Cross-chain transaction auditing method based on truth value discovery technology

Abstract

The invention discloses a cross-chain transaction auditing method based on a truth value discovery technology. The method comprises three stages of preprocessing, reputation evaluation and reputation prediction. A block chain platform audits the reputation of a seller through a reputation prediction value RS. The method has the beneficial effects that the transaction score data is acquired and processed in real time by deploying the intelligent contract on the relay chain, and the reputation evaluation can reflect the latest transaction information in time. By using a truth value discovery technology, a real score can be extracted from scores of a plurality of buyers, subjective deviation and malicious manipulation behaviors in a scoring process are reduced, and the accuracy and credibility of an evaluation result are ensured. Through the statistical characteristics of Dirichlet distribution, the fluctuation of score data can be effectively processed, and the future reputation value of the seller can be predicted according to the score change. All scoring and auditing processes are publicly and transparently processed on the relay chain, and anyone can check and verify the calculation process of scoring and auditing, so that the credibility and transparency of the whole system are enhanced.

Description

Technical Field

[0001] The present invention belongs to the field of cross-chain transaction auditing, and particularly relates to a cross-chain transaction auditing method based on truth discovery technology. Background Art

[0002] With the rapid development of blockchain technology, blockchain has gradually been applied to multiple fields, including finance, supply chain management, and healthcare. The core features of blockchain technology include decentralization, transparency, and immutability, giving it significant advantages in data security and transaction credibility. However, the interoperability issue between different blockchain networks has gradually emerged, especially during cross-chain transactions. Cross-chain transactions refer to the data exchange between different blockchain networks. Since each blockchain network has its own independent rules and mechanisms, ensuring the security and reliability of cross-chain transactions has become particularly important. To address this challenge, researchers have proposed various cross-chain technology solutions aimed at achieving data exchange and asset transfer between different blockchains, breaking the "information silos" and promoting the development of a more extensive blockchain ecosystem. Existing cross-chain technology solutions are mainly divided into two categories: chain-based protocols and bridge-based solutions.

[0003] Chain-based protocols. Side chains: A side chain is a blockchain that operates independently of the main chain and enables cross-chain transfer of assets and data through interaction with the main chain. Typical representatives of side chains include ZeroCross and Zedoo, etc. These solutions usually utilize privacy protection technologies, such as zero-knowledge proofs (zk-SNARK), to ensure the privacy and security of cross-chain transactions. Hash Time Lock Contracts (HTLC): HTLC combines hash lock and time lock mechanisms to facilitate cross-chain exchange activities. It locks transactions using hash values within a specified time to ensure the secure transfer of assets among trading parties. For example, XCLAIM relaxes some strict assumptions of the original HTLC framework, improving the flexibility and efficiency of cross-chain transactions.

[0004] Bridge-based solutions. Relay chains: A relay chain is a blockchain network dedicated to cross-chain communication that enables data exchange and asset transfer between different blockchains by maintaining cross-chain gateways. Typical relay chain solutions include Polkadot and Cosmos. Relay chains usually adopt a star architecture, featuring high scalability and security, and are suitable for scenarios such as cross-chain auditing. Notary schemes: In notary schemes, one or more trusted third-party entities (i.e., notaries) are responsible for verifying the validity of cross-chain transactions. Although notary schemes are simple to implement, they introduce potential single point of failure risks. To improve the security of notary platforms, researchers have introduced secure hardware and encryption technologies. Summary of the Invention

[0005] The object of the present invention is to propose a cross-chain transaction auditing method based on truth discovery technology in view of the deficiencies of the prior art, which can reflect the score changes in real time, handle the fluctuations of score data, and enhance the transparency and credibility of the audit.

[0006] A cross-chain transaction auditing method based on truth discovery technology includes three stages: preprocessing, reputation evaluation, and reputation prediction. The blockchain platform audits the reputation of the seller through the reputation prediction value R S ;

[0007] The preprocessing includes the following steps:

[0008] Step 1: Data collection; when the buyer submits a request to the relay chain, specifying the chain ID and the seller's account address, retrieve all transaction records related to the seller on the specified chain ID; the buyer returns a comprehensive transaction list by calling the Query function of the smart contract and forms a data set, denoted as T1, T2, T3, …, T n};

[0009] Step 2: Data partitioning; divide the data set obtained in Step 1 into different time periods, specifically grouped as {E1, E2, …, E t} and based on a time frame of M / t days;

[0010] The reputation evaluation is automatically executed using a smart contract. The specific steps include:

[0011] Step 1: Dynamically update the buyer weight

[0012] A data set for a period of time contains m transaction scores of m buyers, denoted as s 1 , s 2 , …, s m ; update the buyer weight by combining the consistency between the buyer's transaction score and the estimated true value. The calculation formula for the buyer weight w i is as follows:

[0013]

[0014] For the distance function d(·), mainly calculate std m ; when receiving the buyer's transaction score, calculate Calculate d i =(s i -s avg )2; after receiving all d i for each buyer, calculate Finally, obtain

[0015] Step 2: Update the seller reputation

[0016] After the buyer's ownership weight is updated, the seller's reputation is updated accordingly: for each buyer, the weight w i and its corresponding transaction score s i are weighted and averaged to obtain the seller's comprehensive reputation score, and the seller's reputation s * The updated calculation formula is:

[0017]

[0018] The reputation prediction is automatically executed using a smart contract. The specific steps include:

[0019] Step 1: Perform reputation normalization, normalize the reputation from 0 to 1. The reputation normalization formula is; use the accumulated historical data vector and the posterior Dirichlet distribution for reputation aggregation to obtain the probability distribution p; set Y as the weighted average of p, and the expected value E[Y] is used as the predicted reputation;

[0020] Step 2: For a specific seller AP, let Y (0 ≤ Y ≤ 1) represent the continuous random variable of the true value of AP; represent the C true value levels as {β1, β2, …, β C}, (β i ∈(0, 1], i = 1, 2, …, C, β i <β i+1 ); the probability distribution vector of Y with respect to the C-level is denoted as where p{β i -1 < Y i <β i} = p i (i = 1, 2, …, C); let ξ = {ξ1, ξ2, …, ξ C} represent the cumulative true value vector; for the posterior Dirichlet distribution, p can be modeled as:

[0021] where

[0022] Step 3: Each level β i (i ∈ [1, C]) is assigned a weight value q i ; let p i represent the probability that the true value of AP is classified into β i , where Let Y be the random variable representing the weighted average of the probability of the true value appearing in p. Then the seller's reputation score R S The calculation formula is:

[0023]

[0024] where ξ iIt is the cumulative value of the true value level of the seller.

[0025] Furthermore, the reputation normalization formula is: where s i represents the reputation value of the seller in each period.

[0026] The beneficial effects of the present invention are as follows:

[0027] 1. Reflect rating changes in real time: By deploying smart contracts on the relay chain, the present invention can obtain and process transaction rating data in real time, ensuring that the reputation assessment can promptly reflect the latest transaction information. Using the true value discovery technology, the true ratings can be extracted from the ratings of multiple buyers, reducing subjective biases and malicious manipulation behaviors in the rating process, and ensuring the accuracy and credibility of the assessment results.

[0028] 2. Handle fluctuations in rating data: The present invention uses the Dirichlet distribution to model and predict the historical rating data of sellers. Through the statistical characteristics of the Dirichlet distribution, the fluctuations in rating data can be effectively handled, and the future reputation value of the seller can be predicted based on the rating changes. The Dirichlet distribution can combine the historical rating data of the seller, dynamically adjust the rating weights, and improve the accuracy of the reputation assessment.

[0029] 3. Enhance the transparency and credibility of the audit: The present invention processes all rating and audit processes publicly and transparently on the relay chain, and anyone can view and verify the calculation processes of the ratings and audits, enhancing the public credibility and transparency of the entire system. Through the public execution and recording of smart contracts, it is ensured that all rating data and calculation processes are tamper-proof, further improving the credibility of the audit process. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic flow diagram of a cross-chain transaction audit method based on the true value discovery technology according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0031] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the drawings in the present invention.

[0032] In cross-chain transactions, the ratings given by buyers to sellers not only reflect the quality of the transactions but also provide important information about the credibility of the sellers for other buyers. Therefore, ratings play a key role in cross-chain transaction audits. By comprehensively considering the ratings given by buyers to sellers, auditors can more accurately assess transaction risks and the credibility of sellers, thereby effectively improving the transaction environment of the blockchain. However, existing research mainly focuses on data privacy protection in cross-chain transaction audits and cannot effectively solve the problem of unreliable ratings in cross-chain transaction audits. There are still the following main defects:

[0033] 1. Insufficient authenticity of ratings: Most existing cross-chain transaction audit schemes rely on users' ratings of transactions. However, the authenticity and accuracy of these ratings are often difficult to guarantee and may be affected by vested interests, resulting in the unreliability of audit results.

[0034] 2. Lack of a transparent rating mechanism: The rating process in existing schemes lacks transparency, and the authenticity and accuracy of rating data are difficult to verify. This makes it difficult for auditors to fully trust this rating data, affecting the fairness and credibility of the overall audit.

[0035] 3. Difficulty in dynamically changing reputation assessment: The reputation of sellers changes over time and with transaction behavior. Existing schemes are insufficient in dynamically assessing the reputation of sellers and cannot timely reflect the latest transaction information and reputation changes.

[0036] Based on the defects of existing technologies, the present invention proposes a cross-chain transaction audit method based on truth discovery technology. The key technologies and innovations of this method are as follows: Introducing truth discovery technology: Using truth discovery technology to extract true ratings from the rating data of multiple buyers to ensure the authenticity and reliability of ratings. Transparent rating and audit mechanism: Deploying smart contracts on the relay chain to publicly and transparently process all rating and audit processes to ensure the credibility and verifiability of the calculation process. Dynamic reputation assessment: Using the Dirichlet distribution to predict the future reputation value of sellers based on their historical rating data, and timely reflecting the latest transaction information and reputation changes of sellers.

[0037] The specific content of a cross-chain transaction audit method based on truth discovery technology mainly includes three stages: preprocessing, reputation assessment, and reputation prediction. Assume that each consortium chain has transaction information of each account, and the transaction information includes the transaction records of the corresponding account and the score of each transaction.

[0038] (1) The preprocessing process is as follows:

[0039] Step 1: Data collection

[0040] When a buyer submits a request to the relay chain, specifying the chain ID and the seller's account address, retrieve all transaction records related to the seller on the specified chain ID; The buyer achieves this by calling the Query function of the smart contract, which returns a comprehensive transaction list, denoted as {T1, T2, T3, …, T n}. These transactions are the basis for audit analysis and provide important insights into the financial activities and interactions of the seller in the blockchain;

[0041] Step 2: Data partitioning

[0042] After obtaining the transaction records in Step 1, the next crucial phase involves splitting the acquired dataset into different time periods. For example, if the transaction history spans a time period of M days, we will implement a systematic partitioning strategy to classify the data into smaller intervals. By effectively partitioning the records, such as grouping them into {E1, E2, …, E t} based on a time frame of M / t days, it effectively enhances the granularity of the analysis and facilitates a more structured examination of the seller's financial activities across different time ranges. The specific algorithm process is as follows:

[0043] Input: Chain ID, seller account address

[0044] Output: {E1, E2, …, E t}

[0045] 1. Call the Query function in the smart contract to retrieve all the transaction scores {T1, T2, …, T M} of the seller;

[0046] 2. Divide the scores into multiple time periods, each period being M / t days, and divide them into E1, E2, …, E t .

[0047] (2) The reputation evaluation process is as follows:

[0048] In this stage, the reputation evaluation will be automatically executed using the smart contract. Taking a dataset for a period of time as an example, assume that the dataset contains m scores of m buyers, namely s 1 , s 2 , …, s m .

[0049] The reputation evaluation process algorithm is as follows:

[0050] Input: m scores of the buyer

[0051] Output: Reputation value s *

[0052]

[0053]

[0054] Step 1: Update the weights

[0055] According to the transaction record scores s 1 , s 2 , …, s m, update the weight of each buyer. By combining the transaction record scores for each period and updating the buyer weights based on the consistency between the buyer scores and the estimated true values, the performance of each buyer in different periods can be evaluated more carefully. The buyer weight update scheme helps to improve the accuracy and credibility of the audit process, ensuring that buyers who provide accurate scores play a greater role in the audit while reducing the impact of buyers whose scores deviate from the results. The buyer weight w i is calculated as follows:

[0056]

[0057] For the distance function d(·), mainly calculate std m ; when receiving the transaction score of a buyer, first calculate then calculate d i =(s i -s avg )2. After receiving all d i for each buyer, calculate Finally, obtain

[0058] By dynamically adjusting the buyer weights, more importance can be attached to buyers who provide accurate and reliable data during the audit process, thereby improving the overall accuracy and effectiveness of the audit.

[0059] Step 2: Reputation evaluation

[0060] After all buyer weights are updated, the seller's reputation is also updated.

[0061] Calculate the seller's reputation s * as follows:

[0062]

[0063] where s * represents the final reputation value of the seller. By taking the weighted average of the weight w i of each buyer and its corresponding score s i , the comprehensive reputation score of the seller can be obtained. The weight w i represents the importance of each buyer in the audit process. The higher the weight, the greater the impact of the buyer on determining the seller's reputation.

[0064] By using the seller reputation s * calculation formula and combining the updated weight information, the overall reputation of the seller can be evaluated more accurately. The above reputation update scheme based on buyer weights and scores helps to ensure the objectivity and accuracy of the seller reputation evaluation, providing a reliable basis for subsequent audit results and decisions.

[0065] (3) The described reputation prediction process is as follows:

[0066] At this stage, the smart contract will be used to automatically execute reputation prediction. Reputation prediction mainly includes reputation normalization, reputation aggregation, and reputation evaluation. In order to improve the evaluation of the seller's historical reputation, the Dirichlet distribution is used for reputation prediction. During the audit process, the reputation is standardized, and the Dirichlet distribution is used to predict the seller's future reputation, using the consistently standardized historical reputation. First, reputation normalization is performed to normalize the reputation from 0 to 1. Next, reputation aggregation is performed using the accumulated historical data vector and the posterior Dirichlet distribution to obtain the probability distribution p, and then Y is set as the weighted average of p, and the expected value E[Y] is used as the predicted reputation.

[0067] Step 1: The reputation prediction algorithm is as follows:

[0068]

[0069]

[0070] The specific formula for reputation normalization is:

[0071]

[0072] where s i represents the reputation value of the seller in each period;

[0073] Step 2: For a specific seller AP, let Y (0 ≤ Y ≤ 1) be a continuous random variable representing the true value of AP. Represent the C true value levels as {β1, β2, …, β C}, (β i ∈ (0, 1], i = 1, 2, …, C, β i < β i+1 ). The probability distribution vector of Y with respect to these C-levels is denoted as where p{β i - 1 < Y i < β i} = p i (i = 1, 2, …, C). Let ξ = {ξ1, ξ2, …, ξ C} represent the cumulative true value vector. For the posterior Dirichlet distribution, p can be modeled as

[0074]

[0075] where

[0076] Step 3: Reputation evaluation

[0077] To calculate the seller's reputation score, for each level βi Assign a weight value q to (i ∈ [1, C]) i . Let p i represent the probability that the true value of the AP is classified into β i , where Let Y be a random variable representing the weighted average of the probability of the true value appearing in p. Then the seller's reputation score R S can be calculated as

[0078]

[0079] where ξ i is the cumulative value of the level to which the seller's true value belongs.

[0080] Finally, the blockchain platform can audit the seller's reputation through the predicted reputation value R S , which also facilitates subsequent buyers to decide whether to trade with the seller based on the reputation.

[0081] It should be noted that in addition to being used for cross-chain transaction audit tasks, this solution can also be deployed in other blockchain application scenarios. Specific examples include cross-chain asset transfer tasks for the financial field and cross-chain data audit tasks for supply chain management.

Claims

1. A cross-chain transaction auditing method based on truth discovery technology, characterized in that: The cross-chain transaction auditing method includes three stages: preprocessing, reputation evaluation, and reputation prediction. The blockchain platform audits the reputation of the seller through the reputation prediction value R S ; The preprocessing includes the following steps: Step 1: Data collection; When a buyer submits a request to the relay chain, specifying the chain ID and the seller's account address, retrieve all transaction records related to the seller on the specified chain ID; The buyer returns a comprehensive transaction list by calling the Query function of the smart contract and forms a data set, denoted as T1, T2, T3, …, T n}; Step 2: Data partitioning; The data set obtained in Step 1 is split into different time periods, specifically grouped as {E1, E2, …, E t} and based on a time frame of M / t days; The credit evaluation is automatically executed using a smart contract. The specific steps include: Step 1: Dynamically update the buyer weight A dataset for a period of time contains m transaction scores of m purchasers, denoted as s 1 , s 2 , …, s m ; Update the buyer weight by combining the consistency between the buyer's transaction score and the estimated true value. The calculation formula for the buyer weight w i is as follows: For the distance function d(·), mainly calculate std m ; when receiving the transaction score of the buyer, calculate Calculate d i =(s i -s avg )2; after receiving all d i of each buyer, calculate Finally obtain Step 2: Update the seller's credit After the buyer's ownership weight is updated, the seller's reputation is updated accordingly: the weight w of each buyer i and its corresponding transaction score s i are weighted and averaged to obtain the comprehensive reputation score of the seller, and the seller's reputation s * The update calculation formula is: The credit prediction is automatically executed using a smart contract. The specific steps include: Step 1: Perform credit normalization, normalizing the credit from 0 to 1. The credit normalization formula is; perform credit aggregation using the accumulated historical data vector and the posterior Dirichlet distribution to obtain the probability distribution p; set Y as the weighted average of p, and use the expected value E[Y] as the predicted credit; Step 2: For a specific seller AP, let Y (0 ≤ Y ≤ 1) be a continuous random variable representing the true value of the AP; represent C true value levels as {β1, β2, …, β C}, (β i ∈(0, 1], i = 1, 2, …, C, β i < β i+1 ); the probability distribution vector of Y with respect to the C-level is denoted as where p{β i -1 < Y i < β i} = p i (i = 1, 2, …, C); let ξ = {ξ1, ξ2, …, ξ C} represent the cumulative true value vector; for the posterior Dirichlet distribution, p can be modeled as: Among them Step 3: For each level β i (i ∈ [1, C]) is assigned a weight value q i ; Let p i represent the probability that the true value of the AP is classified into β i , where Let Y be a random variable representing the weighted average of the probability of the true value appearing in p. Then the seller's reputation score R S is calculated as follows: Among them, ξ i is the cumulative value of the level to which the true value of the seller belongs.

2. The cross-chain transaction auditing method based on the truth discovery technology according to claim 1, characterized in that: The reputation normalization formula is as follows: where s i represents the seller's reputation value in each period.

Images