Software risk intelligent research and judgment and closed-loop optimization system based on platform operation data

Abstract

The application belongs to the technical field of software platform operation risk management and control, and particularly relates to a software risk intelligent research and judgment and closed-loop optimization system and method based on platform operation data, which is composed of a main control unit and multiple function modules such as pre-operation data preprocessing, surface code specification research and judgment, business logic risk research and judgment, running performance resource research and judgment, architecture security operation research and judgment, AI model reasoning and execution scheduling. The system adopts a hierarchical research and judgment process, carries out multi-level deep analysis from code surface specification to software global architecture operation safety relying on real-time platform operation data, and provides algorithm support by using an AI model reasoning module to accurately locate the software risk level, fault root cause and specific impact point. Relying on real operation data, the software full-link risk is accurately identified, the platform failure rate is greatly reduced, the software iteration and operation efficiency are improved, and complete intelligent risk management and control support is provided for the long-term stable operation of large and medium-sized online business platforms.

Description

Technical Field

[0001] This invention belongs to the field of intelligent management and control technology of software platform operation risks, specifically involving a software risk intelligent assessment and closed-loop optimization system based on platform operation data. Background Technology

[0002] As internet business platforms continue to iterate, the scale of software systems, the number of microservices, and online business traffic continue to grow. The platforms generate massive amounts of operational data daily, including source code, full-chain business logs, API endpoints, cluster resource monitoring metrics, security vulnerability alerts, user access logs, and other multi-dimensional data. Various risks arising during software operation (code syntax issues, business logic anomalies, cluster performance bottlenecks, and architectural security vulnerabilities) can directly lead to online failures, business errors, service downtime, and data leaks, causing economic losses and security risks to platform operations.

[0003] Traditional software risk management methods have significant shortcomings: First, risk detection is detached from real-world operational scenarios, relying solely on static code scanning without incorporating real-world operational data such as online traffic, resource load, and business errors, thus missing a large number of dynamic operational risks. Second, risk assessment has a single dimension, only testing code or performance, failing to connect the entire security chain of code, business, cluster, and architecture to pinpoint the root cause. Third, risk handling relies on manual operations and maintenance; after identifying risks, manual analysis and optimization are required, resulting in long handling cycles and continuous impact on online business. Fourth, the risk assessment, optimization, and verification processes are fragmented, lacking a standardized closed-loop mechanism, leading to repeated recurrence of similar operational risks and preventing the accumulation of platform risk governance experience.

[0004] Existing AI code inspection tools only focus on static identification of code defects, failing to integrate with platform operational data links. They cannot correlate with real-world online failure scenarios to differentiate the actual impact of risks, and lack tiered optimization and closed-loop verification mechanisms, making them ill-suited for the routine risk management needs of large and medium-sized online business platforms. Therefore, designing a software risk management system that relies on the platform's full operational data, employs hierarchical intelligent analysis, tiered automatic optimization, and end-to-end closed-loop verification is a core technical requirement for improving platform stability and reducing operational costs. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, the first aspect of this application proposes a software risk intelligent assessment and closed-loop optimization system based on platform operation data, comprising: a main control unit and a pre-processing module for front-end operation data, a surface code specification assessment module, a business logic risk assessment module, an operation performance resource assessment module, an architecture security operation assessment module, an AI model inference module, and an execution scheduling module connected in communication with it. The pre-processing module for operational data is used to aggregate all operational data, business logs, source code files, and interface tracking metrics of the platform and complete unified cleaning and standardized structured processing. The surface code specification analysis module is used to conduct surface-level explicit risk assessment of software based on standardized operational code data; The business logic risk assessment module is used to receive the abnormal results output by the surface code standardization assessment module and, in conjunction with business operation logs, identify mid-level business logic risks derived from surface code issues. The operational performance resource assessment module is used to receive the abnormal assessment results from the business logic risk assessment module. Based on the platform's real-time resource usage, interface calls, and service consumption time operation indicators, it identifies hidden performance bottlenecks and resource consumption-related operational risks. The architecture security operation assessment module is used to receive the abnormal assessment results from the performance and resource assessment module. Based on operational data such as cross-service calls, data transmission, access control, online vulnerabilities, and cluster fault tolerance, it assesses the software architecture adaptability, module coupling, interface compatibility, data transmission security, high-risk vulnerabilities, cluster abnormal fault tolerance, and long-term stable operation risks of the production platform. The AI ​​model reasoning module is equipped with a multi-level AI reasoning model, which performs feature extraction and in-depth analysis on the collected first-level, second-level, third-level, and fourth-level hierarchical operation assessment data, locates the software risk level, the root cause of the failure and the precise impact point, and outputs the hierarchical operation risk assessment result. Based on the hierarchical operational risk assessment results, the execution scheduling module outputs start / stop scheduling instructions and risk elimination instructions for each module, driving the execution of risk governance and software optimization actions at the corresponding level.

[0006] The second aspect of this application also provides a method for intelligent assessment and closed-loop optimization of software risks based on platform operation data. Based on the aforementioned system implementation, the specific process includes: S1. The system gathers multi-source operational data in real time, including platform source code, full-link business logs, interface tracking points, cluster resource indicators, and security vulnerability alarms. Through the pre-processing module for operational data, it completes data cleaning, format unification, structured slicing, and dirty data filtering, and outputs standardized and parsable operational datasets. S2. The execution scheduling module continuously initiates the first-level surface code standardization judgment process. The surface code standardization judgment module, in conjunction with the first-level AI inference model, completes the code and basic tracking compliance verification. Based on the judgment results, it determines whether to directly terminate the entire process or rigidly trigger the second-level business logic judgment. S3. Strictly follow the step-by-step rule. Only after identifying abnormal risks at the previous level, initiate the second-level business logic assessment, the third-level performance resource assessment, and the fourth-level architecture security final assessment in sequence. There is no skipping or over-level execution. Deeply explore all dimensions of software operation risks from surface code, middle business, underlying performance to global architecture security. S4, the AI ​​model reasoning module receives operational assessment data from each level, calls the corresponding level's dedicated reasoning model to complete risk feature extraction, root cause analysis, and impact scope determination, and outputs a graded operational risk report to clarify the risk level, fault location, and online impact scope. S5. The execution scheduling module calls the matching level-specific optimization solution based on the hierarchical risk report, and sequentially completes lightweight code standardization, business logic iteration, cluster performance tuning, architecture reconstruction and security hardening. S6. After each level of optimization is completed, the corresponding special closed-loop regression verification is initiated. After the verification is passed, the entire process data of this operation risk is archived, forming a full-link intelligent closed-loop management and control of platform software risk, namely "data aggregation - hierarchical analysis - AI root cause location - hierarchical optimization - regression verification - archiving and accumulation".

[0007] A third aspect of this application also provides an electronic device, comprising: A processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor, which can invoke the program instructions to perform the method described in the second aspect.

[0008] The fourth aspect of this application also provides a computer-readable storage medium storing instructions that, when executed on a computer, perform the method described in the second aspect.

[0009] Fifthly, this application provides a computer program product, including a computer program or instructions that, when executed by a computer, implement the method described in the second aspect.

[0010] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: Compared to traditional software risk management technologies, this invention possesses significant technical advantages and practical value. Its core relies on real, multi-source operational data from the platform for risk assessment, overcoming the drawbacks of traditional static code scanning, which is detached from online scenarios and misses dynamic operational risks, effectively reducing false positives and false negatives. This invention innovatively constructs a four-level hierarchical progressive assessment system, delving into the root causes of risks layer by layer, from surface-level code standards, mid-level business logic, and underlying performance resources to global architectural security. Coupled with a module time-sharing scheduling mechanism, it significantly conserves system computing resources while ensuring 24 / 7 basic monitoring capabilities. Furthermore, it configures differentiated hierarchical optimization strategies and dedicated closed-loop verification mechanisms for different risk levels, achieving precise governance through lightweight standardization, business iteration, performance tuning, and architectural security hardening. Multi-scenario specialized testing prevents risk recurrence. In addition, this invention supports flexible deployment across multiple platforms, adapting to various microservice business platforms and distributed clusters, effectively improving the intelligence level of software operation and maintenance, reducing operation and maintenance costs, and ensuring the long-term stable and secure operation of the software platform. Attached Figure Description

[0011] The present invention will be described by way of example and with reference to the accompanying drawings, wherein: Figure 1 This is a block diagram of the system architecture in this invention; Figure 2 This is a flowchart illustrating the method of the present invention.

[0012] Figure 3 This is a block diagram of the hardware composition of the electronic device of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0014] Example 1:

[0015] This embodiment provides a software risk intelligent assessment and closed-loop optimization system based on platform operation data. (See also...) Figure 1This system is applicable to full-lifecycle risk management scenarios for various microservice architecture software business platforms, cloud-based operation and maintenance service platforms, and distributed software operation systems. It can adapt to the routine risk governance needs of software platforms with different business volumes and iteration frequencies. The system includes: a main control unit and a pre-processing module for operational data, a surface code specification assessment module, a business logic risk assessment module, an operational performance and resource assessment module, an architecture security operation assessment module, an AI model inference module, and an execution scheduling module connected to it. Each functional module completes data interaction and instruction transmission based on a distributed communication protocol. The modules cooperate and constrain each other to achieve fully automated management of the entire process of layered identification, intelligent assessment, precise optimization, and closed-loop verification of software operation risks.

[0016] The pre-processing module for operational data is used to comprehensively aggregate operational data from the software platform across all dimensions. Specifically, it covers multi-source heterogeneous data such as full-link business logs generated during online platform operation, source code files from version iteration updates, event tracking indicators for service interface deployment, cluster operation logs, security alarm data, and user access behavior data. Simultaneously, it performs integrated processing on the aggregated raw and messy data, including noise reduction and cleaning, format unification, standardized structured parsing, field regularization, and invalid data removal. This transforms unstructured, fragmented, and differentiated raw operational data into a standardized structured dataset with a unified format and regular dimensions that can be recognized and parsed by AI models. This provides accurate, effective, and standardized data support for subsequent risk assessment at various levels, avoiding assessment errors caused by data messiness, missing data, and data anomalies from the source.

[0017] The surface code standardization assessment module is used to conduct a comprehensive risk assessment of explicit and standardization issues on the surface of the software source code, based on the standardized and structured operational code data and basic tracking data output by the pre-processing module of the pre-operational data. It focuses on basic dimensions such as code syntax, format standardization, compilation rules, basic configuration, and tracking deployment, and accurately identifies basic and explicit operational risks in the early stages of software operation. It is the first line of defense for software risk management and can quickly identify basic code compliance issues that are not deeply related.

[0018] Business Logic Risk Assessment Module: This module accurately receives the anomaly assessment results from the surface code standardization assessment module. Taking surface code anomalies as the starting point for tracing the source, it combines core data such as the platform's full-link business operation logs, business flow records, and function execution trajectories to deeply identify mid-level business logic anomaly risks derived from basic issues such as non-standard surface code and code defects. It focuses on investigating mid-level operational risks that affect the normal operation of the platform, such as business process failures, logical conflicts, and functional deviations, thus achieving a progressive tracing and assessment from surface code issues to mid-level business risks.

[0019] The performance and resource assessment module receives the business anomaly assessment results from the business logic risk assessment module. It uses business logic anomalies as the assessment trigger condition and relies on the platform's real-time dynamic server resource usage data, microservice interface call frequency and time data, thread running status, memory usage and reclamation status, cluster load parameters and other core operational indicators to deeply explore hidden operational risks that are not easily detected by static scanning, such as performance bottlenecks, resource waste, and resource scheduling imbalances during software operation. This enables in-depth source tracing and assessment from business anomalies to performance and resource risks.

[0020] Architecture Security Operations Assessment Module: This module receives the performance and resource anomaly assessment results from the performance and resource assessment module. As the highest-level in-depth assessment unit of the system, it relies on comprehensive and in-depth operational data, including cross-service call logs, data transmission message records, user access permission logs, online vulnerability scanning reports, cluster fault-tolerant operation data, and module dependency correlation data, to systematically assess the overall software architecture's suitability, microservice module coupling, cross-version interface compatibility, data transmission security level, high-risk vulnerability risks, cluster fault tolerance capabilities, and the long-term stable operation capabilities of the production platform. This allows for the precise identification of systemic and high-risk architectural risks that affect the overall operational security of the platform.

[0021] AI Model Inference Module: This is the core intelligent computing and decision-making unit of the system. It is equipped with multi-level dedicated AI inference models adapted to different assessment levels. It can receive structured data generated by hierarchical operational assessments at levels one, two, three, and four. Through deep learning algorithms, it can automatically extract risk features, perform data correlation analysis, mine abnormal patterns, and conduct in-depth inferences. It can accurately locate the risk level of various operational risks in the software, the root cause of the failure, the precise location of the failure, and the scope of impact on online business. Finally, it outputs standardized, quantifiable, and traceable hierarchical operational risk assessment results, providing core intelligent decision-making basis for subsequent risk optimization strategy output.

[0022] The execution scheduling module is the central scheduling unit for the entire system operation. It can receive the hierarchical operational risk assessment results output by the AI ​​model inference module in real time. Based on different levels and types of software operational risks, it can intelligently generate and output start / stop scheduling instructions, risk governance execution instructions, and optimization verification instructions for each functional module. It can accurately drive the implementation of risk governance and software optimization actions at the corresponding level and in the corresponding scenario, and coordinate and manage the orderly operation of the entire assessment, optimization, and verification process.

[0023] In one specific implementation, the main control unit sets the front-end operational data preprocessing module and the surface code specification assessment module to a 24 / 7 continuous operation state, ensuring that the system can complete the preprocessing of operational data aggregation and basic code risk screening around the clock, and realize real-time monitoring and immediate early warning of basic software risks. At the same time, the business logic risk assessment module, the operation performance resource assessment module, the architecture security operation assessment module, the AI ​​model inference module, and the execution scheduling module are set to an on-demand trigger start state, which is automatically woken up and started only when the front-end module detects abnormal risks, effectively reducing the computing power consumption and resource occupation of the system during normal operation, and balancing the real-time monitoring of the system with the energy efficiency of operation.

[0024] Furthermore, the surface code standardization assessment module pulls the structured code and basic tracking data output by the pre-processing module in real time and transmits them synchronously to the AI ​​model inference module. The AI ​​model inference module calls the first-level code compliance inference model and strictly benchmarks against the platform's preset unified coding standards, code compilation baselines, and basic development standards to conduct comprehensive compliance verification of the full-text syntax, writing format, compilation parameters, and basic configuration of the code. If all detection indicators meet the standard requirements, the software is determined to have no surface operational risks, and the execution scheduling module directly terminates the entire assessment process to reduce unnecessary computing power consumption. If any indicator is abnormal, a first-level surface operational risk is determined to exist, and the execution scheduling module rigidly triggers the second-level business logic assessment process to start the next level of in-depth source tracing detection.

[0025] After receiving the results of the Level 1 anomaly assessment, the business logic risk assessment module automatically initiates the Level 2 in-depth assessment. It accurately collects core operational data such as platform-wide business link logs, program branch execution data, and upstream and downstream service interface input / output matching data. This data is then uniformly organized and sent to the AI ​​model inference module. The Level 2 business logic inference model performs a comprehensive risk assessment of the business process integrity, logical rationality, and functional adaptability. If all operational indicators at this level are normal and business flow is normal, it is determined that only Level 1 surface-level basic risks exist, and no in-depth assessment is required. The execution scheduling module terminates the progressive assessment process and simultaneously executes the Level 1 lightweight optimization solution. If there are anomalies in the indicators at this level, it is determined that there are Level 2 business operational risks, rigidly triggering the Level 3 performance resource assessment process to further investigate the underlying risk causes.

[0026] After receiving the results of the Level 2 anomaly assessment, the performance resource analysis module initiates a Level 3 in-depth performance analysis. This involves batch collecting core operational metrics such as the execution time of platform services, real-time memory usage of each node in the cluster, system resource reclamation logs, frequency and duration of third-party interface calls, and thread pool running status and blocking conditions. These metrics are then organized and sent to the AI ​​model inference module. The Level 3 performance resource inference model performs a comprehensive assessment of the risks associated with software performance, resource scheduling, and cluster load. If all metrics at this level are compliant and meet the standards, it is determined that only Level 2 business logic risks exist. The execution scheduling module then terminates the progressive analysis process and executes the Level 2 logic optimization plan. If any metrics are abnormal, it is determined that there are Level 3 performance resource operational risks, rigidly triggering the Level 4 final architecture security analysis process to investigate systemic architectural security vulnerabilities.

[0027] After receiving the results of the Level 3 anomaly assessment, the architecture security operation assessment module initiates a Level 4 final in-depth assessment. This involves comprehensively collecting deep core data such as platform microservice module dependency coupling data, cross-version and cross-service interface compatibility logs, encrypted records of user privacy data transmission, online vulnerability scanning operation reports, cluster fault tolerance alarm information, and system permission access logs. After standardized processing, this data is sent to the AI ​​model inference module, where the Level 4 architecture security inference model completes the assessment of the platform's systemic operational security risks. If all indicators at this level are compliant and meet the standards, it is determined that only Level 3 performance resource risks exist. The execution scheduling module then terminates the progressive assessment process and executes the Level 3 performance tuning solution. If any indicators are abnormal, it is determined that there is a Level 4 high-risk architecture security operation risk. The execution scheduling module immediately initiates a Level 4 architecture reconstruction and security hardening optimization solution to thoroughly address the high-risk systemic risk.

[0028] Furthermore, in the process of analyzing the first, second, third, and fourth level of tiered operations, the criteria for determining whether data is normal or abnormal at each level are based on the platform's standardized operational baseline, industry-standard technical specifications, and system-preset security thresholds, precisely adapting to the actual online operation scenarios of the software platform. The specific criteria for each level are as follows: Level 1 surface analysis judgment criteria: Based on the platform's unified coding standards and code compilation baseline, when the source code has no syntax errors, complete and standardized variable definitions, uniform and regular code formatting and indentation, no redundant, invalid, or empty code, and a 100% compilation pass rate, and there are no missing or abnormal data collection points in the basic monitoring system, the data is judged to be normal and there is no surface-level operational risk; when the source code has any of the following problems: syntax errors, undefined variables, disordered formatting, redundant or empty code, compilation failure, missing basic monitoring points, or abnormal data collection, the data is directly judged to be abnormal and there is a level 1 surface-level operational risk.

[0029] Level 2 business logic analysis and judgment criteria: Based on the platform's established business process specifications and full-link branch coverage operational baseline, when the software business link has no logical dead zones, all program branches execute completely, the input and output parameters of upstream and downstream service interfaces match perfectly, the business function output results meet online operational expectations, and there are no high-frequency continuous business error logs, the data is judged to be normal and only has a level 1 surface risk; when the business link has any of the following problems: business logic vulnerabilities, missing functions, branch failures, abnormal parameter matching, or high-frequency business abnormal alarms, the data is judged to be abnormal and there is a level 2 business operation risk.

[0030] Level 3 performance resource assessment criteria: Based on platform-preset performance thresholds and cluster resource management specifications, when the average service execution time, node memory usage, connection pool resource reclamation efficiency, third-party interface call frequency, and thread pool blocking rate are all within the compliant range, and the system has no operational alarms such as memory leaks, redundant execution loops, service response timeouts, or resource preemption conflicts, the data is judged to be normal and there is only a Level 2 business logic risk; when the system experiences any of the following issues: execution timeout, excessive memory usage, untimely resource release, redundant interface calls, thread blocking, or cluster load imbalance, the data is judged to be abnormal and there is a Level 3 performance resource operational risk.

[0031] The Level 4 architecture security assessment criteria are as follows: Based on the platform's microservice architecture design specifications, network security protection standards, and long-term stable operation requirements of the cluster, when module dependency coupling is compliant, cross-service interfaces are fully compatible, user-sensitive data transmission is fully encrypted, there are no high-risk vulnerability alerts, the anomaly tolerance mechanism is complete, interface permission verification is fully covered, and the cluster operates without downtime risks over a long period of time, the data is judged to be normal and only has a Level 3 performance resource risk; when the system has any of the following problems: chaotic architecture layering, high-coupling and redundancy of modules, incompatibility of cross-version interfaces, plaintext data transmission, high-risk injection / privilege escalation vulnerabilities, insufficient cluster fault tolerance, or lack of permission control, the data is judged to be abnormal and there is a high-risk Level 4 architecture security operation risk.

[0032] Furthermore, for different risk levels, impact ranges, and severity of risks—Level 1 surface risks, Level 2 business logic risks, Level 3 performance resource risks, and Level 4 high-risk architectural security risks—the system employs differentiated and customized software closed-loop optimization solutions to achieve precise policy implementation and hierarchical governance. The specific optimization solutions for each level are as follows: The Level 1 Surface-Level Operational Risk Optimization Solution specifically addresses fundamental and superficial issues such as code syntax errors, inconsistent code formatting, missing variable definitions, redundant and invalid code, and abnormal data collection. The system automatically performs lightweight rectification and optimization operations without manual intervention or modification of the core software logic. The AI ​​model inference module accurately locates abnormal code lines and invalid data collection points, automatically completing code format standardization, redundant code snippet cleanup, missing variable completion, compilation error repair, and missing basic data collection. After optimization, the system automatically initiates code compilation verification and data collection connectivity testing, comprehensively ensuring that the optimized code can be compiled and run normally, and that basic platform operational data can be collected completely and accurately. The entire process involves zero intrusion into business logic and the overall software architecture, ensuring platform operational stability.

[0033] Level 2 Business Operation Risk Optimization Solution: Specifically targeting mid-level business risks such as business logic dead zones, invalid execution branches, interface parameter mismatches, business function output deviations, and high-frequency online business errors, the system intelligently executes iterative optimization of business logic. The AI ​​model inference module, combined with the platform's massive online real business logs, accurately locates abnormal business links, failed program branches, and logical conflict points. Through intelligent reconstruction, it completes closed-loop optimization of business processes, fills in missing business links, corrects upstream and downstream interface parameter mappings, and eliminates logical dead zones. After optimization, the system automatically replays the platform's real business simulation traffic to conduct full-branch, full-scenario business regression testing, completely eliminating business operation anomalies and ensuring that the online business process is completely closed-loop and that functional output fully complies with the platform's operational standards.

[0034] The Level 3 Performance Resource Operation Risk Optimization Solution specifically addresses performance bottlenecks such as service response timeouts, memory overflows, system resource leaks, redundant code loops, frequent invalid API calls, thread blocking, and cluster load imbalance. The system performs targeted performance resource tuning and optimization. The AI ​​model inference module leverages real-world platform operation data and resource usage metrics to intelligently streamline redundant loop code, reconstruct inefficient execution chains, add an automatic system resource reclamation mechanism, optimize API call merging strategies, resolve thread blocking issues, and balance the load pressure across cluster nodes. After optimization, the system automatically initiates high-concurrency stress tests to simulate peak platform operation scenarios, verifying that all cluster performance metrics have returned to compliance thresholds, and comprehensively improving software performance and resource utilization.

[0035] A Level 4 architecture security and high-risk operational risk optimization solution: Specifically designed to address systemic and high-risk platform risks such as chaotic architecture layering, highly coupled and redundant modules, incompatible cross-service interfaces, leaked user data transmission, high-risk online security vulnerabilities, lack of cluster fault tolerance mechanisms, and ineffective access control. The system undergoes deep architecture reconstruction and operational security hardening optimization; the AI ​​model inference module intelligently reconstructs the microservice layered architecture, splits highly coupled and redundant modules, unifies cross-version and cross-service interface adaptation standards, supplements data transmission encryption mechanisms and access authentication systems, accurately fixes various high-risk online vulnerabilities such as injection and privilege escalation, and improves cluster fault degradation and multi-level fault tolerance systems; after optimization, multi-version compatibility testing, security penetration testing, and 72-hour uninterrupted cluster stability testing are conducted sequentially to comprehensively verify the architecture's security and stability, and completely eradicate systemic high-risk operational risks.

[0036] Furthermore, the system is configured with a dedicated hierarchical operational closed-loop verification mechanism. Differentiated specialized regression verification processes are set up for the risk optimization results at different levels to ensure that each type of risk optimization is thorough, without residue, and without recurrence. After each level of risk optimization is completed, the corresponding specialized regression verification is strictly executed: After the first-level optimization is completed, the system automatically performs code compilation review and instrumentation connectivity verification, comprehensively checking the standardization of surface code, compilation validity, and the completeness of instrumentation data collection, thoroughly confirming that the risks in surface code and basic data collection are completely eliminated; After the second-level optimization is completed, the system builds an online simulation operation scenario, replays the platform's real business traffic to conduct full-link, full-branch business regression testing, checks the operational status of all business processes, confirms that there are no abnormalities in the business processes and that online business errors are completely cleared; After the third-level optimization is completed, the system automatically starts the cluster. High-concurrency stress testing comprehensively verifies all performance indicators, including service execution time, memory usage, resource reclamation efficiency, thread running status, and cluster load, ensuring all indicators return to compliance ranges. After four levels of optimization, the system sequentially conducts cross-version compatibility testing, security penetration testing, and long-term cluster stability testing to comprehensively verify the rationality of the software architecture design, the absence of security vulnerabilities, and the stable long-term operation of the cluster. After all levels of verification pass, the system automatically completes the risk closure and archiving, fully retaining all operational assessment logs, risk analysis reports, and optimization execution records, forming traceable and reusable risk governance data assets. If the verification fails, the system will automatically re-trigger the corresponding level of assessment and optimization process, iterating until all verification indicators meet the standards, achieving 100% closed-loop risk governance.

[0037] Each level of optimization is validated independently after completion. If the validation fails, the optimization is reassessed and re-evaluated. Level 1: Code compilation review + instrumentation connectivity testing; Level 2: Online business traffic simulation full-branch regression; Level 3: Cluster high-concurrency stress performance test; Level 4: Multi-version compatibility testing, security penetration testing, and long-term cluster stability testing; After all verifications are passed, archive all operational data, analysis reports, and optimization records related to this risk, thus completing the risk closure loop.

[0038] Example 2:

[0039] This embodiment provides a method for intelligent assessment and closed-loop optimization of software risks based on platform operation data. It is implemented based on the intelligent assessment and closed-loop optimization system for software risks described in Embodiment 1. This method is adaptable to the routine, automated, and intelligent risk management needs of various software operation platforms, enabling an upgrade from passive handling of software operation risks to proactive prediction, precise management, and closed-loop data accumulation. (See also...) Figure 2 The specific execution process includes: S1: The system gathers various multi-source heterogeneous operational data from the platform in real time, 24 / 7. Specifically, it includes core data such as software iteration source code, full-link business flow logs, service interface monitoring data, cluster resource operation indicators, network security vulnerability alarm information, and user access permission logs. Through the pre-processing module, the raw data is thoroughly cleaned and filtered, standardized in format, structured and segmented, and dirty and invalid data is removed. Finally, it outputs a standardized and parsable operational dataset with a standardized format, unified dimensions, and accurate and effective data, providing high-quality data support for subsequent multi-level risk assessment.

[0040] S2. The execution scheduling module continuously drives the system to start the first-level surface code standardization judgment process. The surface code standardization judgment module analyzes standardized operational data in real time and combines it with the first-level AI inference model to complete a comprehensive verification of source code standardization, basic data entry integrity, and code compilation compliance. Based on the real-time judgment results, it intelligently determines to directly terminate the entire judgment process to save system computing resources, or rigidly and progressively trigger the second-level business logic judgment process to start in-depth risk tracing.

[0041] S3. The system strictly follows the rules of progressive and in-depth analysis, and strictly prohibits skipping or overstepping levels of analysis. Only after the previous level of analysis has clearly identified abnormal risks will the system sequentially initiate the second-level business logic analysis, the third-level performance resource analysis, and the fourth-level architecture security final analysis. It explores the potential operational risks of the software from the surface code specifications, the middle-level business logic, the underlying running performance to the global architecture security in a comprehensive and thorough manner, so as to achieve a comprehensive and thorough tracing of the root causes of risks.

[0042] The S4 AI model inference module receives operational analysis data transmitted from each level in real time, accurately calls the corresponding level's dedicated AI inference model, completes intelligent extraction of risk features, multi-dimensional data correlation analysis, in-depth deduction of fault root causes, and accurate determination of the scope of business impact. Finally, it outputs a standardized and visualized hierarchical operational risk report, clearly marking the risk level, precise fault location, root cause of risk, and actual impact on online business, providing core decision-making basis for subsequent precise optimization.

[0043] S5, the execution scheduling module, intelligently matches the corresponding level of exclusive optimization strategy based on the graded risk report output by the AI ​​model. It precisely drives the system to complete lightweight code regularization optimization, business logic iteration optimization, cluster performance resource tuning, architecture reconstruction and security hardening optimization in sequence. It achieves differentiated and precise governance for different levels and types of risks, taking into account both optimization efficiency and platform operation stability.

[0044] S6. Once all risk optimization work at each level is completed, the system automatically initiates the corresponding level's special closed-loop regression verification work. Through various methods such as scenario-based testing, stress testing, and security testing, the optimization effect is verified. After the verification is passed, the system automatically archives the entire process of risk governance operation data, analysis reports, optimization records, and verification results, forming a full-link intelligent closed-loop management and control system for platform software risks, which includes "data aggregation - hierarchical analysis - AI root cause location - graded optimization - regression verification - archiving and accumulation". This system continuously accumulates platform risk governance experience and constantly improves the platform's risk prevention and control capabilities.

[0045] Example 3:

[0046] This embodiment provides an electronic device, see reference. Figure 3 This device can be independently deployed in various scenarios such as software operation and maintenance servers, cloud management terminals, and local operation and maintenance workstations, adapting to the intelligent risk management implementation needs of different platforms. Specifically, it includes a processor, a memory, and a communication bus. The processor and the memory communicate and transmit data in real time and stably through the communication bus. The memory pre-stores program instructions and algorithms that can be loaded and executed by the processor. The processor calls the program instructions stored in the memory, which can completely and accurately execute all the process steps of the software risk intelligent assessment and closed-loop optimization method based on platform operation data as described in Example 2, realizing fully automatic intelligent assessment, optimization, and closed-loop management of software operation risks.

[0047] Example 4:

[0048] This embodiment provides a computer-readable storage medium, which includes various general-purpose computer storage media such as USB flash drives, portable hard drives, read-only memory, random access memory, disks, and optical discs. The computer-readable storage medium has computer program instructions for implementing this method permanently stored on it. When the instructions are run on terminal devices such as computers, servers, and embedded operation and maintenance equipment, they can automatically execute all the steps of the software risk intelligent assessment and closed-loop optimization method based on platform operation data as described in Embodiment 2. It supports offline batch deployment and normal background operation, providing hardware support for intelligent risk management and control of software platforms.

[0049] Example 5:

[0050] This embodiment provides a computer program product, which includes a computer program or instruction package that can be deployed independently and run across terminals. It supports multiple deployment methods such as cloud deployment, local server deployment, and integrated deployment on an operation and maintenance platform. The key feature is that when the computer program or the instructions are executed by a computer terminal, server, or cloud computing platform, it can fully implement the intelligent software risk assessment and closed-loop optimization method based on platform operation data as described in Embodiment 2. It can automatically complete the entire process of intelligent management and control, including multi-source operation data aggregation, four-level hierarchical intelligent assessment, AI risk root cause location, hierarchical precise optimization, special regression verification, and risk data archiving and accumulation, and is suitable for the risk governance needs of various large, medium and small software business platforms.

[0051] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A software risk intelligent assessment and closed-loop optimization system based on platform operation data, characterized in that: include: The main control unit, along with its communicating modules, including the pre-processing module for operational data, the surface code specification assessment module, the business logic risk assessment module, the operational performance and resource assessment module, the architecture security and operational assessment module, the AI ​​model inference module, and the execution scheduling module; The pre-processing module for operational data is used to aggregate all operational data, business logs, source code files, and interface tracking metrics of the platform and complete unified cleaning and standardized structured processing. The surface code specification analysis module is used to conduct surface-level explicit risk assessment of software based on standardized operational code data; The business logic risk assessment module is used to receive the abnormal results output by the surface code standardization assessment module and, in conjunction with business operation logs, identify mid-level business logic risks derived from surface code issues. The operational performance resource assessment module is used to receive the abnormal assessment results from the business logic risk assessment module. Based on the platform's real-time resource usage, interface calls, and service consumption time operation indicators, it identifies hidden performance bottlenecks and resource consumption-related operational risks. The architecture security operation assessment module is used to receive the abnormal assessment results from the performance and resource assessment module. Based on operational data such as cross-service calls, data transmission, access control, online vulnerabilities, and cluster fault tolerance, it assesses the software architecture adaptability, module coupling, interface compatibility, data transmission security, high-risk vulnerabilities, cluster abnormal fault tolerance, and long-term stable operation risks of the production platform. The AI ​​model reasoning module is equipped with a multi-level AI reasoning model, which performs feature extraction and in-depth analysis on the collected first-level, second-level, third-level, and fourth-level hierarchical operation assessment data, locates the software risk level, the root cause of the failure and the precise impact point, and outputs the hierarchical operation risk assessment result. Based on the hierarchical operational risk assessment results, the execution scheduling module outputs start / stop scheduling instructions and risk elimination instructions for each module, driving the execution of risk governance and software optimization actions at the corresponding level.

2. The system according to claim 1, characterized in that, The main control unit sets the pre-processing module for front-end operational data and the surface code specification assessment module to run continuously 24 / 7, and sets the business logic risk assessment module, operational performance resource assessment module, architecture security operation assessment module, AI model inference module and execution scheduling module to be triggered on demand.

3. The system according to claim 2, characterized in that, The surface code standardization analysis module pulls the structured code and basic data points output by the pre-processing module in real time and transmits them synchronously to the AI ​​model inference module. The AI ​​model inference module calls the first-level code compliance inference model to perform full-text code syntax, format, and compilation parameter compliance verification in combination with the platform's coding standards and compilation baseline. If all indicators are compliant, the software is determined to have no surface-level operational risks, and the entire assessment process is terminated by the execution scheduling module. If any indicator is abnormal, it is determined that there is a first-level surface-level operational risk, and the scheduling module rigidly triggers the second-level business logic analysis process. The business logic risk assessment module takes the results of the first-level anomaly assessment and initiates the second-level assessment. It collects platform business link logs, branch execution points, and upstream and downstream input and output matching operation data and sends them to the AI ​​model inference module. The second-level business logic inference model completes the business risk assessment. If all operational indicators at this level are normal, it is determined that there is only a first-level surface risk. The execution scheduling module terminates the progressive assessment process and executes the first-level lightweight optimization solution. If the indicators are abnormal, it is determined that there is a level 2 business operation risk, which rigidly triggers the level 3 performance resource assessment process; The performance resource assessment module initiates a third-level assessment based on the results of the second-level anomaly assessment. It collects operational indicators such as platform service execution time, cluster memory usage, resource reclamation logs, interface call frequency, and thread pool running status, and sends them to the AI ​​model inference module. The third-level performance resource inference model completes the performance loss risk assessment. If all indicators at this level are compliant, it is determined that there is only a second-level business logic risk. The execution scheduling module then terminates the progressive assessment process and executes the second-level logic optimization plan. If the indicators are abnormal, it is determined that there is a level 3 performance resource operation risk, which rigidly triggers the level 4 architecture security final assessment process. The architecture security operation assessment module takes the results of the third-level anomaly assessment and initiates the fourth-level assessment. It collects data such as platform module dependency coupling data, cross-service interface compatibility logs, encrypted records of user data transmission, vulnerability scanning operation reports, cluster fault tolerance alarms, and access permission logs and sends them to the AI ​​model inference module. The fourth-level architecture security inference model completes the systemic platform operation risk assessment. If all indicators at this level are compliant, it is determined that there is only a level 3 performance resource risk, and the execution scheduling module terminates the progressive analysis process and executes the level 3 performance tuning plan; if the indicators are abnormal, it is determined that there is a level 4 architecture security high-risk operation risk, and the execution scheduling module initiates a level 4 architecture reconstruction and security hardening optimization plan.

4. The system according to claim 3, characterized in that, During the level 1, level 2, level 3, and level 4 tiered operation analysis process, the criteria for determining normal / abnormal data at each level are as follows: First-level surface analysis judgment criteria: Based on the platform's unified coding standards and code compilation baseline, if the source code has no syntax errors, complete variable definitions, unified code formatting and indentation, no redundant or invalid code, and the compilation pass rate meets the standard, and there are no abnormal collection points, then the data is judged to be normal and there is no surface operation risk. If any of the following are found: syntax error, undefined variable, disordered format, redundant empty code, compilation failure, or missing basic data tracking points, the data is considered abnormal and there is a first-level operational risk. Secondary business logic assessment criteria: Based on platform business process specifications, full-link branch coverage of the operational baseline, if there are no logical dead zones in the business link, the branches are executed completely, the upstream and downstream input and output parameters match, the business function output meets online operational expectations, and there are no high-frequency business error logs, then the data is judged to be normal and there is only a primary surface-level risk; if any one of the following exists, such as business logic vulnerabilities, missing functions, branch failures, parameter mismatches, or high-frequency business anomaly alarms, then the data is judged to be abnormal and there is a secondary business operation risk. Level 3 performance resource assessment criteria: Based on platform-preset performance thresholds and cluster resource management specifications, if the average service execution time, node memory usage, connection pool resource reclamation, third-party interface call frequency, and thread pool blocking rate are all within the compliant range, and there are no operational alarms related to memory leaks, circular redundancy, service timeouts, or resource preemption, then the data is considered normal and only there is a Level 2 business logic risk; if any one of the following exists, then the data is considered abnormal and there is a Level 3 performance resource operational risk. The Level 4 architecture security assessment criteria are as follows: Based on the platform's microservice architecture design specifications, network security protection standards, and cluster stable operation requirements, if the module dependency coupling is compliant, cross-service interface is compatible, user sensitive data transmission is fully encrypted, there are no high-risk vulnerability alerts, the anomaly tolerance mechanism is complete, interface permission verification is fully covered, and the cluster has been running for a long time without downtime risk, then the data is judged to be normal and there is only a Level 3 performance resource risk; if any of the following are present: chaotic architecture layering, high module coupling, cross-version interface incompatibility, plaintext data transmission, high-risk injection / privilege escalation vulnerabilities, insufficient cluster fault tolerance, or missing permissions, then the data is judged to be abnormal and there is a Level 4 architecture security high-risk operation risk.

5. The system according to claim 1, characterized in that, The specific software closed-loop optimization solutions for Level 1 surface risks, Level 2 business logic risks, Level 3 performance resource risks, and Level 4 high-risk architecture security risks are as follows: Level 1 Surface-Level Operational Risk Optimization Solution: For surface-level issues such as syntax errors, messy code formatting, missing variables, redundant code, and abnormal data collection, the system performs fully automated lightweight rectification and optimization; the AI ​​model inference module automatically locates abnormal code and data collection points, unifies code formatting, cleans up redundant fragments, completes missing variables, fixes compilation errors, and completes missing basic data collection points. After optimization, it automatically performs compilation verification and data collection connectivity testing to ensure that the code can be compiled normally and that basic operational data is collected completely, without changing the business logic and software architecture throughout the process; Level 2 Business Operation Risk Optimization Solution: For mid-level risks such as business logic dead zones, invalid branches, parameter mismatches, business function deviations, and high-frequency business errors, the system performs iterative optimization of business logic; the AI ​​model inference module combines online business logs to locate abnormal links and failed branches, reconstructs closed-loop business processes, fills in missing business links, corrects upstream and downstream parameter mappings, and eliminates logic dead zones. After optimization, the system automatically replays real business simulation traffic on the platform to conduct full-branch regression testing, ensuring the integrity of online business processes and that outputs meet operational standards. The Level 3 Performance Resource Operation Risk Optimization Solution addresses performance bottlenecks such as service timeouts, memory overflows, resource leaks, redundant loops, frequent API calls, thread blocking, and cluster load imbalances by optimizing system performance resources. The AI ​​model inference module streamlines redundant loops based on operational time consumption and resource metrics, reconstructs inefficient execution chains, adds an automatic resource recycling mechanism, optimizes API call merging strategies, unblocks threads, and balances cluster load. After optimization, it automatically initiates high-concurrency stress tests on the platform to verify that various cluster performance metrics return to compliance thresholds. The Level 4 architecture security high-risk operation risk optimization solution addresses systemic platform risks such as chaotic architecture layering, high module coupling, incompatible cross-service interfaces, data transmission leakage, high-risk online vulnerabilities, lack of cluster fault tolerance, and ineffective access control. The system undergoes deep architecture reconstruction and operational security hardening optimization. The AI ​​model inference module is reconstructed into a microservice layered architecture, highly coupled modules are separated, cross-service interface adaptation standards are unified, data transmission encryption and access authentication mechanisms are improved, online vulnerabilities such as injection / privilege escalation are fixed, and the cluster fault degradation and fault tolerance system is enhanced. After optimization, multi-version compatibility testing, security penetration testing, and a 72-hour long-term cluster stability test are performed to completely eradicate systemic high-risk operation risks.

6. The system according to claim 5, characterized in that, The system is configured with a hierarchical operation closed-loop verification mechanism. After the risk optimization at each level is completed, the corresponding special regression verification is performed: after the first level optimization, code compilation review and data point connectivity verification are performed to confirm that the risks of surface code and basic data collection are completely eliminated; after the second level optimization, the platform's real business traffic is replayed to build a simulation scenario and full-link business regression test is performed to confirm that there are no abnormalities in the business process and online errors are cleared. After Level 3 optimization, cluster stress testing is automatically initiated to verify that all performance indicators, such as service time, memory, resources, threads, and load, have returned to the compliant range. After Level 4 optimization, cross-version compatibility testing, security penetration testing, and long-term cluster stability testing are carried out in sequence to confirm that the software architecture is reasonable, there are no security vulnerabilities, and the cluster is running stably. If the verification is successful, the risk closure loop is completed and archived, and all operational assessment logs and optimization records are retained. If the verification fails, the corresponding level of assessment and optimization process is retried until all verifications are successful.

7. A method for intelligent assessment and closed-loop optimization of software risks based on platform operation data, implemented based on the system described in any one of claims 1-6, characterized in that, The specific process includes: S1. The system gathers multi-source operational data in real time, including platform source code, full-link business logs, interface tracking points, cluster resource indicators, and security vulnerability alarms. Through the pre-processing module for operational data, it completes data cleaning, format unification, structured slicing, and dirty data filtering, and outputs standardized and parsable operational datasets. S2. The execution scheduling module continuously initiates the first-level surface code standardization judgment process. The surface code standardization judgment module, in conjunction with the first-level AI inference model, completes the code and basic tracking compliance verification. Based on the judgment results, it determines whether to directly terminate the entire process or rigidly trigger the second-level business logic judgment. S3. Strictly follow the step-by-step rule. Only after identifying abnormal risks at the previous level, initiate the second-level business logic assessment, the third-level performance resource assessment, and the fourth-level architecture security final assessment in sequence. There is no skipping or over-level execution. Deeply explore all dimensions of software operation risks from surface code, middle business, underlying performance to global architecture security. S4, the AI ​​model reasoning module receives operational assessment data from each level, calls the corresponding level's dedicated reasoning model to complete risk feature extraction, root cause analysis, and impact scope determination, and outputs a graded operational risk report to clarify the risk level, fault location, and online impact scope. S5. The execution scheduling module calls the matching level-specific optimization solution based on the hierarchical risk report, and sequentially completes lightweight code standardization, business logic iteration, cluster performance tuning, architecture reconstruction and security hardening. S6. After each level of optimization is completed, the corresponding special closed-loop regression verification is initiated. After the verification is passed, the entire process data of this operation risk is archived to form a full-link intelligent closed-loop management and control of platform software risk, namely "data aggregation - hierarchical analysis - AI root cause location - hierarchical optimization - regression verification - archiving and accumulation".

8. An electronic device, characterized in that, include: A processor, a memory, and a communication bus, wherein the processor and the memory communicate with each other via the communication bus; The memory stores program instructions that can be executed by the processor. The processor can call the program instructions to execute the software risk intelligent assessment and closed-loop optimization method based on platform operation data as described in claim 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed on a computer, perform the software risk intelligent assessment and closed-loop optimization method based on platform operation data as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or the instructions are executed by the computer, they implement the software risk intelligent assessment and closed-loop optimization method based on platform operation data as described in claim 7.

Images