45 results about "Risk research" patented technology

The invention discloses a construction method and system for volatile organic compound release behavior characterization, and belongs to the technical field of air analysis and risk assessment. According to the method, feature analysis is carried out based on multi-source information such as a molecular structure, a release carrier and environmental conditions, modeling is carried out on a release process by utilizing a mapping operator, and a three-dimensional representation vector V = [I, S, E] reflecting release intensity, environmental sensitivity and time evolution is constructed. The representation is output in a behavior vector or digital label mode, and unified quantitative benchmarking of different substances in a cross-scene mode is achieved. According to the method, depth characteristics such as molecular polarity and desorption activation energy are introduced, so that the dependence on long-period environmental chamber experiments is reduced, the generalization prediction capability on unknown pollutants is improved, the logic connection problem between a bottom-layer physical mechanism and an upper-layer intelligent management and control system is solved, and the reliability of the system is improved. And standardized and explainable behavior input is provided for prospective risk research and judgment.

The invention discloses an electric power emerging business risk keyword identification and early warning method, system and device based on an NLP technology and a medium, and belongs to the technical field of electric power risk, and the method comprises the steps: obtaining multi-source customer service text data, and extracting risk keywords and corresponding voice spectrum data from the multi-source customer service text data; inputting the risk keyword into a power service knowledge graph, and generating a first risk research and judgment result through knowledge reasoning, the first risk research and judgment result comprising a risk traceability path; performing acoustic feature analysis on the speech spectrum data, extracting acoustic feature parameters, calculating an emotional excitation index, and generating a second risk identification result; performing weighted fusion on the first risk research and judgment result and the second risk identification result to generate a comprehensive risk score; and when the comprehensive risk score reaches a preset threshold value, generating a structured early warning work order, and carrying out data binding on the structured early warning work order, the original customer service work order and the call record. According to the invention, the efficiency of risk management and control is improved.

The invention relates to the technical field of safety operation and maintenance of power equipment, in particular to an active anti-condensation control method, system and equipment for a switch cabinet and a medium, which are applied to an active anti-condensation control system for the switch cabinet and used for preprocessing operation environment data of the switch cabinet acquired by a sensing layer based on a data acquisition unit; cabinet body surface temperature fusion and dew point temperature calculation are carried out based on the risk research and judgment unit, a dynamic condensation risk index and a global risk index are obtained, and a risk level is judged through the obtained dynamic condensation risk index and a preset threshold value; and according to the risk level and the global risk index, based on a cooperative control unit and according to a preset multi-mode control strategy, generating a control instruction to perform active condensation prevention of the switch cabinet. Therefore, the early condensation trend is pre-judged, and faults such as insulation reduction and metal corrosion caused by condensation formation are effectively avoided.

The invention discloses a hydro-junction construction period structure safety assessment method and system based on digital twinning, and relates to the technical field of hydro-junction safety assessment, and the method comprises the steps: carrying out the safety correlation analysis of a hydro-junction construction period structure; constructing a multi-level digital twinborn space; the method comprises the following steps: collecting construction monitoring data of a hydro-junction during a construction period, synchronously mapping the construction monitoring data to a multi-level digital twin space for level simulation, and identifying a first type of safety events; and when the first type of security events exist, carrying out influence range tracing and risk conduction analysis, and generating a structure security assessment conclusion. According to the invention, the technical problem of insufficient evaluation comprehensiveness and accuracy caused by the fact that dominant safety event identification and hidden risk mining are difficult to consider in the construction period structure safety evaluation of the hydro-junction in the prior art is solved, and accurate monitoring and targeted risk research and judgment of the construction period structure safety are realized; and the technical effect of improving the comprehensiveness and accuracy of structure safety evaluation in the construction period is achieved.

The invention discloses a product oil storage tank foundation settlement and deformation monitoring and early warning system and method, and relates to the technical field of storage tank safety monitoring, the product oil storage tank foundation settlement and deformation monitoring and early warning system comprises a settlement sensing layer, a reservoir level application layer and a platform application layer; the settlement sensing layer is provided with multiple types of anti-explosion sensors, and is used for collecting storage tank data in real time and processing storage tank foundation settlement and deformation related data; the reservoir-level application layer receives the data transmitted by the settlement sensing layer, and realizes the prediction of the settlement trend of the storage tank and the quantitative evaluation of the risk level; and the platform application layer realizes monitoring data visual display, early warning release and data sharing by building an enterprise security risk intelligent management and control platform or docking with a third-party platform. The invention further discloses a product oil storage tank foundation settlement and deformation monitoring and early warning method. According to the invention, real-time data acquisition, deep analysis, accurate processing and risk research and judgment can be realized, the compatibility is high, data sharing and collaborative management and control are realized, and the security and intelligent management and control level is high.

The invention relates to the technical field of data processing, and discloses a commercial tenant risk rating method and system based on multi-source data fusion, and the method comprises the steps: obtaining the multi-dimensional data of a commercial tenant, and carrying out the heterogeneous data alignment, and obtaining the standard data of the commercial tenant; performing data credibility evaluation on the standard data to obtain a quality score of the standard data; based on the quality score, performing multi-dimensional feature fusion on the standard data to obtain merchant associated feature data of the merchant; based on a preset risk research and judgment standard, carrying out risk quantitative analysis on the merchant associated feature data to obtain a risk index of the merchant; performing comprehensive evaluation on the risk index to obtain risk evaluation information of the merchant; performing level threshold comparison on the risk assessment information to obtain a risk level of the merchant; according to the invention, the accuracy of merchant risk rating based on multi-source data fusion can be improved.

The application discloses a kind of digital risk research management methods and systems for high-risk scene, it is related to industrial manufacturing technical field, the method includes: in each target equipment in high-risk place, real-time monitoring device is installed, and based on the functional attribute of each target equipment, the corresponding data type label is configured to the real-time monitoring data obtained by the real-time monitoring device corresponding to each target equipment;When the data type label of real-time monitoring data is state class signal, then based on the preset determination rule and the state corresponding to real-time monitoring data, fault determination is carried out;When the data type label of real-time monitoring data is numerical value class signal, then based on the preset multiple different alarm levels and corresponding alarm trigger threshold, fault determination is carried out.The application can realize the collection of the key parameters of high-risk equipment, improve equipment fault, environmental anomaly prediction accuracy, convert sudden failure repair to planned maintenance, realize safety management front.

The present application relates to a kind of methods for realizing network content risk analysis based on key points driving, comprising the following steps: depth semantic analysis is carried out to original text, and risk key points are identified and extracted;According to the preset confidence threshold filtering low reliability key points, calling big model intelligent screening risk key points;Multi-source evidence online retrieval is carried out, and multi-source search results are integrated and risk tendency determination;Multi-dimensional risk research and judgment is carried out.The method for realizing network content risk analysis based on key points driving, device, processor and computer readable storage medium thereof of the present application are adopted, by reconstructing analysis process with risk key points as core, and comprehensively deciding by depth fusion local semantic knowledge and external real-time information, surpassing existing technology in precision, explainability and flexibility etc.Dimensions, provide solid technical foundation for constructing next-generation intelligent, efficient, reliable network content risk prevention and control system, and have important practical value and innovation.

The invention discloses a bid procurement decision-making method based on multi-data analysis processing. Comprising the following steps: S1, acquiring electronic bidding files submitted by a plurality of bidders participating in the same bidding item target packet; s2, automatically extracting and structuring preset key information in the electronic bidding file to generate structured information; s3, acquiring and fusing multi-dimensional associated data for a plurality of bidders; s4, performing quantitative analysis on the multi-dimensional associated data to generate a comprehensive risk assessment result; s5, generating a control instruction based on the comprehensive risk assessment result; according to the invention, a structured information processing module verified by an external authoritative data source, a cross-domain risk fusion model adopting a nonlinear collaborative amplification algorithm, and a self-adaptive human-computer interaction interface directly driven by a risk assessment result are integrated; therefore, a set of complete closed-loop processing flow from high-credibility data input to intelligent risk research and judgment and then to guide-type decision support is established.

The invention discloses an intelligent water conservancy multi-modal perception and decision optimization system, and relates to the technical field of urban water network security, and the system comprises a multi-modal perception and fusion module which is used for generating a comprehensive perception data flow; the intelligent diagnosis and early warning module is used for carrying out abnormal feature extraction and event positioning on the comprehensive perception data flow and generating a preliminary diagnosis result; carrying out risk research and judgment to obtain an early warning risk result; the physical constraint risk deduction module is used for deducing by utilizing the physical constraint solving model to generate a risk evolution field; the multi-objective cooperative control module is used for constructing a multi-objective optimization function, reinforcing a learning framework by utilizing a meta-strategy gradient, and generating a cooperative scheduling strategy; the digital twinning and strategy rehearsal module is used for constructing a water network twinning body; and performing verification and screening by using a bidirectional state synchronization mechanism, and outputting an optimal decision set. According to the invention, the core problems of difficult positioning of pipe network aging, low multi-source monitoring precision, rainfall flood regulation lag and the like in urban water network management can be solved.

The invention discloses a garlic whole-process monitoring decision-making platform based on AI and big data, relates to the technical field of artificial intelligence, and solves the problems of data aperture heterogeneity and sample deviation in garlic price prediction in the prior art and the problem that a result deviates from the actual market due to the lack of inventory-behavior feedback modeling in the prediction process. According to the scheme, a multi-level linkage architecture comprising a data protocol layer, a causal deduction layer, an interactive analysis layer, a feedback simulation layer and a parameter calibration layer is constructed, and a consistent data stream is formed by performing time sequence protocol and causal correction on multi-source price data; performing interactive analysis on the price chain and the inventory sequence, introducing a multi-subject response model to simulate market feedback, and finally forming a closed-loop prediction sequence through a dynamic calibration mechanism; according to the method, the prediction stability of garlic whole-process monitoring decision and the reliability of risk research and judgment are remarkably improved.

The application discloses a power grid real-time operation risk automatic identification and grading method based on deep learning, and belongs to the technical field of power system automation, which comprises the following steps: obtaining and fusing power grid real-time and historical data to generate a data set; performing double-flow feature extraction based on causal inference to generate a deep feature vector; using a power grid digital twin model to perform forward deduction to obtain a risk deduction consequence set; developing dynamic risk research and judgment to generate a risk grade and a disposal plan; generating an interpretable early warning report and updating network weights through reinforcement learning according to dispatcher feedback data. The application adopts the double-flow feature extraction based on causal inference and the digital twin deduction technology, can effectively handle the nonlinear relationship between risks and working conditions, realizes automatic identification, quantitative calculation and dynamic grading of risks, and improves the timeliness and accuracy of dispatchers in dealing with real-time operation risks and cascading failures.

The application relates to a fund whole-process digital management and control and on-chain evidence storage method, which comprises the following steps: receiving an account adjustment operation request and analyzing the fund secret level; dynamically calling a corresponding set of security protection parameters based on the secret level; constructing a group behavior baseline model based on a privacy protection mechanism; collecting operation behavior characteristics in real time, and calculating the double-baseline deviation of the operation behavior characteristics from the individual and group baselines; fusing and judging the deviation according to a preset threshold to determine a risk level; executing a graded disposal strategy corresponding to the risk level, and synchronizing whole-process key data to a block chain node for evidence storage. The application realizes the online of the whole process of secret fund account adjustment, the intelligentization of risk research and judgment, and the trustworthiness of operation traces, and achieves the comprehensive effects of improving disposal efficiency, strengthening active risk control, ensuring the consistency of accounts and facts, and meeting the requirements of judicial audit level traceability.

The invention discloses a client interaction data analysis method and device, equipment and a medium. By automatically extracting semantic features and emotional features in customer interaction data, accurately analyzing customer appeals and emotional states, intelligently associating a target business entity with historical cases in combination with a preset domain business knowledge base, and generating associated features with business interpretability, risk analysis is deeply rooted to a professional knowledge system. And performing comprehensive intelligent research and judgment by fusing the multi-dimensional features, and automatically outputting a structured decision result containing a risk type, a risk level and a processing priority. Traditional low-efficiency and lagged risk research and judgment depending on manpower are converted into a real-time and accurate intelligent decision-making process. According to the invention, intelligent discovery and early warning of service risks are realized, and service personnel are assisted to make efficient responses according to clear priorities, so that the service efficiency and the service quality of online customer services are comprehensively improved, and the customer satisfaction is effectively improved.

The invention relates to the technical field of greenbelt management and maintenance, in particular to a greenbelt intelligent management and maintenance system based on cloud computing, which comprises a cloud multi-source integration module, a time sequence disturbance analysis module, a disturbance chain association module, a disease trend identification module and a management and maintenance regulation and control linkage module. And judging the online state of the sensing equipment and the plot numbering accuracy. According to the method, a cloud-oriented data integration system is constructed, multi-source environment parameters and spatial attributes are collected into a dynamic data structure, global association and time sequence connection of information among land parcels are achieved, data are automatically sorted through a cloud platform to form a parameter evolution link, multi-level feature criteria such as coverage, time sequence and trend are established for monitoring objects, and the monitoring efficiency is improved. Abnormal evolution performance of a disease state is directly fed back to a management and maintenance regulation and control process through linkage judgment, and system output regulation and control configuration is generated according to a real-time state, so that the consistency of risk research and judgment and the pertinence of a regulation and control decision are improved.

The application discloses a kind of frozen disaster dynamic comprehensive risk research and early warning method, it is related to highway traffic disaster prevention and control technical field, including obtaining the real-time monitoring data of frozen disaster in 12 hours in monitoring area;Real-time monitoring data is matched with early warning grade division standard, and the risk grade of frozen disaster in monitoring area is researched and judged;Based on the risk grade research result, the highway network characteristics of monitoring area, the reserve base number of deicing and snow-removing equipment and materials are combined to generate the deicing and snow-removing material distribution basic scheme of frozen disaster;Combined with the preset path optimization algorithm, the material allocation and optimal transportation route of provincial and county-level reserve center are determined;The optimized material distribution scheme is issued to the corresponding provincial and county-level reserve center, and the reserve center executes material allocation and transportation according to the scheme.The application realizes the dynamic and accurate research of frozen disaster risk, and improves the scientificity and timeliness of material distribution.

The application relates to a kind of job safety risk identification methods based on multi-modal knowledge graph, comprising the following steps: obtaining multiple modal power field operation data and pre-processing;Schema of multi-modal knowledge graph is constructed;The knowledge extraction of different modal power field operation data is carried out respectively, and the label of power field operation data is obtained;The label of different modal power field operation data is fused;Knowledge storage is carried out through the label after fusion and Schema, and the multi-modal power field operation risk cognitive graph is obtained;The multi-modal power field operation risk cognitive graph is inferred;Through the multi-modal power field operation risk cognitive graph, auxiliary decision is carried out.The application solves the problem that multi-source data of power field operation is difficult to be efficiently utilized by constructing multi-modal power field operation risk cognitive graph, improves the comprehensiveness, timeliness and accuracy of operation risk research and judgment, and provides support for on-site operation safety control decision.

The invention discloses a nuclear power plant hydrogen risk analysis method and system, electronic equipment and a storage medium, and relates to the technical field of nuclear power plant safety analysis. Comprising the steps of establishing a key parameter probability distribution model of a nuclear power plant hydrogen originating accident; generating an initial condition combination sample set of the key parameters based on a probability sampling method, and constructing a thermal hydraulic calculation working condition matrix; executing high-fidelity simulation of a full-sample working condition, and generating a containment multi-compartment spatio-temporal evolution data set; training a time sequence neural network model by using the data set to realize mapping of an accident process and multi-physical field parameters; and inputting a real-time accident type and a monitoring initial value, driving the model to output hydrogen concentration prediction, and synchronously completing dynamic evaluation of the AICC risk and the DDT risk. According to the invention, through efficient calculation and data analysis means, rapid probability evaluation of diffusion, aggregation and explosion consequences of hydrogen in the containment is realized, and the risk research and judgment speed and decision reliability under severe accidents of the nuclear power plant are significantly improved.

The invention discloses a coal sample bag real-time detection and violation replacement early warning method. The method comprises the steps of sample initialization and monitoring starting, sample circulation monitoring and violation early warning and authorized unsealing and monitoring termination. The coal sample bag is provided with a flexible electronic tag and an anti-counterfeiting texture; the radio frequency radar and the camera are deployed at key nodes of a circulation path; the reasoning server is in signal connection with the radio frequency radar and the camera and is used for operating a reasoning algorithm; the personnel identity verification device and the anti-counterfeiting texture recognition device are arranged at the unsealing use node; according to the invention, through three core modules of anti-counterfeiting mark enhancement, multi-dimensional state perception and intelligent risk research and judgment, the tamper-proofing and replacement-preventing capabilities in the sample circulation process are significantly improved.

The invention relates to an information monitoring and tracing method and system based on voiceprint and semantic analysis, which effectively solves the problems of privacy invasion and computing power waste caused by a full-time full-amount analysis mode by introducing a condition triggering mechanism that non-content characteristics such as voiceprint and emotion are analyzed in advance and deep semantic understanding is triggered after the risk confidence reaches the standard. Only when the comprehensive risk assessment shows that the real risk exists, directional analysis is carried out on the voice content, so that the balance between privacy protection and effective monitoring is realized. Meanwhile, risk assessment is carried out by fusing multi-modal features such as voiceprints, emotions and environmental sounds, and an anonymous voiceprint behavior database is established for historical tracing, so that the accuracy and depth of risk identification are remarkably improved, the system can give an alarm in real time, behavior tracks and mode analysis of involved subjects can be provided, and the risk identification efficiency is improved. According to the invention, the spanning from single event alarm to continuous risk research, judgment and traceability is realized, and more comprehensive and more intelligent decision support is provided for managers.

The invention provides a full-period visual analysis method and device for risk events, and relates to the technical field of Internet. According to the method, risk perception is carried out on multi-source heterogeneous data of a plurality of service systems, possible risk events such as social public security events, social stability maintenance events and mass events are identified, then risk research and judgment are carried out on the risk events based on a generated full-period visual interface, and the risk events are subjected to risk analysis and analysis. According to the method, research and judgment results such as event causes, development stages and risk levels of the risk events are determined, and finally, a disposal scheme is generated according to the research and judgment results, so that dynamic visual analysis of the risk events is realized, and the problems of data dispersion and incomplete analysis data and result display in a risk event analysis process are solved; the dynamic visualization ability of total elements, total process and total association of the risk events is realized, and analysts can conveniently and accurately recognize the risk events.

The invention relates to the field of intelligent data processing, and particularly discloses a cross-border trade data processing method and system based on an intelligent port, and the method comprises the steps: processing original data in a local trusted execution environment of each port node, and generating a privacy protection feature vector, generating a verifiable voucher for verifying the calculation compliance by using zero-knowledge proof; verifying the voucher through a distributed consensus network, and fusing each node feature vector by adopting secure multi-party calculation to generate a cross-domain joint feature; the features are input into a decentralized agent network for asynchronous negotiation, and a collaborative risk research and judgment report is output; each node finally optimizes a local model under differential privacy protection according to report feedback; the system comprises four corresponding modules. According to the method, credible collaboration and deep risk insight on the premise that the original data is not out of the domain are realized, and continuous autonomous evolution of the system is guaranteed.

The application discloses a product oil storage tank foundation settlement and deformation monitoring and early warning system and method, relates to the technical field of storage tank safety monitoring, and comprises a settlement sensing layer, a warehouse level application layer and a platform application layer; the settlement sensing layer is configured with multiple types of explosion-proof sensors, can collect storage tank data in real time, and can process settlement and deformation related data of a storage tank foundation; the warehouse level application layer receives data transmitted by the settlement sensing layer, realizes prediction of a storage tank settlement trend and quantitative evaluation of a risk level; the platform application layer can realize visual display of monitoring data, early warning release and data sharing by building an enterprise safety risk intelligent control platform or connecting a third party platform. The application further discloses a product oil storage tank foundation settlement and deformation monitoring and early warning method. The application can realize real-time data collection, deep analysis, accurate processing, risk research and judgment, has high compatibility, realizes data sharing and collaborative control, and has high safety and intelligent control level.

The invention relates to the technical field of information security, and discloses a simulated attack verification method based on a security agent, which comprises the steps of research and judgment result access and context enrichment, security agent PoC harmless verification, dynamic strategy arrangement and man-machine collaborative judgment, automatic execution and strategy issuing, and closed-loop feedback and model self-evolution. According to the method, intelligent screening of the PoC scripts is realized through vector matching, and non-destructive risk verification is completed in the isolation sandbox; a strategy weight model is constructed based on asset attributes and business contexts, strategy dynamic arrangement and conflict resolution are realized, and a man-machine co-processing boundary is clarified; all-link data feedback is opened, and risk research and judgment model self-optimization is achieved. According to the method, real and theoretical risks are accurately distinguished, the error disposal probability is reduced, the continuity of government affair cloud core services is guaranteed, continuous self-evolution of a security protection system is realized, and the strict requirements of government affair cloud security protection are met.

The application discloses a kind of data security risk automatic evaluation method and system based on artificial intelligence, it is related to AI model calling technical field, load evaluation index and preset task parameter, the initialization matching of to-be-tested environment is carried out, and evaluation benchmark model is obtained;Standardized data asset list is obtained by processing multi-source original data using NLP and machine learning algorithm, and compliance comparison analysis is carried out with the index in evaluation benchmark model, and preliminary risk list is obtained;The preliminary risk list is automatically desensitized, and the implicit risk mining and causal inference are carried out in the private large model, and the deep risk research result is obtained, then weighted calculation is carried out, and the quantized risk grade numerical value is obtained;According to the quantized risk grade numerical value, reasoning processing is carried out, and risk disposal suggestion priority and standardization evaluation report are obtained;Through evaluation report rectification retest, dynamic security risk situation data is obtained, realizes the timely early warning of data, solves the short board of traditional static evaluation.

The application discloses a full-dimension digital enterprise new-quality scientific and technological innovation risk intelligent early warning and closed-loop prevention and control system and method, and belongs to the technical field of enterprise scientific and technological innovation big data risk control, new-quality productivity safety management and intelligent risk research and judgment. In view of the industry pain points that traditional enterprise scientific and technological innovation risks rely on manual inspection, post-repair, non-quantitative models, no pre-warning, uncontrollable patent risks and digital risks, and low fault tolerance of scientific and technological innovation investment, the application constructs a six-digital monitoring system of scientific and technological innovation risks including policy, technology, patent, fund, talent and market, realizes real-time data collection, AI risk rating, threshold intelligent early warning, hierarchical disposal and iteration of the whole link closed-loop management and control, and realizes the pre-judgment, in-process control and post-cure of enterprise new-quality scientific and technological innovation risks, and builds a complete enterprise scientific and technological innovation safety system, and guarantees the stable, continuous and high-quality growth of new-quality productivity.

The invention discloses a weak grating array sensing landslide state observation system, and belongs to the field of landslide state observation, and the system comprises a risk division unit which is used for carrying out the risk grade division of a landslide disaster area; the equipment deployment unit is used for deploying a multi-field weak grating sensing array monitoring optical cable according to the risk level of each monitoring area; the data acquisition unit is used for adaptively adjusting a sampling time interval and acquiring monitoring information of a multi-field weak grating sensing array monitoring optical cable; the threshold criterion unit is used for correcting the threshold criterion of the multi-field monitoring parameters; the risk studying and judging unit is used for studying and judging the risk of the landslide disaster to obtain a risk studying and judging index of each monitoring point; and the early warning and forecasting unit is used for determining an early warning and forecasting level according to the risk research and judgment index of each monitoring point and sending out early warning and forecasting information based on measures corresponding to the early warning and forecasting level. According to the invention, the problem that the existing method cannot meet the requirements of timeliness and accuracy monitoring of the landslide is solved.

The invention relates to the technical field of information security, and provides a taint data tracking method and device based on byte code instrumentation, and the method comprises the steps: respectively building corresponding shadow storage regions for a stack frame and a stack which run in a virtual machine; inserting a data operation monitoring code into the key byte code instruction of the virtual machine; in the program execution process, circulation of data between stack frames and heaps is tracked based on data operation monitoring codes, and stain marks are synchronously updated in the shadow storage area. According to the method and the device provided by the invention, the corresponding shadow storage areas are respectively established for the stack frame and the heap, so that fine-grained management of the stack frame and the heap object is realized, and accurate tracking of the whole process of data from input to output is ensured; a data operation monitoring code is inserted into the key byte code instruction, and the data flow direction in the whole program execution process can be captured; the flow direction of the data input by the user is accurately tracked, the potential safety risk is identified, and the false report and the missing report of risk research and judgment are reduced.

The application relates to the technical field of sensitive information auditing in the government-enterprise field, and specifically provides a sensitive auditing method for the government-enterprise field based on a multi-modal knowledge graph and a large model. Based on a standardized multi-modal data set, entities, relationships and attributes are extracted, and information of different modes is associated and fused through cross-modal association to construct a knowledge graph of fused multi-modal information. Based on the knowledge graph of fused multi-modal information, graph retrieval technology is adopted to search for multi-modal context information associated with to-be-audited input information in the knowledge graph. The to-be-audited input information and the searched multi-modal context information are input into a multi-modal large model, sensitive information recognition and risk research are performed by the multi-modal large model, it is judged whether the to-be-audited input information involves sensitive information and the sensitive level thereof is reasoned, and finally, an auditing result containing the sensitive level, the confidence and the research basis is output. The accuracy and reliability of sensitive information auditing in the government-enterprise field are improved.