CN-122022814-A - Multi-layer business audit system and method for network transaction

CN122022814ACN 122022814 ACN122022814 ACN 122022814ACN-122022814-A

Abstract

The invention provides a multi-layer business audit system and a multi-layer business audit method for an algorithm network transaction, wherein the multi-layer business audit system comprises the steps of S1, distributed acquisition of multi-source transaction data streams, S2, multi-layer audit analysis of the data streams, S21, time series analysis, detection of time dimension abnormality by adopting an LSTM or a Transformer model, S22, graph network learning analysis, identification of abnormal transaction behavior interaction by using GNN, S23, statistics analysis, detection of abnormal behaviors deviating from normal distribution based on a statistical model, S3, dynamic compliance verification of transaction behaviors based on lightweight intelligent contracts, verification content comprising identity authentication, access control and rule matching, S4, real-time response and tracing of the identified abnormal behaviors, S5, and generation of a business audit report. The invention combines dynamic and multidimensional audit model with intelligent contract verification to realize real-time and cross-domain compliance audit of the whole life cycle of the transaction of the computing network.

Inventors

WANG CHENG
LIU WEI
CHEN XUE
TANG HAO
LI ZHIWEI
XIE EN
JIANG CHANGJUN

Assignees

同济大学

Dates

Publication Date: 20260512
Application Date: 20260115

Claims (9)

1. The multi-layer business audit system for the transaction of the computing network is characterized by comprising a distributed probe module, a data processing and audit analysis module, an intelligent contract verification module, an exception processing and tracing module and a business audit report generation module; the distributed probe module is arranged at a plurality of nodes of the computing power network and is used for collecting multi-source transaction data in real time; The data processing and auditing analysis module is in communication connection with the distributed probe module and is used for carrying out multi-layer auditing analysis on the acquired data and comprises a time sequence analysis unit, a graph network learning unit and a statistical analysis unit; the time sequence analysis unit is used for executing time sequence analysis; the map network learning unit is used for executing map network learning analysis; the statistical analysis unit is used for performing statistical analysis; the intelligent contract verification module is in communication connection with the data processing and audit analysis module and is used for carrying out dynamic compliance verification on transaction behaviors based on lightweight intelligent contracts; the exception handling and tracing module is in communication connection with the data handling and audit analysis module and the intelligent contract verification module and is used for responding to the exception behaviors in real time and recording logs; and the business audit report generation module is in communication connection with the exception handling and tracing module and is used for generating and outputting a business audit report.
2. A method of a multi-layer business audit system for an algorithm-based network transaction according to claim 1 including the steps of: Step S1, distributed data acquisition, namely acquiring multi-source business data related to the network transaction in real time at a plurality of network nodes of a power computing network, wherein the multi-source business data comprises time sequence data, network behavior logs and transaction task characteristics; S2, carrying out parallel or serial multi-layer analysis based on the multi-source business data, wherein the multi-layer analysis comprises time sequence analysis, graph network learning analysis and statistical analysis; the time sequence analysis is used for modeling the time sequence of the task request based on the time sequence analysis module so as to detect an abnormal mode of the transaction request in the time dimension; the graph network learning analysis is used for constructing a graph model based on interaction relations among transaction nodes, and the graph neural network is used for extracting interaction characteristics so as to identify abnormal transaction behavior interaction; The statistical analysis is carried out, the statistical feature extraction is carried out on the multi-source business data flow, and abnormal behaviors deviating from normal distribution are detected based on a statistical model; Step S3, intelligent contract verification, namely carrying out dynamic compliance verification on transaction behaviors based on a preset lightweight intelligent contract, wherein the compliance verification comprises identity authentication, access control and rule matching; S4, performing real-time response operation on the identified abnormal transaction behaviors based on the multi-layer audit analysis and compliance verification results in the steps S2 and S3, and storing an abnormal behavior record in an audit log for tracing analysis of the abnormal behaviors; and S5, generating a business audit report, namely automatically generating the business audit report comprising compliance analysis, abnormal behavior description and improvement suggestion based on the multi-layer audit analysis result, the intelligent contract verification result and the abnormal response record.
3. The method of claim 2, wherein in step S1, distributed probes are disposed at a plurality of nodes of the computing network for collecting the multi-source transaction data stream in real time.
4. A method for a multi-layer business audit system for an online transaction according to claim 3 wherein in step S1, the distributed probe collects data at an adjustable frequency, the frequency ranging from 1 millisecond to 10 seconds.
5. The method of claim 2, wherein in step S2, the multi-layer business data is analyzed in parallel or in series by a data processing and audit analysis module.
6. The method of claim 2, wherein the time series analysis module uses a model based on a long short time memory network LSTM or a transducer model based on a self-attention mechanism.
7. The method of claim 2, wherein in step S3, the rule base of the lightweight intelligent contract supports dynamic update, and the rule synchronization and consistency check across different trust domains are implemented by a distributed ledger technique.
8. The method of claim 7, wherein in step S3, the audit rule base based on the lightweight intelligent contract supports rule collaboration between different trust domains, and the rule collaboration includes sharing rules for transaction compliance audit in each trust domain when the transaction involves a cross-trust domain scenario, and performing consistency check on audit results of transaction behaviors, so as to ensure consistency of the cross-trust domain transaction audit rules.
9. The method of claim 8, wherein in step S3, when verifying compliance of transaction behaviors across trust domains, verification is performed first by local intelligent contracts, and then consistency verification is performed with cross-domain rules synchronized in the distributed ledger.

Description

Multi-layer business audit system and method for network transaction Technical Field The invention relates to the technical field of audit systems, in particular to a multi-layer business audit system and a multi-layer business audit method for an algorithm network transaction. Background With the rapid development of cloud computing, edge computing and Internet of things technologies, the deep fusion of computing resources and network resources has prompted the generation of a novel transaction mode of a computing power network. In a network transaction scenario, a user may dynamically request distributed computing, storage, and network services. However, the complexity and openness of the network transaction presents the following problems: First, the network transaction has obvious dynamic and real-time requirements. The resource scheduling and the business flow in the network transaction are in a continuously changing state, the traditional auditing mode based on static rules is difficult to track the transaction behavior in time, and quick response to abnormal behaviors is also impossible. Second, the web transaction involves multiple source data streams and complex transaction scenarios. In the transaction process, multi-source data such as calculation tasks, network flow, service interaction logs and the like exist simultaneously, and multidimensional attributes such as time, space and user behaviors are presented, so that the existing single-dimension audit technology is difficult to accurately capture abnormal behaviors and potential safety risks. Furthermore, web transactions typically occur between different trust domains, each trust domain having independent security policies and compliance requirements. The prior art lacks unified cross-domain audit rules and collaboration mechanisms, and is difficult to ensure the consistency and compliance control of cross-domain transactions. Meanwhile, in a distributed computing network transaction environment, the risk of privacy disclosure and data tampering is faced in the process of collecting, transmitting and sharing transaction data, and how to ensure the safety and the integrity of the data is also a problem to be solved. The technical scheme closest to the invention mainly comprises the following two types. The method is characterized in that an audit module is deployed at each node of an algorithm network, the business flow data is subjected to matching analysis through a predefined rule, and the result is recorded in a local log. However, this solution has the following problems: 1) Lacking in dynamics, static rules are difficult to adapt to real-time changing network transaction scenes. 2) Audit isolation, that is, each node audit module works independently, so that cross-domain transaction behaviors are difficult to cooperatively analyze. 3) Abnormal response hysteresis, namely low manual processing efficiency and incapability of meeting the real-time requirement in a dynamic transaction scene. Therefore, the scheme depends on static rules, lacks of dynamic performance, is difficult to cooperate with audit results among nodes and has delayed abnormal response. Therefore, the scheme is wholly dependent on static rules, lacks of dynamic property and synergetic property, and is difficult to be suitable for complex network transaction environments. The other type is a transaction verification method based on a lightweight intelligent contract, which performs identity authentication and authority verification on transaction behaviors through the intelligent contract and records a transaction log. However, this solution has the following problems: 1) The intelligent contract can only verify the compliance of transaction data, and cannot realize the global monitoring of service flow and abnormal behaviors. 2) Lacking full life cycle support, the scope of contract validation is limited to transactions that do not cover the full life cycle of the transaction. 3) The cross-domain support is insufficient, namely, definition and verification modes of intelligent contract rules among different trust domains are not uniform, and uniform audit and collaborative verification of the cross-trust domains cannot be realized. The proposal has single audit dimension, only performs verification in the transaction occurrence stage, cannot cover the complete life cycle of the transaction, and also has difficulty in supporting unified audit across trust domains. In conclusion, in the prior art, the audit dimension is single, the audit time point is limited, the cross-domain coordination capability is insufficient, the dynamic, real-time and cross-domain compliance audit of the multidimensional service flow in the computational network transaction scene is difficult to realize, and the actual application requirements in the computational network transaction environment cannot be met. Disclosure of Invention The invention aims to provide a multi-layer business audit metho