A Security Assessment Method for Cryptographic Applications in IoT Systems Based on a Large Model

By adopting a security assessment method for cryptographic applications in IoT systems based on large models, the problems of automation and standardization in the security assessment of cryptographic applications in industrial IoT scenarios are solved. This method enables the traceability of assessment conclusions and the identification and elimination of logical vulnerabilities, thereby improving the reliability and accuracy of the assessment.

CN122316730BActive Publication Date: 2026-11-17北京合泰信安信息技术有限公司
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202610446651.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-04-07
Publication Date
2026-11-17
Estimated Expiration
2046-04-07

AI Technical Summary

Technical Problem

In industrial IoT scenarios with multi-vendor equipment in mixed networking, existing technologies struggle to automate and standardize compliance assessments of cryptographic application security in complex heterogeneous devices and multi-protocol environments. This results in assessment results that cannot be reproduced or verified, and a closed-loop, verifiable chain of evidence cannot be formed.

Method used

This paper adopts a security assessment method for cryptographic applications in IoT systems based on a large model. It constructs asset identifiers by acquiring target asset information, collects heterogeneous evidence and calculates hashes to generate evidence indexes, integrates cryptographic compliance standards into a control item library, uses a domain-pre-trained large model for semantic feature vector matching, combines deterministic verifiers and active probing scripts for verification and correction, generates target assessment results, and outputs a visual report.

Benefits of technology

It enables full auditability of the evaluation process, ensures one-click traceability between evaluation conclusions and original physical evidence, identifies and eliminates logical loopholes, improves the objectivity and effectiveness of evaluation conclusions, and solves the problems of evidence fragmentation and broken evidence chains.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122316730B_ABST
    Figure CN122316730B_ABST
Patent Text Reader

Abstract

This invention relates to the field of cryptographic application evaluation technology, and discloses a method for cryptographic application security evaluation of Internet of Things (IoT) systems based on a large model. The method includes: acquiring target asset information to construct asset identifiers and determining evaluation boundaries based on network topology; collecting heterogeneous evidence within the evaluation boundaries and generating an evidence index by calculating hashes based on the heterogeneous evidence; integrating the cryptographic compliance standards corresponding to the asset identifiers into a control item library, the control item library including compliance entries, and converting the compliance entries into compliance semantic feature vectors; and converting the heterogeneous evidence into evidence semantic feature vectors. This invention employs a heterogeneous evidence hash encapsulation and evidence index intrinsic association technology to achieve full auditability of the evaluation process, enabling one-click traceability from the evaluation conclusion to the original physical evidence, and addressing the shortcomings of existing technologies such as evidence fragmentation, lack of verifiability, and broken evidence chains.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of cryptographic application evaluation technology, specifically to a method, device, and medium for cryptographic application security evaluation of Internet of Things (IoT) systems based on a large model. Background Technology

[0002] In the context of industrial IoT scenarios with multi-vendor equipment in a mixed network, the increasing demands of critical information infrastructure for commercial cryptographic compliance have made it crucial to address the novel and urgent technical challenges in ensuring the secure operation of the industrial internet. This is due to the need for automated and standardized compliance assessments of cryptographic applications in complex heterogeneous devices and multi-protocol environments.

[0003] Currently, the evaluation of cryptographic applications in the aforementioned scenarios mainly relies on manual interviews, on-site configuration verification, and automated scanning tools based on fixed rule bases. Technical personnel typically use a pre-set compliance checklist to manually extract device firmware information, capture network traffic packets, and retrieve system logs to obtain evaluation evidence. This evidence is then combined with traditional static logical matching algorithms to logically determine the compliance of cryptographic algorithms, the security of key management, and the correctness of protocol implementation, ultimately generating an evaluation report.

[0004] While existing technologies can perform basic compliance checks, they generally suffer from deep-seated flaws such as broken cross-layer evidence chains and untraceable assessment conclusions. Because the firmware version, configuration status, and certificate information of industrial IoT devices change dynamically with maintenance operations, existing static evidence collection methods can retain fragmented descriptive conclusions or isolated screenshot evidence, but lack unified semantic encapsulation and intrinsic correlation of key original evidence. This makes the assessment process like a castle in the air; once the environment shifts even slightly, the original assessment results cannot be reproduced or verified, making it difficult to support continuous security audits and liability determination, and failing to form a closed-loop, verifiable evidence chain. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for security assessment of cryptographic applications in Internet of Things (IoT) systems based on large-scale models, thereby at least partially solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a security assessment method for cryptographic applications in Internet of Things (IoT) systems based on a large-scale model, wherein the security assessment method for cryptographic applications in IoT systems includes:

[0007] Obtain target asset information to construct asset identifiers, and determine the assessment boundaries based on network topology;

[0008] Collect heterogeneous evidence within the evaluation boundary, calculate a hash based on the heterogeneous evidence to generate an evidence index, and associate the evidence index with the heterogeneous evidence;

[0009] The cryptographic compliance standards corresponding to the asset identifiers are integrated into a control item library, which includes compliance items. The compliance items are then converted into compliance semantic feature vectors.

[0010] The heterogeneous evidence is transformed into an evidence semantic feature vector, which is then matched with the compliance semantic feature vector to generate a preliminary evaluation result.

[0011] The preliminary evaluation results are verified and corrected to obtain the target evaluation results.

[0012] Preferably, the step of collecting heterogeneous evidence within the evaluation boundary and generating an evidence index by calculating hashes based on the heterogeneous evidence includes establishing an evidence time-series chain, performing consistency comparisons on the heterogeneous evidence collected at different timestamps, and identifying and recording the dynamic changes in the evidence index caused by device firmware upgrades or configuration tampering.

[0013] The heterogeneous evidence includes device firmware images, system configuration files, network communication traffic fragments, and security operation and maintenance logs. The evidence index and the physical storage path of the heterogeneous evidence have a one-to-one mapping relationship in the evidence library.

[0014] Preferably, the step of converting the compliance entry into a compliance semantic feature vector is to use a domain pre-trained large model to extract entities from the compliance entry, identify the cryptographic algorithm type, key length and authentication protocol involved in the compliance entry, and calculate the dimensional weights of the compliance semantic feature vector according to the importance of the entity.

[0015] The process of converting the heterogeneous evidence into an evidence semantic feature vector and matching it with the compliant semantic feature vector to generate a preliminary evaluation result involves pre-setting evidence slots in the prompt word template, forcing the preliminary evaluation result generated by the large model to include and associate the corresponding evidence index. If the generated judgment text does not contain a valid evidence index tag, the reasoning result is determined to be invalid and an automatic re-examination mechanism is triggered.

[0016] Preferably, the step of verifying and correcting the preliminary evaluation result to obtain the target evaluation result includes: inputting the preliminary evaluation result into a deterministic verifier for verification. The deterministic verifier includes a certificate compliance parsing engine, a cryptographic algorithm fingerprint recognition module, and a protocol handshake status monitor, which are used to verify the objectivity and authenticity of the preliminary evaluation result from the physical feature level.

[0017] If the preliminary evaluation result conflicts with the physical rule parsing result of the deterministic verifier, the evaluation confidence level is reset.

[0018] Preferably, the step of verifying and correcting the preliminary evaluation result to obtain the target evaluation result further includes: triggering an active probe script to perform protocol verification. The active probe script includes a cryptographic suite exhaustive operator and a protocol degradation stress test module, which is used to initiate interactive probes on the target device in a controlled environment to obtain real-time feedback on the dynamic operating status.

[0019] The preliminary evaluation results are corrected based on the detection results of the active detection script to eliminate conflicts, and the final target evaluation result is obtained.

[0020] Preferably, after obtaining the target assessment result, the method further includes: combining weights, violation indicators, and confidence scores to perform risk diffusion analysis using the asset identifier, including:

[0021] Extract the asset identifier of the risky device, automatically index existing assets with the same hardware model and firmware version within the assessment boundary, and simultaneously mark potential compliance risks.

[0022] Preferably, the target assessment result includes a comprehensive risk score, which is calculated using the following formula:

[0023] ,

[0024] in, As a comprehensive risk score, To control the number of items, For control items The weight, This is the violation determination coefficient. The confidence score is the result of feedback from the deterministic validator. This is the index number of the control item;

[0025] The violation determination coefficient The value is dynamically determined by the detection results of the active detection script. When a protocol downgrade vulnerability or the use of an unauthorized algorithm is detected in the device, the violation determination coefficient is adjusted. It is assigned the highest weight level.

[0026] The confidence score fed back by the deterministic verifier Controlled by the integrity verification status of the heterogeneous evidence, if the hash verification of any original evidence in the digital signature evidence package fails, the confidence score of the corresponding control item will be... It was automatically set to zero.

[0027] Preferably, after obtaining the target assessment result, the method further includes: outputting an assessment report, which displays the assessment conclusion, evidence index, and rectification suggestions in conjunction with the report;

[0028] The associated display assessment conclusions, evidence index, and rectification recommendations include:

[0029] The visualization interface provides hyperlinks based on the evidence index, supporting one-click tracing from the evaluation conclusion to the corresponding original hexadecimal traffic data or configuration file line number in the digital signature evidence package, thus achieving closed-loop verification of the evaluation results.

[0030] An electronic device, comprising:

[0031] A processor; and a memory storing program instructions that, when executed by the processor, cause the electronic device to perform any one of the methods described above.

[0032] A storage medium, which is a computer-readable storage medium, stores computer-readable instructions thereon that, when executed by one or more processors, implement any one of the methods described above.

[0033] This invention provides a method for security assessment of cryptographic applications in Internet of Things (IoT) systems based on a large-scale model. It offers the following advantages:

[0034] 1. This invention adopts a heterogeneous evidence hash encapsulation and evidence index endogenous association technology solution to achieve the technical effect of full auditability of the evaluation process, realize one-click traceability from the evaluation conclusion to the original physical evidence, and solve the shortcomings of existing technologies such as evidence fragmentation, inability to verify and broken evidence chain.

[0035] 2. This invention directly extracts the physical operating characteristics of IoT devices through a deterministic verifier, no longer relying on self-declared documents or configuration files of the devices. Instead, it obtains "immutable truth values" from the original data level such as encrypted protocol handshakes and digital certificate parameters, providing a physical reference standard for semantic reasoning of large models. By using an active probing script to simulate attacker behavior in a controlled environment, it actively initiates interactive probing to the target device. This aims to verify whether the device will illegally degrade when faced with weak encryption requests or interference, thereby obtaining real-time security feedback of the device in dynamic operating states. This compensates for the shortcomings of static evaluation in failing to discover logical vulnerabilities, achieving the effect of real-time alignment between semantic inference and physical operating states. It also automatically eliminates logical conflicts between configuration declarations and the real environment, solving the shortcomings of existing evaluation methods in failing to identify operational degradation and the poor authenticity of evaluation conclusions. Attached Figure Description

[0036] Figure 1 A schematic flowchart illustrating the cryptographic application security assessment method for IoT systems based on a large model, provided in an embodiment of the present invention;

[0037] Figure 2This diagram illustrates a comparison between the security assessment method for cryptographic applications in IoT systems based on a large model, as provided in this embodiment of the invention, and existing technologies. Detailed Implementation

[0038] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0039] The present invention will now be described in detail with reference to the accompanying drawings:

[0040] Please see the appendix Figure 1 This invention provides a method for security assessment of cryptographic applications in Internet of Things (IoT) systems based on a large model. The method includes:

[0041] Obtain target asset information to construct asset identifiers, and determine the assessment boundaries based on network topology.

[0042] Specifically, the steps of acquiring target asset information, constructing asset identifiers, and determining assessment boundaries based on network topology offer the advantage of comprehensively and accurately grasping the true physical characteristics and logical states of all online devices in the IoT environment, preventing any hidden assets from escaping the scope of security oversight. Relying on a structured network topology map, the scope of security reviews for cryptographic applications can be scientifically and rationally defined. This measure effectively prevents the review process from becoming scattered or creating blind spots by overlooking core business nodes. Simultaneously, this operation lays a solid data foundation for building a tight, static binding relationship between physical nodes and logical review units, ensuring that the cryptographic application security analysis process of the IoT system can be steadily and orderly advanced under a clear and well-defined objective.

[0043] Collect heterogeneous evidence within the evaluation boundary, calculate a hash based on the heterogeneous evidence to generate an evidence index, and associate the evidence index with the heterogeneous evidence.

[0044] Specifically, heterogeneous evidence is collected within the evaluation boundary, and hashes are calculated based on this heterogeneous evidence to generate an evidence index. The advantage of this step is that it allows for the direct capture of completely different dimensions of underlying raw operational data from the IoT system's core. This operation provides the most objective, authentic, and unforgeable underlying raw evidence for cryptographic security analysis. By performing rigorous one-way hash calculations on all the massive and complex underlying raw evidence, uniquely linked digital tags can be generated. This process completely locks in the original appearance and timestamp attributes of the underlying raw evidence, preventing any malicious tampering. An unbreakable chain of trust, capable of reverse tracing and verification, is built between the underlying raw evidence and the upper-level analysis results, ensuring the absolute authority and credibility of the security review report.

[0045] The cryptographic compliance standards corresponding to asset identifiers are integrated into a control item library, which includes compliance items. The compliance items are then transformed into compliance semantic feature vectors.

[0046] Specifically, the cryptographic compliance standards corresponding to asset identifiers are integrated into a control item library, and compliance items are transformed into compliance semantic feature vectors. The advantage of this step is that it completely breaks through the technical bottlenecks of inefficiency and susceptibility to subjective misjudgments inherent in traditional manual verification of security checklists. By introducing advanced domain-pre-trained large-scale models, deep entity extraction and weight allocation operations are performed on complex and obscure cryptographic specification clauses. These actions perfectly transform natural language text, which could originally be read and understood by professional security experts, into mathematical vectors that can be directly read, calculated, and compared in a high-dimensional manner by computer systems and artificial intelligence large-scale language models. This transformation endows IoT automated systems with a powerful ability to learn autonomously and deeply understand the connotations of cryptographic security specifications, providing an indispensable benchmark coordinate system for achieving fully automated and highly intelligent machine review of cryptographic application compliance.

[0047] Heterogeneous evidence is transformed into semantic feature vectors of evidence, which are then matched with compliant semantic feature vectors to generate preliminary assessment results.

[0048] Specifically, heterogeneous evidence is transformed into semantic feature vectors, which are then matched with compliance semantic feature vectors to generate preliminary assessment results. The advantage of this step lies in the innovative introduction of cutting-edge retrieval enhancement generation technology, enabling the underlying original evidence and compliance standard clauses to undergo similarity measurement and collision in a high-dimensional mathematical semantic space. This collision allows the large language model to output security and compliance judgments without relying on internal black-box divergent associations, but rather to be strictly constrained to logical reasoning within the framework of the truly extracted underlying original evidence's digital tags. This mechanism completely eliminates the inherent problem of AI large language models easily generating illusions and fabricating conclusions when faced with complex cryptographic security logic. It ensures that the preliminary output security judgment has the advantage of efficient generalized semantic analysis and is supported by solid and reliable underlying objective physical evidence.

[0049] The preliminary assessment results are verified and corrected to obtain the target assessment results.

[0050] Specifically, the preliminary assessment results are verified and corrected to obtain the target assessment result. The benefit of this step is that it constructs an unbreakable security defense for the IoT cryptographic security review system. By introducing a pure physical rule verification engine unaffected by any AI black-box thinking, and a highly adversarial interactive active probe attack script, the system performs rigorous secondary cross-validation and conflict cleansing on the initial judgment conclusions output by the large language model. This verification process accurately identifies and thoroughly eliminates false positives and false negatives caused by logical reasoning flaws or lag in underlying static data. Ultimately, the system forges a final security conclusion that can withstand repeated scrutiny from any cybersecurity audit expert and completely objectively and truthfully reflects the dynamic cryptographic operation status of IoT devices, achieving a crucial leap from probabilistic correctness to absolute correctness in the review results.

[0051] Collect heterogeneous evidence within the evaluation boundary, calculate hashes based on the heterogeneous evidence to generate an evidence index, including establishing a temporal chain of evidence, performing consistency comparisons on heterogeneous evidence collected at different timestamps, and identifying and recording the dynamic changes in the evidence index caused by device firmware upgrades or configuration tampering.

[0052] Heterogeneous evidence includes device firmware images, system configuration files, network communication traffic fragments, and security operation and maintenance logs. The evidence index and the physical storage path of heterogeneous evidence have a one-to-one mapping relationship in the evidence repository.

[0053] Transforming compliance entries into compliance semantic feature vectors involves using a domain-pre-trained large model to extract entities from compliance entries, identifying the cryptographic algorithm type, key length, and authentication protocol involved in the compliance entries, and then weighting the dimensions of the compliance semantic feature vectors according to the importance of the entities.

[0054] Heterogeneous evidence is transformed into semantic feature vectors of evidence and matched with compliant semantic feature vectors to generate preliminary evaluation results. Evidence slots are preset in the prompt word template, and the preliminary evaluation results generated by the large model must contain and be associated with the corresponding evidence index. If the generated judgment text does not contain a valid evidence index label, the reasoning result is deemed invalid and an automatic re-examination mechanism is triggered.

[0055] The preliminary evaluation results are verified and corrected to obtain the target evaluation results. This includes inputting the preliminary evaluation results into a deterministic verifier for verification. The deterministic verifier includes a certificate compliance parsing engine, a cryptographic algorithm fingerprint recognition module, and a protocol handshake status monitor, which are used to verify the objectivity and authenticity of the preliminary evaluation results at the physical feature level.

[0056] If the preliminary assessment results conflict with the physical rule parsing results of the deterministic checker, reset the assessment confidence level.

[0057] The preliminary evaluation results are verified and corrected to obtain the target evaluation results. Further, this includes triggering an active probe script to perform protocol verification. The active probe script includes a cryptographic suite exhaustive operator and a protocol degradation stress test module, which is used to initiate interactive probes on the target device in a controlled environment to obtain real-time feedback on the dynamic operating status.

[0058] The initial assessment results are revised based on the detection results from the active detection script to eliminate conflicts and obtain the final target assessment results.

[0059] After obtaining the target assessment results, further steps include: combining weights, violation indicators, and confidence scores, and utilizing asset identification to conduct risk diffusion analysis, including:

[0060] Extract the asset identifiers of devices with risks, automatically index existing assets with the same hardware model and firmware version within the assessment boundary, and simultaneously mark potential compliance risks.

[0061] The target assessment results include a comprehensive risk score, which is calculated using the following formula:

[0062] ,

[0063] in, As a comprehensive risk score, To control the number of items, For control items The weight, This is the violation determination coefficient. The confidence score is the result of feedback from the deterministic validator. This is the index number of the control item;

[0064] Violation determination coefficient The value is dynamically determined by the detection results of the active detection script. When a device is detected to have a protocol downgrade vulnerability or to be using an unauthorized algorithm, the violation judgment coefficient increases. It is assigned the highest weight level.

[0065] Confidence score returned by the deterministic validator Controlled by the integrity verification status of heterogeneous evidence, if the hash verification of any original evidence in the digital signature evidence package fails, the confidence score of the corresponding control item will be affected. It was automatically set to zero.

[0066] After obtaining the target assessment results, the process further includes: outputting an assessment report, which displays the assessment conclusions, evidence index, and rectification recommendations.

[0067] The associated assessment conclusions, evidence index, and rectification recommendations include:

[0068] The visualization interface provides hyperlinks based on evidence indexes, enabling one-click tracing of evaluation conclusions to the corresponding original hexadecimal traffic data or configuration file line numbers in the digital signature evidence package, thus achieving closed-loop verification of evaluation results.

[0069] Matching with compliant semantic feature vectors, and employing retrieval enhancement generation techniques, semantic similarity is measured using the following cosine similarity formula:

[0070] ,

[0071] in, Evidence semantic vector semantic feature vector of control term The cosine similarity values ​​between them are as follows: For evidence semantic vectors, For the semantic feature vector of the control term, This represents the total number of dimensions of the semantic vector. As a vector-dimensional traversal index variable, when the similarity calculation result exceeds a preset threshold, it is determined that the heterogeneous evidence is valid to support a compliant conclusion.

[0072] Obtain target asset information to construct asset identifiers, and use active fingerprint detection operators to extract IoT devices. Protocol stack fingerprints and application layer services Information, combined with the device firmware version, is used to calculate the asset feature vector, which is then mapped to a unique asset identifier. This identifier is used to establish a static binding relationship between physical assets and logical evaluation units within the evaluation boundary. The static binding relationship supports associating the asset identifier with a specific security baseline in the control item library. When the matching degree between a newly added asset identifier and an existing identifier reaches a preset threshold, the historical evaluation rule template for that type of asset is automatically inherited.

[0073] Target asset information is obtained by actively probing and extracting the fingerprints and service banner data of the underlying transmission control protocol stack of IoT devices. This serves as the most basic analysis object and source data for the cryptographic security assessment system.

[0074] Asset identification is a unique identifier obtained by calculating and mapping the physical device based on the extracted target asset information and device firmware version data. This identifier is used to accurately locate specific devices in a cluttered network environment and establish a static binding association between the physical device and the logical evaluation unit.

[0075] Network topology is obtained by systematically sorting and arranging the physical connection lines and logical communication paths between hardware nodes in an IoT system. It serves to intuitively display the overall communication architecture of the system and provide spatial coordinate guidance for defining the scope of security review.

[0076] The assessment boundary is determined by delineating and isolating the core nodes and related links that require cryptographic security verification based on the network topology. This effectively eliminates interference from irrelevant peripheral network assets and concentrates computing resources on core cryptographic application review.

[0077] Heterogeneous evidence is obtained by aggregating and collecting all device firmware images, system configuration files, network communication traffic fragments, and security operation and maintenance logs generated within the assessment boundary. This serves to comprehensively and realistically restore the objective physical state of the underlying cryptographic operation of IoT devices.

[0078] A hash is calculated by performing irreversible mathematical feature compression on all the collected multi-source heterogeneous raw data using a one-way hash algorithm in cryptography. It serves to give each piece of underlying raw physical data a unique and unforgeable digital fingerprint and ensure that the raw physical data is absolutely tamper-proof.

[0079] Evidence index is obtained by mapping and binding the calculated unique digital fingerprint with the absolute storage path of the underlying original physical data in the database. It plays a role in providing one-click accurate tracing and locating of the underlying hexadecimal original code position during the large language model reasoning and verification process.

[0080] Cryptographic compliance standards are compiled by collecting and summarizing all mandatory regulations and security requirements issued by the state or industry regarding cryptographic applications in IoT systems. These standards serve as the most authoritative basis for judging the compliance and legality of cryptographic applications on target devices.

[0081] The control item library is created by modularly splitting, classifying, and archiving all the complex cryptographic compliance standard texts collected using structured data management methods. This library facilitates computer systems in quickly retrieving and accessing security specification review details from different dimensions, and improves the overall efficiency of compliance comparison.

[0082] Compliance entries are obtained by further breaking down the macro-level compliance check categories within the control item library into micro-level textual description specifications for specific cryptographic algorithms or key lengths. These compliance entries serve as the finest-grained compliance semantic analysis benchmark targets for the large language model.

[0083] The compliance semantic feature vector is calculated by weighting and transforming the algorithm type and key length entities contained in the compliance entry text using a domain-pre-trained large language model. This process transforms natural human language into a digital form that can be precisely measured and compared mathematically by computers and artificial intelligence.

[0084] The evidence semantic feature vector is obtained by inputting the collected heterogeneous evidence original text and code fragments into a neural network extraction model to extract and transform the underlying structured features. This enables the underlying physical original operational data to cross the data form gap and directly collide and measure the mathematical dimension of the compliance standard vector for accurate measurement and comparison.

[0085] The preliminary assessment results, based on the cutting-edge technology of retrieval enhancement, are obtained by calculating the extreme value of cosine similarity between the semantic feature vector of evidence and the semantic feature vector of compliance, and by forcibly binding the original data labels to the output text. This preliminary assessment results play a role in quickly applying the generalized understanding ability of artificial intelligence to provide an initial judgment and conclusion with theoretical data support on the password security status of a massive number of devices across the entire network.

[0086] A deterministic verifier is constructed by combining a certificate parsing engine with an algorithm fingerprint module and a handshake monitor to create a purely physical judgment sandbox that is unaffected by any divergent thinking in artificial intelligence. This deterministic verifier plays a role in strictly verifying the objective truth of logical reasoning conclusions at the underlying network communication and real interaction level, thereby completely eliminating the generation of false judgment results by intelligent machines.

[0087] The active probing script is compiled by integrating the brute-force cipher suite testing module and the protocol degradation stress attack module based on the network attack and defense adversarial thinking. The active probing script plays a role in actively sending probe data packets to the tested IoT device in a controlled and isolated security experimental environment, so as to force the tested device to expose its true underlying security status.

[0088] The target evaluation result is obtained by inputting the initial judgment into the physical engine to perform conflict comparison and correction, and then outputting the final judgment without any illusion of a large language model. This serves to provide security operation and maintenance managers with an authoritative conclusion on cryptographic application security that is absolutely accurate and supported by physical probe data.

[0089] The comprehensive risk score is calculated by weighting and multiplying the base score corresponding to each review control item with the penalty multiplier for violation judgment and the confidence level of the integrity of the evidence chain. It serves to show security auditors the severity level of security threats faced by the cryptographic applications of the entire Internet of Things system in the most intuitive and quantitative numerical indicators.

[0090] The assessment report, generated by the report generation engine, formats and packages the target assessment conclusion text, comprehensive risk score chart, and underlying original data traceability hyperlinks, and embeds a globally tamper-proof digital signature to generate the assessment report. It serves as a graphic and textual guide for system rectification and reinforcement, and supports third-party regulatory agencies in transparently reviewing and verifying all review conclusions.

[0091] Example 1:

[0092] Scenario Background: This embodiment is applied to an IoT system on the distribution side of a provincial power grid, involving smart distribution terminals and gateway devices from multiple manufacturers. The main challenge in this scenario is that frequent equipment maintenance leads to rapid firmware updates. Reports generated by traditional manual sampling inspections often suffer from fragmented evidence and a lack of indexing, making it impossible for auditors to verify compliance conclusions from six months ago, resulting in a serious problem of broken evidence chains.

[0093] Execution steps and procedures:

[0094] Asset fingerprinting and assessment boundary delineation: Utilizing asset identification tools to proactively detect heterogeneous terminals within the distribution network, and extracting... The device's hardware serial number, kernel firmware version number, and based on The protocol stack uses service response fingerprints to calculate and generate unique asset fingerprint vectors, and defines the current substation subnet as the logical evaluation boundary to ensure that all subsequently collected evidence is bound to a unique physical device.

[0095] High-frequency acquisition of heterogeneous evidence and generation of hash indexes: Within the evaluation boundary, the system automatically retrieves the terminal device's runtime configuration file and acquires... The original traffic packets during the handshake phase are exported as system hardening logs, and each piece of collected evidence is executed in real time. Hash operations are performed to generate a globally unique evidence index, and the correspondence between this index and the physical storage path is recorded in the evidence repository.

[0096] Evidence time-series chain construction and signature encapsulation: For configuration changes caused by power distribution terminal firmware upgrades, the system compares evidence indexes under different timestamps, records the dynamic drift process of evidence fingerprints, forms an evidence time-series chain, and finally encapsulates all evidence and indexes into a digital signature evidence package, applies an asymmetric encryption signature from the evaluation system, and ensures the immutability of the original material by the physical layer.

[0097] Semantic alignment and large-scale model loop closure determination: The large-scale model utilizes a vector model pre-trained in the domain and applies the national cryptographic algorithm. The compliance control items are transformed into semantic feature vectors. The semantics of matching traffic packets are located in the evidence package through retrieval enhancement generation technology, and the corresponding evidence index tags are forcibly embedded in the generated judgment conclusion.

[0098] One-click traceability and reliability verification: Outputs an evaluation report of the integrated evidence package hash. The verifier clicks the evidence index hyperlink next to the judgment conclusion on the visualization interface. The system jumps to the corresponding conclusion with one click based on the mapping relationship and highlights the algorithm negotiation field in the original hexadecimal traffic message. This achieves a complete closed loop from conclusion to physical evidence and completely solves the problem of unverifiable evidence caused by fragmentation.

[0099] Example 2:

[0100] Scenario Background: This embodiment is applied to the automated painting workshop of an automobile manufacturing company, involving a large number of communications via industrial protocols. Controller and Human-machine interface. The characteristic of this scenario is that the device claims to support high-strength encryption in the static configuration document, but in actual operation, due to compatibility or human misconfiguration, it is often in a protocol degradation state, resulting in serious dynamic drift and distortion of static evaluation conclusions.

[0101] Execution steps and procedures:

[0102] Semantic transformation and initial judgment of static control items: Integrating the "Industrial Internet Security Standard" to construct a control item library, the large model transforms prohibited plaintext or weakly encrypted protocol entries into semantic vectors, which are then matched with production lines. The system configuration file was initially found to declare that certificate authentication was enabled, generating an initial compliance intermediate conclusion.

[0103] Deterministic Verifier Physical Rule Verification: Input the above intermediate conclusions into the deterministic verifier, and the certificate compliance parsing engine extracts them. The actual loaded certificate chain is parsed, and if it is found that the certificate has expired or the signature algorithm is outdated... If the verifier determines that a conflict with the physical rules is found, the confidence score for that determination will be immediately reset.

[0104] Triggering active detection script interaction verification: The system automatically calls the brute-force operator for the cipher suite to target... Initiate an interactive handshake request, simulating a client attempting to force the use of a disabled device. Alternatively, a weak encryption suite can be used to connect and obtain the actual feedback messages of the device in its operating state. The detection found that the device did not intercept the downgrade request, but instead generated a downgrade response.

[0105] Conflict resolution and final conclusion correction: Based on the evidence of downgraded vulnerabilities reported by proactive detection, the system triggers conflict resolution logic, forcibly corrects the intermediate conclusion of "configuration compliance" before the large model, assigns the violation judgment coefficient of the control item to the highest level, and calculates the actual risk score according to the risk scoring formula.

[0106] Risk diffusion analysis and full-line synchronization: based on this The system uses asset fingerprints to automatically index assets with the same hardware batch and firmware version within the assessment boundaries of the painting workshop. tower Assets are simultaneously marked with potential protocol downgrade risks. This approach, which combines semantic inference with physical detection, addresses the persistent problem of inconsistencies between conclusions in static compliance assessments and actual operational status.

[0107] Example 3:

[0108] Scenario Background: This embodiment is applied to the distributed control system environment of a large petrochemical enterprise. This scenario includes thousands of pressure sensors, flow meters, and other devices distributed across different plant areas. Control unit.

[0109] Execution steps and procedures:

[0110] Global asset fingerprint profile construction and topology mapping: Utilizing asset identification tools, deep fingerprint detection is performed on all online IoT terminals within the petrochemical plant area to extract equipment model and serial number, and active fingerprint detection operators are used to capture... The unique latency characteristics of the protocol stack and application layer services Description information, combined with the current firmware kernel version, for each unit The terminal generates a unique asset fingerprint vector and establishes a static binding relationship between physical devices and logical evaluation units on the global network topology map.

[0111] Benchmark Node Deep Compliance Assessment and Evidence Binding: The system first selects a representative control center node within the factory area as the benchmark assessment object, collects heterogeneous evidence packages, uses a large model combined with retrieval enhancement generation technology to semantically align the compliance of cryptographic algorithms, and uses a deterministic verifier to parse the algorithm features in the binary firmware to generate a compliance assessment conclusion with associated unique evidence index, identifying that the benchmark node has a high compliance risk due to the use of unauthorized custom encryption and decryption components.

[0112] Risk diffusion analysis algorithm trigger and fingerprint comparison: The system extracts the asset fingerprint of the risk benchmark node as the risk master, starts the automatic indexing mechanism within the defined assessment boundary, and uses the fingerprint similarity algorithm to perform a global search on the existing asset database to search for homogeneous assets with the same hardware model, the same firmware version and similar protocol stack characteristics.

[0113] Dynamic mapping and automatic inheritance of risk assessment scores: When the matching degree between the fingerprint of a newly added or existing asset and the risk master reaches a preset high threshold, the system automatically determines that the asset has homogeneous compliance defects. There is no need to perform tedious manual evidence collection and large model reasoning again. Instead, it directly inherits the assessment rule template and risk judgment logic of the benchmark node, and automatically synchronizes the violation judgment coefficient in the risk scoring formula to the highest weight level.

[0114] Global Risk Map Generation and Batch Rectification Suggestions: The system ultimately outputs a global diffusion analysis report integrating the evidence chain, displays the violation details of the baseline node on the visualization interface, and simultaneously highlights the distribution map of all assets affected by the risk throughout the entire plant area. It also provides hyperlinks for batch rectification suggestions based on asset fingerprints, realizing an efficient closed loop of risk discovery at one point and synchronous early warning across the entire network, significantly improving the overall security assessment efficiency in a large-scale heterogeneous IoT environment.

[0115] Example 1: By introducing a unique evidence index generated by a hash algorithm and an evidence time-series chain technology, an end-to-end closed-loop mapping is achieved from the logical conclusions generated by the large model to the underlying physical evidence files. This achieves the technical effect of full auditability of the evaluation process. The advantage is that it can ensure that even with frequent device firmware changes or dynamic environmental shifts, the verifier can trace and verify the original hexadecimal traffic data or configuration file line numbers with one click based on the index tags integrated in the report. This fundamentally eliminates the persistent problems of evidence fragmentation, easy tampering, and lack of traceability in traditional evaluation models.

[0116] Example 2: By constructing a dual verification system consisting of a deterministic verifier and an active probing script, real-time alignment and logical verification of the semantic inference conclusions of the large model and the actual operating status of the device are achieved. This enables automatic identification and elimination of technical conflicts between false configuration declarations and protocol degradation operations. The advantage is that it can use hard physical rules and interactive probing feedback to correct the reasoning illusions that artificial intelligence may produce, ensuring that the evaluation conclusions remain on the compliance of static documents. It can also truly reflect the cryptographic defense level of IoT devices in complex attack and defense environments, greatly improving the objectivity and effectiveness of the evaluation conclusions.

[0117] Example 3: By utilizing asset fingerprint similarity comparison technology and risk diffusion analysis algorithm, after discovering compliance risks at the baseline node, the risk attributes and judgment logic are automatically mapped synchronously to existing assets with the same characteristics across the entire network. This achieves a highly efficient assessment effect of discovering risks at one point and providing synchronous early warnings globally. The advantage is that it significantly reduces the time cost and manpower investment in conducting full manual assessments in large-scale, high-density industrial control networks, provides management decision-makers with a global view of the distribution of security risks, and enables rapid response and batch closed-loop handling of security risks.

[0118] Embodiments of the present invention have been presented and described. It will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to the embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A large model-based Internet of Things system password application security evaluation method, characterized in that, The security assessment method for cryptographic applications in the Internet of Things system includes: Obtain target asset information to construct asset identifiers, and determine the assessment boundaries based on network topology; Collecting heterogeneous evidence within the evaluation boundary, and generating an evidence index by calculating hashes based on the heterogeneous evidence, including: Establish an evidence timeline chain, and identify and record the dynamic changes in the evidence index caused by device firmware upgrades or configuration tampering by performing consistency comparisons on the heterogeneous evidence collected at different timestamps. The evidence index is associated with the heterogeneous evidence, wherein the heterogeneous evidence includes device firmware images, system configuration files, network communication traffic fragments, and security operation and maintenance logs, and the physical storage paths of the evidence index and the heterogeneous evidence have a one-to-one mapping relationship in the evidence library; The cryptographic compliance standards corresponding to the asset identifiers are integrated into a control item library, which includes compliance items. The compliance items are then converted into compliance semantic feature vectors. The heterogeneous evidence is transformed into an evidence semantic feature vector, which is then matched with the compliance semantic feature vector to generate a preliminary evaluation result. The preliminary evaluation results are verified and corrected to obtain the target evaluation results.

2. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 1, characterized in that, The process of converting the compliance entries into compliance semantic feature vectors involves using a domain-pre-trained large model to extract entities from the compliance entries, identifying the cryptographic algorithm type, key length, and authentication protocol involved in the compliance entries, and then weighting the dimensions of the compliance semantic feature vectors according to the importance of the entities. The process of converting the heterogeneous evidence into an evidence semantic feature vector and matching it with the compliant semantic feature vector to generate a preliminary evaluation result involves pre-setting evidence slots in the prompt word template, forcing the preliminary evaluation result generated by the large model to include and associate the corresponding evidence index. If the generated judgment text does not contain a valid evidence index tag, the reasoning result is determined to be invalid and an automatic re-examination mechanism is triggered.

3. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 1, characterized in that, The step of verifying and correcting the preliminary evaluation result to obtain the target evaluation result includes: inputting the preliminary evaluation result into a deterministic verifier for verification. The deterministic verifier includes a certificate compliance parsing engine, a cryptographic algorithm fingerprint recognition module, and a protocol handshake status monitor, which are used to verify the objectivity and authenticity of the preliminary evaluation result from the physical feature level. If the preliminary evaluation result conflicts with the physical rule parsing result of the deterministic verifier, the evaluation confidence level is reset.

4. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 3, characterized in that, The step of verifying and correcting the preliminary evaluation result to obtain the target evaluation result further includes: triggering an active probe script to perform protocol verification. The active probe script includes a cryptographic suite exhaustive operator and a protocol degradation stress test module, which is used to initiate interactive probes on the target device in a controlled environment to obtain real-time feedback on the dynamic operating status. The preliminary evaluation results are corrected based on the detection results of the active detection script to eliminate conflicts, and the final target evaluation result is obtained.

5. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 1, characterized in that, After obtaining the target assessment results, the process further includes: combining weights, violation indicators, and confidence scores, and using the asset identifiers to perform risk diffusion analysis, including: Extract the asset identifier of the risky device, automatically index existing assets with the same hardware model and firmware version within the assessment boundary, and simultaneously mark potential compliance risks.

6. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 4, characterized in that, The target assessment result includes a comprehensive risk score, which is calculated using the following formula: , in, As a comprehensive risk score, To control the number of items, For control items The weight, This is the violation determination coefficient. The confidence score is the result of feedback from the deterministic validator. This is the index number of the control item; The violation determination coefficient The value is dynamically determined by the detection results of the active detection script. When a protocol downgrade vulnerability or the use of an unauthorized algorithm is detected in the device, the violation determination coefficient is adjusted. It is assigned the highest weight level. The confidence score fed back by the deterministic verifier Controlled by the integrity verification status of the heterogeneous evidence, if the hash verification of any original evidence in the digital signature evidence package fails, the confidence score of the corresponding control item will be... It was automatically set to zero.

7. The method for security assessment of cryptographic applications in an Internet of Things system according to claim 1, characterized in that, After obtaining the target assessment results, the process further includes: outputting an assessment report, which displays the assessment conclusions, evidence index, and rectification suggestions. The associated display assessment conclusions, evidence index, and rectification recommendations include: The visualization interface provides hyperlinks based on the evidence index, supporting one-click tracing from the evaluation conclusion to the corresponding original hexadecimal traffic data or configuration file line number in the digital signature evidence package, thus achieving closed-loop verification of the evaluation results.

8. An electronic device, characterized in that, include: processor; And a memory storing program instructions that, when executed by the processor, cause the electronic device to perform the method according to any one of claims 1-7.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium storing computer-readable instructions thereon, which, when executed by one or more processors, implement the method as described in any one of claims 1-7.

Citation Information

Patent Citations

  • Commercial password evaluation system and evaluation method

    CN121261910A

  • Method and system for dynamically generating a hint to recall a password for a user account of a user

    US20180069843A1