A method and system for application security testing capability acceptance and upgrade control based on a closed world of sample scenarios.

By constructing a closed-world detection method for example scenarios, combined with an upgrade controller and a pseudo-acceptance rejection mechanism, the problems of false alarms and false alarms in static application security detection are solved, achieving a detection effect of zero false alarms and zero false alarms, and ensuring the auditability and reproducibility of detection capabilities.

CN123113593APending Publication Date: 2026-11-17BEIJING ZHONGSOFT INTELLIGENT CONTROL INFORMATION TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611467271.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

When existing static application security testing products are deployed in industry, there are problems of false alarms rising and false alarms falling back, and there is a lack of effective scenario-based acceptance and upgrade control methods, resulting in unclear testing capability boundaries and an inability to achieve zero false alarms and zero false alarms.

Method used

We adopt a closed-world detection method based on sample scenarios. By constructing a scenario envelope and an upgrade controller, we only verify false positives and false negatives for locked scenarios, enforce the Docs-first upgrade process, and implement a veto mechanism in the pseudo-acceptance method to ensure the accuracy of detection capabilities.

Benefits of technology

It achieves a detection effect of zero false negatives and zero false positives without relying on the complete corpus clearing and proper noun suppression, while avoiding the misleading effects of pseudo-acceptance methods and ensuring the auditability and reproducibility of detection capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113593A_ABST
    Figure CN123113593A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for acceptance and upgrade control of application security detection capabilities based on closed world sample scenarios. A machine-readable scenario envelope is created for each detection rule, recording the scenario identifier, entry / taint / guard / sink tokens, and hit / quiet pairs. The detection capability boundary is limited to locked scenarios; outside of these scenarios, the rule is considered correct by default and cannot drive expansion engines. The sole acceptance criterion is zero false positives and zero missed false positives in locked meshes, requiring that the rule remain green even after renaming. Before upgrade, four fields must be entered into the disk query: class, exit line, and modification record. If these are not present, modification of the judgment code is prohibited. A rejection table is used to block actions that clear all false positives, suppress proper names, allow clause number ranges, or scan method body lines to green as acceptance. This invention does not involve syntax tree disk entry contracts, layered detection pipelines, or large model gates themselves.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the interdisciplinary technical field of application security testing engineering and software quality assurance. Specifically, it relates to a method and system for accepting and judging code changes based on the closed world of a sample scenario as the capability boundary and the control rules of machine-readable scenario envelopes and Docs-first liftgates. Background Technology

[0002] When application security testing products, such as static application security testing, are deployed in industry, they generally face the problem of "impressive alarm numbers in the corpus" and "unclear boundaries of actual capabilities." Existing practices often use methods such as clearing all alarms from the entire corpus, suppressing proper names, allowing clause number ranges, or scanning method body text lines to quickly obtain "test pass" results. This leads to a rebound in false alarms after switching projects, or missed alarms in similar scenarios.

[0003] Another type of problem is the disordered upgrade of rules: the judgment code is modified first, the scene documentation is added later, and even non-zero hits are created by using wiring post code to pretend that it has been fixed. There is a lack of executable control methods that take "zero missed reports and zero false reports in locked scene meshes" as the sole acceptance criterion and make document storage a prerequisite for modifying the engine.

[0004] It should be noted that existing technologies address the issue of the detection execution architecture by implementing a unique source for the syntax tree, isolating detection paths, using a layered detection pipeline, and establishing a large model post-verification gate. This invention does not reiterate these architectural features but focuses on the scenario-based acceptance testing of detection capabilities and the shutdown control for upgrade failures. Summary of the Invention

[0005] The technical problem to be solved by this invention is: how to define the detection capability boundary with a machine-readable scenario envelope without relying on clearing the entire corpus and suppressing proper names, force Docs-first upgrade, and shut down the failure of pseudo-acceptance methods, so as to achieve zero false positives and zero false negatives in locked sample scenarios and still be valid even with name changes.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution.

[0007] Firstly, a method for acceptance and upgrade control of application security detection capabilities based on closed world sample scenarios is provided: A scenario envelope is constructed containing scenarioId and entry / taint / guard / repository tokens and hit / quiet pairs; only locked scenarios are counted for false positives and false negatives, while external scenarios are considered correct by default and the driver expansion engine is prohibited; before upgrade, four fields—query class, exit line, and modification record—are verified; full clearing, name suppression, number segment release, and method body line scanning are rejected; locked meshes are accepted and renamed for verification; only after all passes are upgrade permission issued.

[0008] Secondly, a corresponding system is provided, including a scene envelope library, rule documents and machine library, decision code repository, and closed world acceptance and upgrade controller.

[0009] The beneficial effects of this invention are as follows:

[0010] (1) The detection capability is reduced from “corpus data” to “locked scene grid”, and the acceptance is auditable and reproducible.

[0011] (2) The default is correct outside the scenario, blocking the blind expansion and recall driven by the corpus not included in the envelope.

[0012] (3) The Docs-first state machine avoids engine-first and green-first-later document-filling.

[0013] (4) The fake acceptance rejection form intercepts the green number segment, green line scan and special name suppression accuracy from the mechanism.

[0014] (5) Rename verification ensures that the judgment lock is based on the scene structure rather than the business name.

[0015] (6) Decouple from the detection execution architecture and large model gate to avoid overlapping with the protection scope of existing architecture-type inventions. Attached Figure Description

[0016] To more clearly illustrate the embodiments of the present invention, the accompanying drawings will be briefly described below.

[0017] Figure 1 System overall architecture UML component diagram.

[0018] Figure 2 Example scenario: UML package diagram of envelope and capability boundary.

[0019] Figure 3 Docs-first UML state machine diagram for upgrade control.

[0020] Figure 4 Locked grid acceptance UML activity diagram.

[0021] Figure 5 The pseudo-acceptance rejection and failure closes the UML activity diagram.

[0022] Figure 6 The overall steps of the method are illustrated in the UML activity diagram.

[0023] Figure 7 The system's horizontal composition is shown in the UML component diagram.

[0024] Figure 8 System flowchart.

[0025] In the diagram: 101 – Scene Envelope Library; 102 – Rule Document and Machine Library; 103 – Decision Code Repository; 104 – Closed World Acceptance and Upgrade Controller; 105 – Output Interface; 105a – Upgrade Permit; 105b – Engine Modification Prohibited. All diagrams are drawn according to UML 2 specifications. Figure 1 , Figure 7 For component diagrams; Figure 2 For the package diagram; Figure 3 State machine diagram; Figure 4 , Figure 5 , Figure 6 This is an illustration of an activity. Figure 6 This is a vertical activity diagram (including initial / termination nodes, action nodes, and decision nodes). Figure 7 This is a horizontal component diagram. The abstract diagram uses a vertical S1–S5 step-by-step block sequence, independent of the specification diagram and not related to it. Figure 6 Activity diagram or Figure 7 Mixed use of component diagrams. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the invention.

[0027] Example 1 System Composition See Figure 1 The system includes a scene envelope library 101, a rule document and machine library 102, a judgment code repository 103, and a closed-world acceptance and upgrade controller 104. Controller 104 has read-only access to 101 and 102, and can only authorize changes to the code in 103 that matches the registered exit line after an upgrade permit is output. Output 105 contains an upgrade permit, a prohibition on engine modification instructions, or an acceptance report with a defect list.

[0028] Example 2: Scenario Envelope and Capability Boundary See Figure 2 For rule R, write the scenario envelope Ei. Example fields include: scenarioId = external_entry_to_sink_unguarded; Entry token = isExternalEntry; Taint token = ExternalScopeReach; Guard token = isGuardEffective = false; Sink token = standard API sink token. Hits must be detected at the sink; quiet isomorphic twins only change the guard or reachability to make the polarity negative. Code structures not written to the envelope are not counted as missed reports in this round and cannot be used as a reason to expand the engine.

[0029] Example 3: Docs-first Upgrade State Machine See Figure 3 Upon receiving a request to modify the judgment, the controller checks in sequence: whether the scene envelope is locked; whether the query class and the hole exit line are written to the machine; and whether the four fields of the modified record are written to disk. If any step fails, the engine modification is prohibited. Example of a query class: taint reachability analysis class; example of a hole exit line: sink failure closing predicate name. Method body scan classes or clause number range release classes are prohibited from being registered as valid hole exit lines.

[0030] Example 4: Fake Acceptance Rejection See Figure 5 If the submitted materials claim "all corpus alarms have been cleared, therefore approval is granted," or demonstrate a business method name suppression table, allow access by clause number range, or obtain non-zero hits using method body regular expression stubs, the controller will hit the rejection table and fail to close. It should be emphasized that this invention rejects the use of the above methods as a means to pass scene grid acceptance; it does not advocate or change other existing controls of the external engine regarding the original source code scanning path.

[0031] Example 5: Locked Grid Acceptance and Renaming See Figure 4 For each locked scenarioId, perform hit and quiet acceptance tests in parallel or serially. Then, rename the business identifier using a bijective while retaining the standard API name. After rerunning, hit must still be detected, and quiet must still be silenced. Only after all scenarios pass can you enter the constitutional modification engine window that matches the already deployed disk exit line; after modification, you must run the mesh again to prevent fixing A from damaging B.

[0032] Example 6: Decoupling from External Architecture See claim 10 and Figure 1 This system can be deployed as an independent upgrade access control within a continuous integration pipeline. Regardless of whether the external static engine uses a specific intermediate representation for disk loading or a hierarchical pipeline, as long as it can provide detection / quiet results for hit / quiet examples, it can be invoked by the controller. This invention exclusively excludes syntax tree unique source contracts, hierarchical hard boundaries, or large model gate patterns.

[0033] Example 7: Overall Method Flow See Figure 6 The overall steps are as follows: S1 Constructing the envelope → S2 Defining the capability boundary → S3 Docs-first gate → S4 Pseudo-acceptance rejection → S5 Mesh and renaming acceptance → S6 Outputting permission or prohibition. This order must not be reversed: it is forbidden to modify the judgment code before filling in the envelope, and it is also forbidden to skip S5 using evidence rejected in S4.

[0034] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for acceptance testing and upgrade control of application security detection capabilities based on a closed world of sample scenarios, characterized in that, The method is applied to a system consisting of a scene envelope library, rule documents and machine libraries, a decision code repository, and a closed-world acceptance and upgrade controller; the method includes: S1. Build or update a machine-readable scenario envelope set for at least one application security detection rule; each scenario envelope includes at least a scenario identifier (scenarioId), an entry token, a taint token, a guard token, a sink token, and a set of hit samples and a set of quiet isomorphic twin samples bound to the scenario identifier. S2. The capability boundary is defined by the closed world acceptance and upgrade controller: only locked scene envelopes are included in the statistics of missed and false alarms; code structures that do not fall into any locked scene envelopes are judged as correct code by default, and it is prohibited to trigger the expansion and recall modification of the judgment code repository based on this. S3. Execute Docs-first upgrade gate: Before allowing modification of the decision code repository, verify whether the rule document and machine library have been written to disk the four fields corresponding to this change: query class, exit line class name or predicate name, and modification record; the four fields must include at least the reason, solution, hot issue, and rollback of previous explanation; if any is missing, output the command to prohibit engine modification; S4. Perform pseudo-acceptance rejection on the proposed acceptance evidence: If any of the following is used as the basis for acceptance: clearing all alarms in the full corpus, suppressing business names, allowing clause number segments or path tainted predicate number segments, scanning method body lines or greening wiring post codes, then the application will fail, be closed, and upgrade will be prohibited. S5. Perform grid acceptance for each locked scenario identifier: the hit sample set must have zero false alarms, and the quiet isomorphic twin sample set must have zero false alarms; and perform randomization and renaming on the business identifiers in the samples, while keeping the polarity unchanged; if any scenario fails, upgrading is prohibited. S6. If and only if all of S3 to S5 pass, output upgrade permission and allow constitutional modification of the predicate in the decision code repository that is consistent with the already dropped plate exit line; otherwise, output prohibition of engine modification instruction and defect list.

2. The method according to claim 1, characterized in that, The quiet isomorphic twin in step S1 is isomorphic to the corresponding hit twin in terms of nested structure and sink token family. The only difference is the polarity field in scenarios such as guard establishment or taint unreachability. It is prohibited to construct quiet instances by using empty classes, deleting sinks, or changing rule identifiers.

3. The method according to claim 1, characterized in that, The query class mentioned in step S3 is one of the following: taint reachability analysis class, syntax structure attribute class, configuration domain class, or graph predicate class; if the method body line scan class or clause number segment release class is to be used as the exit line, then the Docs-first upgrade gate will fail and close directly.

4. The method according to claim 1, characterized in that, The clause number range release mentioned in step S4 includes at least the behavior of using the vulnerability clause number range or the path taint predicate to cover the number range as the main signal for exiting the vulnerability or releasing the vulnerability; the method body line scan includes at least the behavior of applying regular expression matching to the concatenated method body text as the main signal for exiting the vulnerability.

5. The method according to claim 1, characterized in that, In step S5, during the renaming randomization, the standard application interface name and annotation name of the language or framework are preserved, and bijective replacement is performed on the local variable name, business class name, and business method name; if the polarity is reversed after renaming, the acceptance of this scenario fails.

6. A test and upgrade control system for application security testing capabilities based on a closed world of sample scenarios, characterized in that, For implementing the method of any one of claims 1 to 5, comprising: Scene envelope library, used to store machine-readable scene envelopes and their hit / quiet pairs; The rules document and machine library are used to store four fields: query class, hole line, and modification record. The decision code repository is used to store the implementation of the exit predicate; The Closed World Acceptance and Upgrade Controller is used to perform capability boundary delineation, Docs-first upgrade gate, pseudo-acceptance rejection, locked grid acceptance and renaming review, and output upgrade permission or prohibition of engine modification commands.

7. The system according to claim 6, characterized in that, The closed world acceptance and upgrade controller maintenance audit log records the scene envelope hash, document hash, rejection hit item and grid acceptance result corresponding to each upgrade request. The audit log is not written back as a new exit condition.

8. The system according to claim 6, characterized in that, The system also includes a single-outlet registration form; when the same detection rule registers two or more launchable paths at the same time, or when an old path is not marked as decommissioned but can still be launched, the controller determines that the upgrade has failed.

9. The system according to claim 6, characterized in that, The scenario envelope library supports generating four types of samples—direct, loop, lambda, and static_block—based on a nested structure, and pairs them with hit / quiet; Scene identifiers for incomplete grids must not be marked as locked.

10. The system according to claim 6, characterized in that, The closed-world acceptance and upgrade controller is decoupled from the external static detection engine: it does not declare or change the unique source disk contract of the syntax tree, the hard boundary of the layered detection pipeline, or the presence and mode of the large model's post-gate; it only controls the capability acceptance and upgrade permission.