels are not accounted for in current credit ratings5. SCF systems struggle to reduce stakeholder financial risk and overcome system skepticism. Blockchain and AI can overcome these problems6. Since they are decentralized, immutable, and transparent, blockchain ledgers remove middlemen and guarantee data veracity. Multi-agent systems may develop dynamic, real-time financial risk and trust models using software components that monitor, analyze, and decide7. We provide safe, transparent, decentralized, and market-flexible supply chain financing solutions tailored to your needs. Research problem and objectives Financial fraud, inefficiencies, and stakeholder mistrust are frequent in conventional supply chain finance (SCF) systems due to centralized data management, poor transparency, and delayed risk detection8. Due to human error and static credit rating algorithms, these systems exaggerate supply network risk9. SCF ecosystem credit approval delays and fraud cost global enterprises $40 billion yearly, whereas conventional systems average transaction clearing waits of over 6 days. Bi-MATRA-like real-time, automated, and transparent risk processes are needed given these figures. Insufficient decentralized trust hinders financial transactions between participants. A decentralized, intelligent system that can dynamically evaluate risk and compute trust as global supply networks grow more complicated10. Research objective Bi-MATRA enhances accountability and trust by providing immutable, transparent supply chain transaction records via blockchain. Bi-MATRA enhances accountability and trust by providing immutable, transparent supply chain transaction records via blockchain. Autonomous multi-agent systems detect real-time danger by monitoring, analyzing, and assessing financial behaviors without a central authority or manual processes. Autonomous multi-agent systems detect real-time danger by monitoring, analyzing, and assessing financial behaviors without a central authority or manual processes. A dynamic EigenTrust algorithm calculates trust scores based on both direct and indirect interactions, enabling reliable supply chain financing decisions. A dynamic EigenTrust algorithm calculates trust scores based on both direct and indirect interactions, enabling reliable supply chain financing decisions. Methodology Blockchain, multi-agent systems, and trust modeling may solve the issues11. Bi-MATRA (Blockchain-Integrated Multi-Agent Trust and Risk Assessment) records supply chain interactions in a tamper-proof, transparent, decentralized ledger1. Autonomous network agents evaluate stakeholder behavior based on transaction timeliness, contract fulfillment, and financial stability. Dynamic trust scores are calculated using EigenTrust12. Agents evaluate each other directly, and the system calculates global trust ratings from local trust values via transitive links. Bi-MATRA agents use an adaptive trust algorithm to change their behavior and rule-based decision-making for reactive choices. Smart contracts automate financial agreements and risk reduction by storing, updating, and applying trust ratings in real time13. In an agent-based blockchain-based simulation testbed, responsiveness, accuracy, and robustness are compared to standard SCF models14. Explainability methods that show how characteristics affect agent network trust assessments improve interpretability and user trust. In this paradigm, risk assessment and trust propagation are different processes with functional roles. Transaction-level risk assessment that is predictable, interpretable, and rule-based uses weights and thresholds to link operational, financial, and compliance data to a likelihood score. Risk signals must be visible, repeatable, and auditable without data-driven model training. A methodical distribution strategy based on interaction outcomes and agent transitivity linkages updates the trust score as the blockchain verifies events. Unlike adaptive parameter updates based on interaction history and trust propagation dynamics, optimum control based on supervised learning, reinforcement learning, or a predictive model is not adaptable. Thus, the approach’s essential concepts are kept, unrealistic learning ability assumptions are avoided, and the framework is positioned as a practical risk management and trust system rather than a prediction model. Research Contributions are: To present a blockchain-based, multi-agent SCF risk assessment framework for real-time, transparent, and persistent risk assessments. To present a blockchain-based, multi-agent SCF risk assessment framework for real-time, transparent, and persistent risk assessments. To construct an upgraded EigenTrust method on blockchain for dynamic, decentralized trust score calculation based on agents’ behaviour. To construct an upgraded EigenTrust method on blockchain for dynamic, decentralized trust score calculation based on agents’ behaviour. To automate SCF financial agreements, reduce intermediaries, and improve efficiency, transparency, and decision-making accuracy via dynamic trust-based smart contracts. To automate SCF financial agreements, reduce intermediaries, and improve efficiency, transparency, and decision-making accuracy via dynamic trust-based smart contracts. To provide a proactive multi-agent monitoring system that tracks financial activity, stakeholder behaviors, and risk indicators in decentralized SCF networks. To provide a proactive multi-agent monitoring system that tracks financial activity, stakeholder behaviors, and risk indicators in decentralized SCF networks. To evaluate Bi-MATRA’s risk detection, trust validation, and operational efficiency benefits through a blockchain simulation experiment using agent-based models. To evaluate Bi-MATRA’s risk detection, trust validation, and operational efficiency benefits through a blockchain simulation experiment using agent-based models. EigenTrust’s system-level operationalization in a blockchain-enabled, multi-agent Supply Chain Finance (SCF) scenario, and the dynamic relationship between risk and trust in real-time financial decision-making, are two of the study’s distinctive features. Unlike earlier SCF projects that use rule-based logic or EigenTrust as static or offline components, Bi-MATRA directly integrates a modified EigenTrust mechanism into an on-chain, event-driven workflow. When fraud alarms, payment delays, and compliance breaches occur, smart contracts immediately increase confidence. The architectural and operational “intelligence” of the proposed system is achieved by autonomous agents monitoring locally, trust is dynamically recalculated using blockchain-verified interaction histories, and financing decisions are made automatically without centralized intervention. Bi-MATRA prioritizes decentralized trust responsibility, real-time responsiveness, and auditability above predictive model complexity, unlike machine learning-driven credit scoring, system dynamics-based SCF risk analysis, or metaheuristic-optimized classifiers. Therefore, the paper proposes a deployable, integrated SCF trust-and-risk assessment system that turns trust computation beliefs into a realistic, multi-agent, blockchain-native financial infrastructure. Related work This study applies Blockchain, AI, and machine learning to SCF and trust-based systems. Each part covers methodology, dataset, results, constraints, and research needs. Transparency, security, fraud detection, and trust are positives, but simulated data, restricted scalability, and no real-world validation are drawbacks. These results demonstrate the necessity for cross-industry, real-world testing to close research gaps and make SCF technology development more practical. Blockchain-enabled supply chain finance Tsai et al.15 developed a consensus-based, innovative contract algorithm, and blockchain technology was introduced to automate and validate supply chain finance (SCF) financing transactions. Simulated financial transaction records and supplier payment histories assess system performance under various operational conditions. Results reveal that the blockchain-enabled SCF model speeds transaction processing by 25%, reduces delayed payments by 30%, and eliminates double-spending and invoicing fraud. Simulated data may not accurately reflect the complexity and unexpected behavior of real-world SCFs, thereby limiting our ability to model them. Scalability remains a concern for decentralized networks as transaction volume grows. Li et al.16 proposed the Fabric-SCF, a blockchain-based system built on Hyperledger Fabric and an attribute-based access control (ABAC) algorithm, enables secure, fine-grained data access in supply chain finance. The system is assessed for typical SCF operations using simulated financial credit records and transaction logs from small and medium-sized firms. Fabric-SCF offers secure, traceable, immutable data storage with dynamic access control and rapid transaction throughput. Simulation-based validation’s main limitation is that it may not completely account for different, large-scale real-world deployment challenges in heterogeneous supply chains. Li et al.17 improved pandemic monitoring and response via a blockchain-enabled decentralized surveillance and reporting algorithm and real-time smart contracts. With a compartmental SEIR model parameter and a contact history dataset, detection speed, data bias, and reporting delays were improved. Results demonstrate that blockchain technology has reduced estimation errors, enhanced data privacy for data sharing, and improved epidemic control. Synthetic data simulations may not accurately capture the complexities and network scalability of real-world scenarios during unexpected public health crises. Guo et al.18 presented BC4Regu, a blockchain-based consensus technique that increases SCF transparency and coordination, using an information management platform and the Internet of Things. We simulated transaction data and principal-agent scenarios to assess system performance and risk management. BC4Regu streamlines processes, increases data transparency, and facilitates supply chain commerce and finance. Complex and dynamic supply chains have scalability and IoT interoperability issues since these principles are theoretical but not implemented. Bai et al.19 offered a qualitative case study using topic analysis and semi-structured interviews. Workers, managers, and experts from Chinese state-owned firms participated. Interview transcripts were analyzed qualitatively. These studies emphasize framework compatibility and cross-chain compatibility issues, societal concerns about altering business practices, and operational obstacles of utilizing blockchain for SCF and changing organizational operations. This study relies heavily on qualitative research and one-off case studies, hence it may not apply to other nations or industries. Risk intelligent assessment in supply chain finance Liu et al.20 developed a hybrid risk assessment model for construction SMEs’ online supply chain financing utilizing XGBoost, SMOTENC, and Random Forest algorithms. The model beat earlier methods by 6.39% in assessment accuracy and 6.95% in AUC utilizing data from upstream SMEs in China’s building supply chain. Its industry and regional dataset may restrict its applicability to other industries or global financial ecosystems. Li et al.21 used the Heterogeneous Graph Transformer (HGT) approach to identify financial statement fraud in a supplier-customer knowledge network using economic and transaction data from Chinese listed companies. HGT outperformed conventional machine learning approaches by 5.19%, with an AUC of 85.10% and stable performance across historical periods. The main problem is that it uses publicly accessible supplier-customer data, which may not capture private or informal interactions needed to identify fraud. Balmaseda et al.22 recommended GNNs with C2R labeling to forecast financial network systemic risk. Network structure and node-edge attributes are used. On two financial network datasets with inter-entity exposures and risk indicators, GNNs beat conventional ML with C2R, increasing MCC by 94% and 15% and reducing quantile RMSE. In real-world financial systems, high-quality, full network data may be scarce or partial, limiting system capabilities. Wang et al.23 theoretically and qualitatively studied blockchain’s integration into Industry 4.0 without a research methodology or dataset. Blockchain improves sustainability reporting, corporate transparency, and operational accountability, but it also boosts energy consumption, scalability, legal uncertainty, and incentive misalignment. Since these insights require empirical confirmation and quantitative performance evaluation, future research should employ operational models and real-world pilot studies. Trust mechanism construction in blockchain-based systems Pattanayak et al.24 advised 31 supply chain experts and semi-structured interviews for qualitative research. Blockchain technology fosters trust and relational and dynamic capacities in supply chains. This link is moderated by supply chain unpredictability. Qualitative, interview-based research lacks quantitative validation or cross-industry comparison, therefore future research should use empirical, data-driven analysis to generalize its results. Shu et al.25 introduced multi-layer metric algorithms to compute complete trust values for underlying suppliers and blockchain-stored transitive signature receivable warrants, thereby offering a blockchain-based transitive trust model for supply chain finance. Experimental simulations using synthetic SCF transaction data verified the methodology, thereby improving the reliability of trust transfer and access to SME funding. Future studies should address the lack of real-world deployment data and scalability assessments across various supply chain configurations to enhance broader applicability. Jovanovic et al.26 offered blockchain, federated learning, and explainable AI (XAI) for credit rating, using privacy-preserving decentralized models and SHAP and LIME for model explainability. Decentralized models were less transparent, dependable, and explainable than centralized ones in synthetic federated financial dataset experiments that replicate multi-institutional credit systems. The absence of real-world deployment, scalability analysis under financial restrictions, and many data sources for operational validation should be addressed in future study. Jiang et al.27 employed a trust transitivity model for SMMEs in supply chain finance using blockchain-based EPVs to evaluate direct and indirect trust. Simulations of hypothetical multi-tier manufacturing supply chains demonstrated that the approach enhanced trust transfer and alleviated financial constraints for lower-tier suppliers. Future research should include empirical real-world deployment and sensitivity testing under different market and policy settings to broaden applicability. Reinforcement learning (RL) agents for adaptive risk judgments and trust scoring in financial systems were introduced between 2023 and 2025. Cross-chain smart contracts facilitate interoperability among blockchain networks. Trust learning, blockchain, and automation are rarely combined in these studies. To predict the evolution of credit risk in supply chain finance, Li et al.28 created an automated analytical framework based on system dynamics that relies on centralized modelling assumptions and highlights the significance of temporal risk transmission. Although the method is not intended for real-time transactional trust enforcement, Patro et al.29 extended SCF research toward sustainability by putting forth a generative-AI-driven carbon finance indexing framework, showing how advanced analytics can support strategic financial decision-making. Lei et al.30 demonstrated improved forecast accuracy by using machine learning approaches to avoid SCF risk; however, their approach prioritizes offline risk classification over decentralized operational management. Zhang et al.31 introduced a support vector machine optimized using the firefly algorithm for credit risk assessment. Although static datasets and supervised learning are still needed, classification performance has improved. Online SCF risk assessment by Lyu and Zhao32 used compressed sensing in big-data situations to handle high-dimensional data efficiently. However, automated execution and decentralized trust propagation were not explored. These studies emphasize the need for predictive risk modeling and analytics and integrated, blockchain-enabled, real-time trust and risk governance in decentralized supply chain finance systems. Liu et al.33 use smart contracts, reputation models, and blockchain technology to manage trust, scalability, and security in dynamic multi-entity networks. Their research focuses on blockchain-enabled IoT trust management. They suggest supply chain finance ecosystems might benefit from blockchain-based trust-propagation methods like reputation aggregation and transitive trust.Adjei et al.34 introduced a multi-agent reinforcement learning system that uses a graph attention network to find weaknesses in smart contracts. They demonstrated how the robustness and accuracy of detection in decentralized blockchain systems are significantly increased by agent-based learning and graph-aware representations. This article demonstrates how effectively blockchain technology and multi-agent intelligence can automatically detect abnormalities and threats. Furthermore, Bodo and De Filippi35 examined how the regulatory environment affects trust in blockchain and DeFi, and argued that technical trust mechanisms need to be complemented by governance and regulatory congruence to ensure sustained adoption. The aggregate emphasis of these studies on the need to incorporate smart contract security, automated multi-agent analysis, decentralized trust computing, and regulatory considerations strengthens the case for the proposed blockchain-integrated trust and risk assessment paradigm. The paper advances blockchain-based SCF systems by operationalizing trust and risk assessment as event-driven on-chain processes. Previous SCF models compute reputation or risk off-chain or in centralized modules and store the results on the blockchain. Still, Bi-MATRA considers trust propagation as part of the blockchain workflow and distributes evaluations among independent agents without aggregating them. Bi-MATRA continuously refreshes trust based on confirmed transactional events and disseminates it through a structured trust graph, enabling timely, context-sensitive decisions. Current solutions are based on pre-determined policies or periodically renewed reputation ratings that fail to respond to transactional performance in real time. Previous deployment systems focused on auditability or decision support, requiring human or institutional action to respond to trust or risk evaluation. Bi-MATRA links trust scores with smart-contract logic to automatically enforce financing approvals, suspensions, or alerts at execution time. This focused synthesis shows that Bi-MATRA integrates and executes trust-aware SCF processes in real time, rather than proposing new rule sets and reputation algorithms. The research on blockchain technology, machine learning, computational intelligence methods, risk, and trust mechanisms in supply chain finance is presented in Table 1. The vast majority of research uses simulated data, resulting in advancements in transaction speed, trust transfer, and risk assessment accuracy. The lack of real-world implementation, scalability testing, and cross-industry validation is a major issue. Stresses the need for multi-actor, live operational pilots. Fabric-SCF16 and transitive trust models for SCF environments25 are blockchain-enabled supply chain finance systems; however, they employ static rule-based mechanisms, centralized logic, or lack seamless integration with real-time smart contracts. Decentralized multi-agent trust assessment automation is rare. First-ever integration of a distributed multi-agent system for neighborhood trust evaluation into Bi-MATRA. Automatic, real-time financial decisions are enabled via an upgraded EigenTrust algorithm that updates trust ratings based on blockchain-triggered smart contracts and behavioral indications. Bi-MATRA’s design and BI distinguish it from other SCF systems. Tri-layered architecture makes the trust and risk assessment technique continuous, tamper-proof, and behavior-aware. RL agents offered adaptive risk assessments and trust ratings to financial systems in 2023–2025. Smart contracts that cross blockchains improve interoperability. Few of these research mix blockchain with trust learning. Table 1Summary of related work.Ref.No./ authorMethodDatasetKey findingsLimitationResearch gapTsai et al.16Smart Contracts + blockchainSimulated dataFaster processing, reduced delays, no fraudSimulated data, scalabilityReal-world multi-actor testingLi et al.17Hyperledger Fabric + ABACSimulated recordsSecure, fast, dynamic data accessNo large-scale validationCross-industry real-world trialsLi et al.18Blockchain + SEIRSynthetic dataBetter detection, privacy, and speedLacks real crisis dataValidate in live crisis scenariosGuo et al.19Blockchain + IoTSimulated trade dataLower costs, better transparencyNo IoT deployment testedReal IoT-SCF integration pilotsBai et al.20Qualitative Interviews31 interviewsExposed tech and policy barriersSingle-case, no numbersBroader, cross-sector studiesLiu et al.21XGBoost + SMOTENC + RFReal SME data (China)Higher accuracy and AUCIndustry, region-specificTest globally, across different sectorsLi et al.22Heterogeneous Graph TransformerPublic transaction dataHigher fraud detection AUCMisses private dealsInclude informal/private dataBalmaseda et al.23GNN + C2RNetwork dataBig MCC and RMSE gainsNeeds complete network dataHandle partial, missing dataWang et al.24Conceptual StudyNoneHighlights blockchain benefitsNo practical validationRun operational pilot studiesPattanayak et al.25Qualitative Interviews31 expertsBlockchain builds trustNo quantitative dataEmpirical trust modellingShu et al.26Multi-layer Trust MetricsSynthetic SCF dataBetter trust transfer, SME financingNo real deploymentLive multi-chain SCF pilotsJovanovic et al.27Blockchain + Federated Learning + XAISynthetic dataImproved transparency and explainabilityNo real-world testTest in regulated finance settingsJiang et al.28Intuitionistic Fuzzy + Dynamic WeightsSimulated supply chainStronger trust transfer, SME supportNo market data validationReal-market pilots and sensitivity analysis Summary of related work. Bi-MATRA framework for blockchain-based SCF risk and trust management Blockchain-enabled supply chain finance (SCF) operating procedure and Bi-MATRA framework for trust mechanisms and risk assessment are shown in Fig. 1. Trust persistence and transaction-level risk assessment are architecturally distinct. Deterministic, rule-based risk assessment employing operations, finances, and regulations may be repeatable and transparent. Trust propagation resembles a structured aggregate of agent interactions rather than data-driven model training, reinforcement learning, or supervised learning. Systematic risk indicators, weighting algorithms, and blockchain-verified events that automatically update trust make the system more user-friendly. Agent and transaction system behavior audits may be easy. Smart contract, environment, and transaction data are sent to Transaction Monitoring and Data Parsing.Transaction risk scores are calculated using risk-related fields. These risk profiles are provided to a Multi-Agent-Based Local Trust Evaluation module, where agents use EigenTrust to calculate localized trust scores and construct a Local Trust Matrix. Bi-MATRA framework architecture for blockchain-enabled supply chain finance risk assessment and trust management. Next, a Blockchain-Embedded Global Trust Computation module merges local trust computations into a Global Trust Vector for decentralized, tamper-proof trust management. Using this trust vector, a Smart Contract-Based Risk Decision Engine automatically enforces financial and contract regulations based on dynamic trust levels. Finally, risk-based decisions optimize TPE, RTA, CTOR, and DTR. Transaction monitoring and data parsing This module analyzes supply chain financing (SCF) transactions to find patterns. A thorough financial and logistical analysis of supplier-customer interactions helps identify issues early. First, payment status, delivery scheduling, quantity discrepancies, contract status, and environmental data like temperature and humidity are read. Risk-labeling each transaction is automatic using this data. Late deliveries, payments, or smart contract triggers prompt system evaluation. Temperature-sensitive, environmental goods are also assessed. Transaction profiles vary. That profile influences trust and risk in subsequent Bi-MATRA phases. Without human inspections or centralized monitoring, the module automates data extraction and risk labeling for quicker, more consistent decision-making. Risk assessment in field extraction Field extraction extracts Order Status, Payment Status, Time to Delivery, Quantity Mismatch, Fraud Indicator, and Compliance Check from raw SCF data. Quantifying operational and financial behavior underpins risk and trust assessments. In the Bi-MATRA system, each transaction ​ Known as a feature vector: . Where ​ Corresponds to the extracted value of the \:j-th feature (e.g., payment status, delivery delay, mismatch, temperature). Features are encoded numerically (e.g., 1 for overdue, 0 for on-time; scaled values for temperature/humidity). Encapsulates the entire field extraction, feature weighting, risk scoring, and optimization process used in Bi-MATRA for trust and risk assessment: Where in Eq. 1 Feature \:j from transaction \:i (e.g., payment status, fraud flag, delivery time, etc.). ​ Learnable weight for feature \:j.bis known as the Bias term. Predicted risk score. : Actual risk label for transaction \:i. Binary cross-entropy loss to be minimized over all \:m transactions. This equation describes Bi-MATRA’s field processing pipeline, from parsing transaction details to computing trust-based risk scores and optimizing forecasting accuracy. Risk scoring in blockchain-based supply chain finance Blockchain-enabled Supply Chain Finance assigns a Risk Score (\:R) to each transaction using rule-based logic. The score includes delayed delivery (more than 5 days), missed payments, quantity mismatches (more than five units), fraud indicators, and triggered smart contracts. Fraud detection weighs more than other factors in the risk score. Mathematically, the Risk Score \:R for a transaction can be defined as: Where in Eq. 2\:R is known as Total Risk Score,is known as Delayed Delivery Indicator (1 if Time to Delivery > 5, else 0), \:O is defined as Overdue Payment Indicator (1 if Payment Status is Overdue, else 0), \:Q is represented as Quantity Mismatch Indicator (1 if mismatch > 5 units, else 0), \:F is kown as Fraud Indicator (1 if Fraud = 1, else 0),is known as Smart Contract Trigger Indicator (1 if status = Triggered, else 0), is defined as the Weight coefficients, where ​=2 (higher fraud weight), and others = 1. Additionally, environmental compliance is flagged if: . triggers a cold chain or moisture alert. An automated decision engine for trust management called Bi-MATRA receives its input from this logic’s risk score \:R, which incorporates signals from operations, finances, and the environment. The Bi-MATRA uses transaction data to monitor risks in blockchain-based supply chain financing, as shown in Fig. 2. Current fraud, compliance, payment status, and estimated delivery dates are key data elements. This data is arranged for analysis. The program assesses transaction risk using this data. It detects suspicious activity, smart contract triggers, late payments, odd volumes, and delivery delays. For delicate things, the gadget monitors humidity and temperature and alerts the user if they exceed restrictions. The Bi-MATRA trust engine utilizes all risk signals and alarms to assess trustworthiness and financial viability. Automatic, consistent, and dispersed risk assessments enhance dependability and efficiency without requiring human intervention. Bi-MATRA transaction monitoring and risk scoring flow. The Bi-MATRA paradigm combines rule-based risk scoring and learning-based trust propagation into a single pipeline, describing their interaction and distinguishing them. Rule-based systems can only calculate transparent transaction risk scores utilizing operational, financial, and compliance aspects. This trust assessment layer uses deterministic ratings as structured inputs but does not learn from them.Learning and adaptation only occur at the trust computation step, which combines past interactions and transitive linkages between agents to generate global trust estimates. Let a transaction be represented by a feature vector Connecting each element to a defined signal like payment status, delivery delay, fraud warning, or compliance result. Equation (3) calculates each transaction’s risk score using a weighted linear rule-based function. The fixed significance weight of the \:j -th indicator is . Risk score has no effect on the decision-learning process’s reinforcement or punishment of agent interactions to manage local trust updates. The business-wide linear EigenTrust/PageRank approach is utilized by Bi-MATRA to represent trust. Equation (4) is in the normalized local trust matrix and Eq. (5) calculates the global trust vector periodically. It is the cumulative positive feedback from agent i to agent j, adjusted for transaction risk (greater risk means less ). The damping factor and the pre-trusted vector p is used. Using the following symbols: for transaction feature vectors, for rule-based risk weights, \:C for the normalized local trust matrix, t for the global trust vector, and for the damping parameter, this linear formulation ensures a unique stationary trust vector under standard stochasticity conditions. It avoids exponential or ambiguous update expressions. Transaction risk scoring and logistic alert evaluation. Multi-agent-based local trust evaluation Bi-MATRA for Blockchain-enabled Supply Chain Finance deploys software agents to suppliers and customers for localized trust assessment. These agents employ a modified EigenTrust algorithm to assess counterparties, transaction integrity, and behavior. Agents build global trust on local trust ratings, which are impacted by past encounters and peer feedback. Decentralized assessment reduces fraud, improves transparency, and enables dynamic risk assessment. Real-time trust evaluations without centralized control are possible with the multi-agent architecture. Improves SCF ecosystem resilience, accountability, and efficiency while reducing financial and regulatory concerns. Events like delayed payments, non-compliance, and fraud flags prompt reactive, rule-based Bi-MATRA agents to execute predetermined logical processes. An updated EigenTrust algorithm changes trust ratings based on earlier interactions and indirect trust propagation to evaluate agent trust. Deep reinforcement learning and supervised classifiers are not yet implemented, but the system is meant to support future anomaly detection and risk categorization using machine learning. The Bi-MATRA’s blockchain-enabled supply chain finance trust evaluation procedure. The Trust Evaluation Process utilizes fraud, compliance, payment, and delivery indicators derived from Trust Chain Transaction Data, as shown in Fig. 3. The trust matrix is generated, normalized, and built from local trust values. These help software agents (suppliers and customers) calculate the Eigen Trust global trust scores. Transactional data is transparent and secure because decentralized agents use an immutable blockchain ledger. This dynamic evaluation improves supply chain ecosystem risk assessment, compliance monitoring, and trust-based decision-making. Trust evaluation flow in the Bi-MATRA system using eigen trust. Trust matrix construction using weighted EigenTrust in blockchain-enabled SCF In the Bi-MATRA framework, each agent ​ (supplier or customer) Computes a local trust score for every interacting agent ​, forming a trust matrix , where represents the normalized trust agent ​ places in the agent . Behavioural measurements dynamically updated from transactional features stored on the blockchain are the Source of this trust. Features of transactions form the basis of each trust increment: Where in Eq. (6) is the Fraud Indicator (0 = no fraud), is known as Compliance Check (0 = compliant), Payment Status (1 = paid), Timely Delivery (1 = < 5 days), are tunable trust weights. Then, trust scores are normalized to produce the trust matrix: , . To compute global trust \:t, the adapted EigenTrust algorithm is applied: Where Global trust vector, damping factor (e.g., 0.15), \:e pre-trusted peers vector, or uniform distribution. A recursive computation employs agents’ direct and bidirectional trust to guarantee unchangeable assessments. Smart contracts rapidly update ratings following blockchain events like delivery, compliance, and payment status, improving risk-conscious decision-making and latency. Blockchain-embedded global trust computation (EigenTrust adaptation) The decentralized, dynamic Bi-MATRA system evaluates supply chain financial trust using Blockchain and EigenTrust. Suppliers and purchasers trust blockchain records of prior contacts, which include payment timeliness, fraud, and compliance. Transitive propagation creates a global trust vector from local evaluations to locate trustworthy people. Smart contracts can update trust-based actions and scores instantly. This strategy reduces dependence on a single entity and improves accountability, improving early danger identification and stakeholder confidence in supply chain finance systems. Since it employs an iterative, deterministic EigenTrust algorithm, the Bi-MATRA trust calculation mechanism struggles with dynamic learning induced by environmental or agent behavior changes. Deterministic rule-based trust updates are used by Bi-MATRA agents, although reinforcement learning or Bayesian trust optimization may improve their adaptive decision-making. Following these stages may help agents enhance their decision-making in hostile or unclear settings, deal with context-specific risk patterns, and update their trust norms. Trust and risk assessment modules use adaptive learning to improve deterministic rule-based approaches36. Agents use a Beta distribution updated online based on real-time transaction outcomes including payment timeliness, compliance fulfillment, and fraud occurrence to measure local trust. This distribution allows a Bayesian trust updating method to estimate the counterparty’s reliability.By allowing agents to learn from the past, account for uncertainty, and dynamically adjust trust levels in response to changing risk scenarios, the probabilistic update provides data-driven adaptability to the deterministic EigenTrust propagation. Labeled datasets and centralized training pipelines are needed for supervised learning models like XGBoost and deep neural networks in decentralized supply chain finance. However, the hybrid technique may work online and unsupervised without ground-truth labeling.In multi-agent blockchain settings, reinforcement learning-based trust optimization has high computational overhead, delayed convergence, and limited explainability. Bayesian–EigenTrust integration has faster convergence, lower complexity, and stable trust evolution compatible with smart contract execution constraints. Evidence shows that learning-augmented trust updating improves early risk sensitivity and adaptive trust adjustment while maintaining transparency, scalability, and robustness, making the proposed framework an automated risk assessment and event-driven updates trust mechanism. In Fig. 4, the Bi-MATRA trust mechanism is summarized. Transaction behavioral indicators are standardized into trust ratings using the stepwise EigenTrust-based local trust assessment in panel (a). The blockchain immutably stores the global trust vector, which is iteratively merged from various agents’ local trust scores, as shown in Panel (b). Abstraction preserves trust transmission and decision-making while reducing complexity. Figure 4 shows how Bi-MATRA derives local trust scores between agents using transaction histories, fraud indications, compliance status, and payment fulfillment. Highly rated agents are trustworthy. Damping factors balance personal trust values with pre-trusted peers’ influence in EigenTrust’s trust vector updating process. Secure, intelligent decision-making is possible by blockchain-based supply chain finance systems’ decentralized, adaptive, and transparent trust management. Repeated updates stabilize global trust vector t. If “if I trust A, and A trusts B, then I might trust B” for new suppliers/customers, trust can be automatically inferred. Detects fraud and dependable agents without direct contact. See Table 2. Illustration of the (a) EigenTrust-based trust evaluation and (b) global trust score computation in the Bi-MATRA framework. Table 2Agent trust and compliance assessment.AgentVerified? (for p)Fraud?CompliancePaymentS4Yes010S6No100S8Yes011 Agent trust and compliance assessment. Global trust scores enable Bi-MATRA smart contracts to make objective, automated supply chain financing choices. Agent trust score determines invoice financing or trade credit approval. A threshold is exceeded. Low trust transactions are refused or reported for manual or automated review. Smart contracts update agents’ creditworthiness based on blockchain-logged trust. Decentralised, trust-driven automation eliminates manipulation and single-point failures, providing a fairer, more robust supply chain finance environment. Step 4: on-chain trust update Bi-MATRA updates and securely stores trust scores on the blockchain to ensure transparency and reliability. The trust score tt is calculated using an adapted EigenTrust algorithm: . component \:a balances pre-trusted interactions(\:P), while component records agent feedback. After the scores level, this is repeated. Each blockchain transaction has a trust-based hashtag. Smart contracts may use these ratings to automatically approve or reject financial circumstances and identify suspicious behavior. The lack of confidence in instance TX00003 prompted a fraud notice and cautious contract response.Bi-MATRA is more secure and efficient than typical supply chain finance systems since its trust decisions are automated, tamper-proof, and auditable. EigenTrust provides scalable, decentralized trust propagation, but its output may be misleading to stakeholders without technical knowledge. Our suggestion for improving openness when autonomous AI decision-making is not available is to use SHAP and LIME from explainable AI (XAI) to understand EigenTrust algorithm trust ratings. Agents and auditors may use these methods to find out how trust ratings are affected by non-compliance and late payments. Fuzzy rule-based logic applied to conventional language rules (such as “If payment is delayed and fraud is flagged, then trust is low”) may improve trust evaluations. We seek to explain automatic trust in decentralized systems with these changes. Transactions influence Bi-MATRA trust ratings due to compliance concerns, payment delays, and fraud alerts. SHAP and LIME help non-technical stakeholders grasp these implications. The example “if payment is delayed and fraud is flagged, then trust is low.” shows that fuzzy rule explanations may be understood. These methods give openness and practicality for trust assessments, although user explainability may be improved. Smart contract-based risk decision engine The Bi-MATRA system automates Supply Chain Finance (SCF) financing decisions utilizing EigenTrust-based trust and risk scores. Smart contracts can review and execute financial agreements in real-time, enhancing speed, transparency, and reliability. Trust scores are calculated using the following recursive formula: Where: The normalized trust matrix (C) is based on agent feedback (interaction history). The damping factor is usually 0.15. The vector of initial trust anchors (pre-trusted agent values) is . Each transaction generates a new trust score , which is recorded on-chain via smart contracts and related to the Transaction Hash. Smart contract decision logic Smart contracts reference updated trust scores (t) and risk criteria (e.g., fraud indications, compliance check flags). It assigns a contract status using the reasoning below: In Bi-MATRA, TX00002 has a trust score. With , no fraud detected, and successful delivery and payment, the Smart Contract Status is “Completed”. While TX00003 failed compliance tests and had a fraud indication of 1, it was automatically classified as “Triggered”. TX00005’s excellent trust score, 0% fraud, and perfect compliance earn it a “Active” rating, suggesting risk decision rules approval for funding. Details of the baseline implementation are supplied to assure comparison assessment impartiality, repeatability, and openness. All baseline systems were developed using the same dataset, simulation environment, and assessment parameters as Bi-MATRA, without conducting any additional experiments. The Fabric-SCF baseline16 was developed using a permissioned blockchain with smart contracts based on static rules, taking into account compliance flags, order quantity, delivery status, and payment status. The simulation was executed more than thirty times. The XGBoost–Random Forest baseline20 employed offline-trained machine learning models, using features such as payment delay, quantity mismatch, order value, historical default signals, and compliance status, in the absence of blockchain or decentralized trust mechanisms. The outcomes were cross-validated ten times and averaged over 20 runs. Without multi-agent autonomy or dynamic risk assessment, the blockchain-based transitive trust model25 evaluated trust across 25 iterations using static direct and indirect transaction histories. To simulate a traditional bank-led SCF system, centralized credit thresholds and manual verification delays were used, based on industry averages. Each model was evaluated using the same agent populations, transaction volumes, network conditions, performance indicators (TPE, RTA, CTOR, and DTR), and fixed random seeds to ensure an equitable and reproducible comparison. Result analysis All of the quantifiable performance metrics displayed here come from extensively monitored blockchain-based simulation tests. Bi-MATRA decreased the average transaction clearance time in our simulations by over 35%, from 4.0 days (Fabric-SCF) and 6.2 days (bank-led SCF baseline) to 2.6 days. This increased early risk identification by 28% from 68.2% (traditional SCF) to 87.4% (Fabric-SCF). Trust propagation converged in 12–30 iterations for networks with 100–5,000 agents, maintaining over 93% trust score accuracy. These findings show Bi-MATRA’s performance in simulations, albeit outcomes may vary by participant, network, and regulatory situation. Dataset source information The simulation was strengthened by using a synthetic dataset obtained from publicly accessible blockchain-enabled supply chain financing components. Humidity, temperature, transaction and supplier/customer IDs, payment status, delivery delays, fraud indications, and compliance checks are inputs. Trust and risk evaluations increase when anonymized, artificially created datasets match real-world SCF dynamics. Table 3 of the Blockchain-Enabled Supply Chain Dataset contains transaction and operation data37. This data shows how blockchain technology may improve supply chain visibility, risk reduction, and operational analytics. The well-documented supply chain contains time stamps, transaction identifiers, item and supplier IDs, humidity and temperature, and GPS positions. Operations, finances, payment procedures, order totals, shipment tracking numbers, and delivery dates are all recorded. The blockchain’s inherent compliance checks for contract statuses allow trust verification, fraud prevention, and transaction hashing in real-time supply chain financing. Graph anomaly detection and blockchain audit analytics might benefit from this collection. Using blockchain-enabled, agent-based SCF test environments, we created a synthetic, simulation-based dataset. The public Kaggle Blockchain-Enabled Supply Chain Dataset37 provided structure instead of transaction data. No financial or operational papers exist in the collection. We used programming to simulate smart contract interactions to build hypothetical transactions. The large database of simulated records includes smart contract status, environmental parameters like humidity and temperature, delivery time, payment status, fraud indicators, amounts that don’t match, and supplier and customer IDs. Features were dispersed using controlled probability distributions, rule-based triggers, and temporal limitations to imitate SCF characteristics from the literature. Partially completed orders, late payments, compliance concerns, and fraud may arise. The dataset shows real-world SCF systems’ structural, transactional, and behavioral aspects, including as trust transfer, risk escalation, and clearing time volatility. Lack of financial data precludes operational validation, real-world financial deployment, and transaction performance. Here, SCF algorithms demonstrate their promise and efficiency. Table 3Illustration of dataset description.AttributeDescriptionTransaction IDUnique identifier for each transactionTimestampThe time at which the transaction occurredItem IDUnique product/item identifier involved in the transactionSupplier IDIdentifier for the supplier executing the transactionCustomer IDIdentifier for the customer receiving the transactionLocationThe city or region where the transaction takes placeTemperatureRecorded temperature during transaction or shipmentHumidityRecorded humidity during the transaction or shipmentGPS coordinatesExact geographical coordinates for the transaction eventOrder amountTotal value of the orderQuantity shippedNumber of units/products shipped in the transactionOrder statusCurrent status of the order (e.g., Shipped, Delivered, Pending)Payment statusPayment completion status for the transaction (Paid, Overdue)Time to deliveryNumber of days it took to deliver the shipment.Quantity mismatchDifference between the quantity ordered and the quantity shippedFraud indicatorBinary indicator flagging potential fraud (1 = suspicious, 0 = clean)Transaction hashA blockchain-generated unique transaction hash for integrity verificationSmart contract statusStatus of the associated smart contract (Active, Completed, Triggered)Compliance checkWhether the transaction passed (1) or failed (0), the blockchain compliance protocols Illustration of dataset description. Decentralized supply chain finance systems require trust convergence, risk detection accuracy, transaction efficiency, and resilience to adversarial behavior, which motivates the proposed evaluation metrics, TPE, RTA, CTOR, and DTR. TPE shows that dynamic multi-agent networks consistently reach trust equilibrium by computing the inter-iteration trust variance and measuring the EigenTrust-based global trust vector’s stability and convergence rate. RTA uses statistical detection theory to balance fraud and compliance detection by accounting for true positives, false positives, and false negatives while analyzing smart contract-driven risk activation. Operational efficiency study drives CTOR, which compares blockchain automation to conventional SCF transaction clearing time. DTR, based on distributed system resilience theory, shows decentralized trust computing’s resilience by assessing trust loss and recovery after hostile or uncommon occurrences. To ensure statistical validity, all metrics were examined using randomized interaction sequences across 30 simulation runs. The mean, standard deviation, and 95% CIs were determined for each statistic. A paired two-tailed t-test comparing baseline and Bi-MATRA models confirmed substantial performance enhancements (p < 0.01 for all improvements). Wilcoxon signed-rank tests confirmed these findings and reduced distributional bias. To do a sensitivity analysis, increase the network’s agents from 100 to 5,000 and the percentage of hostile or non-compliant agents from 10% to 40%. TPE and DTR decrease as network stress increases, whereas RTA and CTOR stay constant within reasonable confidence ranges. Thus, the suggested measurements are scalable and valid across agent behavior types and network sizes. Trust propagation efficiency (TPE) Trust Spreading Efficiency affects blockchain-enabled SCF network agents by measuring how well the customized EigenTrust algorithm distributes trust ratings. This metric demonstrates how effectively the system depicts agents’ real-time activities, highlighting trustworthy ones and downplaying malicious ones.The Bi-MATRA system ensures decentralized network security and integrity with a high TPE and accurate and dynamic trust environment. EigenTrust measures how well the decentralized multi-agent network distributes trust ratings. Agents swiftly and reliably build global trust from local encounters. Where I Eq. (7), ​ The trust score of agent \:i at iteration.\:N is the number of agents. Lower TPE indicates faster convergence and better trust propagation. There are 100 buyer agents and 500 supplier agents in the simulation. In terms of value, transactions fall into two categories: 70% are of low value (<$5,000), and 30% are of high value (> $20,000). To test the system under various structural assumptions, we created models of both centralized (hub-and-spoke) and decentralized (peer-to-peer) network topologies. Decentralized, blockchain-enabled supply chain finance (SCF) networks enhance trust accuracy, as illustrated in Fig. 5, which plots the Global Trust Score Accuracy (%) across agent interactions to evaluate Trust Propagation Efficiency. The algorithm for TPE calculates the average trust score fluctuation over iterations defined from Eq. 7: Where ​, and represent agent i’s trust scores at iterations t and t−1, where N is the number of agents. Due to unstable trust scores , TPE is strong at first, but as interactions increase, reliable agents receive positive feedback and malicious ones are penalized, reducing score variance , TPE. Fast trust convergence outperforms Fabric-SCF16, XGBoost-RF20, and Blockchain-based Transitive Trust25, which propagate more slowly and exhibit worse steady-state accuracy. Bi-MATRA’s EigenTrust technology rapidly stabilizes network trust, minimizing TPE early and leading trust accuracy across all interaction levels. We found that Bi-MATRA’s practical and adaptive trust propagation in SCF networks creates a durable, decentralized trust environment. Table 4 compares baseline selection to learning-based SCF risk models such Bayesian networks and GNNs9,22. Previous studies using GNNs to detect fraud or predict systemic risk have mostly focused on the design of financial networks. Such models frequently lack the entire and high-quality network data needed for genuine SCF situations, particularly for private or unauthorized transactions. Bayesian networks also reason probabilistically. They cannot respond instantly to agents’ changing actions because they employ past data and conditional dependencies. Comparative trust score accuracy of different models across increasing agent interactions. Table 4Comparative analysis of SCF risk assessment approaches.ApproachData requirementReal-time adaptationDecentralizationML componentLimitationBi-MATRA (proposed)Transaction + compliance + environmental indicatorsYes (EigenTrust + smart contracts)Fully decentralizedMinimal (rule-based trust evaluation)No supervised ML, relies on transaction data qualityGNN-based models22,23Complete financial network dataLimited, retraining requiredCentralized/partially decentralizedGNNRequires full network data, poor handling of private/informal transactionsBayesian network models9Predefined conditional dependenciesModerateCentralizedProbabilistic inferenceRequires historical data and domain knowledge, with limited real-time adaptationXGBoost + RF21Historical transaction featuresLimited, offline trainingCentralizedSupervised MLA static model does not capture transitive trust or automated enforcement Comparative analysis of SCF risk assessment approaches. Risk trigger accuracy (RTA) Risk trigger precision shows how effectively smart contracts and machine learning algorithms identify fraudulent transactions. Good and bad outcomes are measured by the percentage of true positives, or appropriately detected threats.RTA validates Bi-MATRA’s BIAS (Blockchain-Integrated Anomaly Score) mechanism and smart contract-based compliance checks to detect fraud and compliance violations without overwhelming the system with notifications. RTA evaluates Bi-MATRA’s risky transaction detection using legal checks, fraud indicators, and automated decision outcomes based on smart contracts. The SCF (Fabric-SCF16, XGBoost-RF20, and Blockchain-based Transitive Trust25 trust and risk evaluation systems’ average response times with time, confirming Risk Trigger Accuracy (RTA) in Fig. 6. RTA compares true positives (risky transactions accurately recognized) against false positives (valid transactions wrongly reported) and false negatives is given by Eq. (8). Risk trigger accuracy (RTA) performance across smart contract frameworks over time. Where \:TP is known as True Positives (correctly identified risky or fraudulent transactions), \:FP is known as False Positives (everyday transactions incorrectly flagged as dangerous), \:FN is defined as False Negatives (risky transactions missed by the system). Decentralized supply chain financing requires rapid fraud and compliance detection to ensure operational security. Smart contracts and algorithms accelerate compliance and risk evaluations, reducing undetected irregularities. According to the RTA formula, effective systems have a high True Positive (TP) ratio and low False Positives (FP) and False Negatives. Bi-MATRA detects threats in real time with the quickest response speeds. Reactive smart contracts use BIAS to reduce system overload and enable dynamic compliance. Bi-MATRA’s timeliness and greater RTA allow it to discover faults without flooding the network. Clearance time optimization rate (CTOR) The optimization rate of clearing time measures how much smart contract automation of SCF choices speeds transaction clearance. Traditional SCF systems’ decentralized intermediates are included when comparing Bi-MATRA’s processing speed. Intelligent trust and risk procedures boost supply chain efficiency and throughput, as evidenced by a greater CTOR. The automatic contract automation and real-time trust evaluation in Eq. (9) let CTOR evaluate how much quicker Bi-MATRA clears transactions than standard SCF systems. Where ​ Average transaction time in a conventional SCF system, Average transaction time under Bi-MATRA. The Clearance Time Optimization Rate (CTOR) quantifies the percentage improvement in transaction clearance time achieved through automated intelligent contract decision-making in blockchain-enabled SCF systems, defined from Eq. 9 where Traditional​ is the average clearance time of baseline systems (e.g., Fabric-SCF, XGBoost-RF, and Blockchain Trust), and Represents Bi-MATRA’s optimized clearance time. As illustrated in Fig. 7, Bi-MATRA consistently demonstrates superior performance with a CTOR of approximately 35% across transaction volumes ranging from 100 to 5000, compared to 15% for Fabric-SCF, 25% for XGBoost-RF, and 30% for Blockchain Trust. The outcomes of Bi-MATRA were contrasted with those of two reference systems: (1) an implementation of Fabric-SCF16 and (2) a model of bank-led SCF that relied on conventional clearance procedures. The existing model took an average of 6.2 days for clearance, whereas Fabric-SCF took 4.0 days, and Bi-MATRA only took 2.6 days. Also, compared to 68.2% and 59.1%, Bi-MATRA achieved a risk detection rate of 87.4%. This improvement is calculated from Bi-MATRA’s integration of real-time trust computation. , where is defined by the Change in trust score for agent i between two consecutive iterations (or transactions), : Time interval or number of interactions between updates, multi-agent consensus, and adaptive brilliant contract execution, driven by its EigenTrust-based trust propagation and the BIAS risk evaluation mechanism. The constantly increasing CTOR proves Bi-MATRA’s effectiveness in decreasing latency, automating compliance checks, and expediting clearing procedures outperforms conventional and semi-intelligent SCF systems in scale and speed. Clearance time optimization rate (CTOR) performance comparison of SCF models. Decentralized trust resilience (DTR) Decentralized Trust Resilience assesses the system’s capacity to sustain trust assessments and secure operations despite hostile agents, system failures, or inadequate data. This statistic evaluates the Bi-MATRA algorithm’s stress-resilience and trust rating adaptation. A high DTR suggests that the blockchain-integrated multi-agent method may enable sustainable and transparent SCF ecosystems even under turbulent settings, as in Eq. (10). Where ​ The trust score before disruption, The trust score after recovery,, is the number of update phases that return to the previous trust spectrum, and a larger DTR indicates more resilience and faster restoration. Table 5: 100–5,000 agent Bi-MATRA scalability metrics. Despite increasing contact volume, convergence is under 30 iterations, trust accuracy reduces somewhat, and latency is realistic. Bi-MATRA is load-resistant and scalable. Table 5Bi-MATRA scalability evaluation under increasing agent counts.Metric100 agents500 agents1,000 agents2,500 Agents5,000 agentsGlobal trust convergence time12 iterations18 iterations22 iterations26 iterations30 iterationsAverage trust score accuracy94.8%94.2%93.7%93.4%93.1%Communication overheadLowLowModerateModerateModeratePeak memory usage12 MB38 MB70 MB115 MB148 MBSmart contract trigger latency0.5 s0.7 s1.1 s1.5 s1.9 sStatusStableStableStableStableStable Bi-MATRA scalability evaluation under increasing agent counts. The Decentralized Trust Resilience (DTR) statistic measures a decentralized SCF ecosystem’s capacity to swiftly recover from trust shocks and reach trust equilibrium. Based on DTR values from transactions (\:TX00001-TX00009), Bi-MATRA is more robust than Fabric-SCF16, XGBoost-RF20, and Block chain-based Transitive Trust25. Equation 10 calculates DTR precisely. and indicate an agent’s trustworthiness before and after a disturbance like fraud or an oddity. After the disruption, many cycles or actions (\:R) stabilized or restored the trust score. A tiny pre- and post-trust time gap indicates a strong DTR. Graph transactions show negligible trust erosion and quick convergence to stability with DTR values of 0.49 (Fig. 8). Decentralized trust resilience (DTR) performance across SCF transactions. Limitation of simulation Despite improving operating efficiency, trust propagation, and risk detection in simulated SCF situations, Bi-MATRA ignores real-world aspects when assessed. Simulating decentralized supply chain restrictions, communication failures, and network slowness using fake data may influence real-time trust and risk estimations. Since technology cannot replicate regulatory operations and compliance standards, the framework may need human intervention in many policy and legal concerns, which may delay or necessitate overrides. Inconsistent, imprecise, or delayed oracle data might hamper automated financial choices, even with quick smart contracts and oracle confidence. Existing SCF networks do not employ human interventions or overrides in contested or high-risk transactions, which might degrade system performance. In addition to sophisticated risk assessment and trust computation, Bi-MATRA needs additional testing for live, multi-actor SCF datasets, real-time blockchain networks, and operational limits. The present assessment models simulate SCF system operational, transactional, and environmental aspects such payment delays, fraud, irregular deliveries, and compliance using a synthetic dataset (Bi-MATRA)37. From publicly accessible SCF operating data and industry standards, all parameter values, distributions, and interdependencies were calculated. According to sensitivity research, transaction volume, fraud frequency, and agent interaction patterns will not affect the current performance trends of 35% faster clearing and 28% better risk identification. The approaches employed make this a proof of concept rather than a deployment validation. Future research may address deployment problems, improve realism, and apply their results by using rule-based synthetic data, anonymized or semi-real SCF datasets, or both. Conclusion and future enhancement Traditional supply chain finance (SCF) systems use Bi-MATRA to manage fraud, integrity, and financial risk. Bi-MATRA creates a trustworthy, secure SCF platform using blockchain technology’s immutability, scalability, intelligent agent-based trust computation, and enhanced EigenTrust. This technique streamlines operations using smart contract transactions, clearance optimization, and real-time dynamic trust evaluations. Bi-MATRA outperformed standard SCF methods in early risk assessment by 28% and transaction clearance by 35%. The design maintains high Decentralized Trust Resilience (DTR) to prevent destructive shocks and anomalies in uncertain times. Bi-MATRA may assist autonomous SCF networks become more transparent, trustworthy, and responsible financially. Suppliers can expedite credit acquisition with real-time trust ratings, while lenders may decrease risk with automated fraud detection and dynamic creditworthiness updates. Blockchain transactions’ immutability and auditability help auditors and regulators stay compliant. Bi-MATRA’s technology helps real-world SCF ecosystems achieve trust, efficiency, and accountability. Bi-MATRA now evaluates trust using adaptive criteria. Future anomaly detection powered by machine learning might increase risk and fraud detection. IoT sensors like humidity, temperature, and real-time GPS tracking may be coupled with transaction data to enhance context-aware trust judgments. The multi-agent system that generates dynamic credit ratings may improve financial choices using statistical analysis. Finally, Bi-MATRA might be linked to Hyperledger Fabric or Ethereum private networks to demonstrate its security, interoperability, and capabilities in real-world settings. Clustering-based outlier identification, Isolation Forests, and other lightweight anomaly detection methods may improve intelligent risk assessment. These technologies will boost trust and flexibility by detecting fraud before it happens, even with unlabeled data. To make Bi-MATRA fully flexible, we will employ Bayesian networks for probabilistic reasoning and reinforcement learning for trust determination criteria. In environmentally conscious and risk-aware SCF contexts, the system earns value by adopting trust behaviors. This makes it more reliable and suitable for complex supply chain finance situations. Bi-MATRA will soon audit EigenTrust’s trust judgments utilizing XAI technologies including SHAP values, LIME, and fuzzy inference methods. Trust is needed for agent-based autonomous financial systems to meet regulations. Combining synthetic testing with certified real-world SCF datasets from industry partners or anonymized data improves sector-specific outcomes and operations. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions Both authors contributed equally to all aspects of this research. All authors reviewed the manuscript. Funding The project has not received any funding support. Data availability All data that support the findings of this study are included within the article. Code availability The complete implementation of the proposed Bi-MATRA framework is publicly available on Zenodo (DOI: 10.5281/zenodo.19275521). The repository includes all source code, execution scripts, dependency files, and sample outputs required to reproduce the reported results. Detailed instructions for dataset acquisition and execution are also provided to ensure full reproducibility. 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 Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement All data that support the findings of this study are included within the article. The complete implementation of the proposed Bi-MATRA framework is publicly available on Zenodo (DOI: 10.5281/zenodo.19275521). The repository includes all source code, execution scripts, dependency files, and sample outputs required to reproduce the reported results. Detailed instructions for dataset acquisition and execution are also provided to ensure full reproducibility.