144 results about "Content security" patented technology

The application discloses a feature editing method for large model content security, which compares and analyzes the sparse coding features of a chat assistant constructed based on a large language model under positive user input and negative user input, extracts the internal response differences of the model to different semantic directions, and the mechanism can automatically and accurately identify the key feature dimensions highly related to the semantic direction of the target attribute. The model activation is mapped to a sparse feature space by using a sparse autoencoder, and each dimension of the feature has independent and interpretable semantic meaning. By injecting a feature guide vector in the space, the interference of the control process on the text grammar, fluency and information density is significantly reduced. The sparse representation mechanism is introduced to structure the intermediate activation features in the reasoning process of the large language model and to intervene in a targeted manner, so that the reply of the chat assistant to the user input conforms to the preset safety specification, and the safety and controllability of the chat assistant in the interaction with the user are improved.

The invention relates to the technical field of computers, and discloses a model content security control method and system, electronic equipment and a storage medium, and the method comprises the steps: combining a received user input with a historical dialogue record of a current session to form a dialogue context; performing risk identification based on the dialogue context, generating at least one risk category and a corresponding initial risk value, and generating a risk integral corresponding to the current input according to a preset integral strategy; accumulating the risk points to accumulated risk points of the current session; comparing the accumulated risk integral with a dynamic risk threshold value obtained by calculation according to a reference threshold value, a logarithmic function attenuation item of the dialogue round and a user historical behavior adjustment item; and when the accumulated risk integral reaches or exceeds a dynamic risk threshold value, triggering a preset risk management and control measure. According to the method, the risk content can be effectively identified and controlled, attacks bypassing a security mechanism through slow induction can be defended, and the security boundary is not easy to detect.

The invention provides a large model dynamic protection method and system based on a zero-trust architecture. The method comprises the steps that a security proxy gateway receives an access request; authenticating an initiating main body of the access request, collecting context information and transmitting the context information to a strategy decision point; the strategy decision point calculates a trust score in real time based on a dynamic trust evaluation model and performs real-time evaluation in combination with an access control strategy to generate a dynamic authorization judgment result; if the access is allowed, forwarding the access request to a large language model server, and performing input security filtering; the large language model server generates response content and performs output security filtering; and returning the final response subjected to the output security filtering to the initiating main body through the security proxy gateway. According to the method, a multi-layer protection framework is constructed, a dynamic trust evaluation model is introduced, and a content filtering layer is deployed, so that continuous permission verification, risk adaptive control and full-link content security protection are realized, and the service security of a large model is effectively guaranteed.

The invention relates to the technical field of natural language processing, in particular to an AIGC content safety monitoring system and method based on dynamic reasoning and context awareness, and the system comprises a semantic graph construction module which is used for extracting entity nouns and predicate verbs in a text according to a received AIGC interaction text flow, generating a semantic concept node set, and sending the semantic concept node set to a database; and performing directed connection and hierarchical nesting on the concepts in the semantic concept node set according to a logic direction according to a subject-predicate-object dependency relationship rule. According to the method, the curvature value of the semantic track is calculated, and the similarity between the direction vector and the center of the sensitive semantic cluster is combined for double verification, so that sudden turning of an intention in a dialogue process or progressive induction to a sensitive field can be perceived, and abnormal mutation can be recognized through a curvature pulse form; therefore, hostile attack behaviors are accurately captured in real time in dynamic interaction, and the defense capability for context dependent attacks and implicit induction behaviors is improved.

The invention provides a video evidence obtaining method based on Transform and quantum features, and belongs to the technical field of crossing of multimedia content security and computer vision, and the method comprises the steps: S1, collecting multi-modal original data; s2, preprocessing the multi-modal data; s3, single-mode authenticity preliminary detection is carried out; s4, multi-modal feature fusion of quantum optimization is carried out; and S5, multi-modal authenticity comprehensive judgment is carried out. Through multi-mode cooperation and quantum technology innovation, the problems of insufficient robustness, inaccurate positioning and the like of a traditional evidence obtaining technology are effectively solved, and an efficient, accurate and feasible technical scheme is provided for authenticity verification of video content.

The invention provides a large model content generation auditing method and system based on man-machine collaboration, and relates to the technical field of large models, and the method comprises the steps: deconstructing the generation process data of to-be-audited content, constructing a content risk propagation map, mapping a multi-level auditing strategy library, and recognizing an uncertainty region needing manual intervention. And obtaining manual auditing judgment and correction operation, forming auditing knowledge precipitation and perfecting an auditing strategy library. According to the method, the content risk source can be accurately positioned, the auditing efficiency is improved, a closed-loop optimized man-machine collaborative auditing mechanism is formed, the misjudgment rate is effectively reduced, and the large model content security is improved.

The invention belongs to the technical field of computer information security, particularly discloses a content security protection method and device based on semantic consistency, and aims to solve the problem that high-level cue word injection attacks are difficult to effectively recognize in the prior art. Carrying out real-time interception on dominant illegal texts through a content filtering module; quantizing generation rationality differences of prompt words between attack and normal language models by using a word vector confusion degree calculation module; evaluating the logic coherence of the sentence structure of the prompt word through a statement semantic consistency judgment module; integrating the multi-dimensional feature data and carrying out risk classification by adopting a machine learning algorithm; and routing the suspected attack request to a value fine tuning model for processing according to a judgment result. According to the technical scheme, high-precision recognition and response to complex cue word injection attacks are achieved, and the content safety protection capacity and compliance guarantee level of a large model in an interaction scene are remarkably improved.

The invention relates to the technical field of network content security, in particular to a network media video sensitive information intelligent identification and early warning system, which comprises a data acquisition module for acquiring network media video data; the video information extraction module performs multi-dimensional analysis on the acquired data to obtain depth features; the AIGC video detection module identifies whether the content is the content generated by the AIGC technology or not according to the depth feature, and if yes, an AI generation mark is marked; the causal inference and evaluation module is used for carrying out causal correlation diagnosis on intention and context, carrying out deep analysis through an anti-fact inference technology, judging whether the intention is sensitive information or not, evaluating dangerousness and obtaining a dangerousness level; and the early warning module takes the minimum intervention cost and the maximum risk control as targets, matches and dynamically generates an optimal intervention scheme from a strategy library in combination with the risk level, and intervenes the sensitive video. Therefore, the problems of incomplete multi-modal information analysis, lack of an AI generation content identification mechanism and the like are solved.

The invention discloses a multi-modal content real-time detection and dynamic replacement method and system. The method comprises the following steps: buffering an input audio / video stream for a fixed time length and synchronously decomposing the input audio / video stream into a video image, an audio text and a picture text; performing sensitive content detection on the three types of data in parallel, and judging overall violation according to at least one violation; and according to the violation type, selecting a global replacement mode or a local mask mode to carry out real-time replacement processing. The system comprises corresponding function modules. According to the invention, multi-modal decomposition and parallel detection are combined with cross-modal cooperative determination, so that the problem of high missed determination rate of single detection is solved; the detection precision and the real-time performance are balanced through fixed buffering and dynamic replacement strategies; particularly, a text detection and recognition fusion model is adopted to improve the dynamic subtitle recognition rate; wide-temperature hardware and fan-free design are combined, stable operation of the system in severe environments such as outdoors is guaranteed, and accuracy, real-time performance and reliability of content security control are remarkably improved.

This invention discloses a batch multimodal data alignment method and system based on the CLIP model, belonging to the field of information processing technology. The method includes: S1: receiving batch multimodal data; S2: data preprocessing; S3: batch feature extraction based on the CLIP model to achieve feature alignment; S4: batch classification based on a Prompt template; S5: result generation and output, and visualization processing. This invention is applicable to user behavior analysis, abnormal traffic detection, and network content security governance in a big data communication environment. It fully utilizes CLIP's powerful cross-modal semantic alignment and zero-sample transfer capabilities, combined with batch optimization strategies, dynamic task scheduling, and a batch classification method based on a Prompt template, thereby achieving efficient, flexible, and scalable multimodal data processing, accurately identifying malicious text, malicious images, and cross-modal risk links, and enabling real-time monitoring and intelligent prevention of malicious network behavior.

This invention discloses a method and system for preventing false information injection using a large language model. The method involves: dynamically evaluating the credibility of user input to obtain an input credibility score; calculating a user reputation score based on historical user behavior data and mapping it to a generation permission level; retrieving authoritative knowledge fragments from a closed-loop trusted knowledge graph for the input and constructing generation constraint instructions based on the permission level; calling the large language model to generate response content according to the permission level and constraint instructions; performing consistency verification on the response content and then outputting it; finally, generating a lineage log containing end-to-end interaction data and storing it on the blockchain. This invention, by integrating input credibility assessment, user reputation coupling, knowledge tracing constraints, and end-to-end auditing, achieves pre-emptive identification, process blocking, and post-event traceability of false information, effectively solving the problem of the lack of systematic defense against false information injection attacks in existing technologies, and significantly improving the content security and compliance of generative artificial intelligence applications in key areas.

This invention discloses a network security browsing system and method. First, a user request forwarding module sends user requests to an RBI server, with all subsequent interactions completed on the RBI server. This achieves physical isolation between the user's local device and internet content, preventing direct attacks from malicious websites and ensuring user access security. Second, a request analysis and processing module performs in-depth analysis of the content and destination of user requests, proactively identifying and intercepting potential threats such as phishing websites and malicious code injection, compensating for the shortcomings of traditional technologies in ensuring website content security. Third, when the target website is secure, the content acquisition and presentation module allows the RBI server to request and process content in a secure environment, making it difficult for malicious code to cause damage and preventing attacks on the system. All modules work together to create a safe and reliable network browsing environment for users.

A generative artificial intelligence-driven method for detecting inappropriate content relates to the field of inappropriate content detection technology, solving the problems of inaccurate detection and reliance on manual labor. The method includes: automatically collecting inappropriate content to obtain a first inappropriate content dataset, and performing detection through a content security detection platform to obtain a first detection result; extracting sample labels from the data in the first inappropriate content dataset to obtain undetected inappropriate features and their combination patterns; manually analyzing the data to obtain highly concealed inappropriate features and novel combination methods, generating prompt words for the AIGC model; constructing a multivariate, highly concealed inappropriate content dataset based on the inappropriate content generated by the AIGC model and not detected by the content security detection platform; and training a content security detection model using the labels and multimodal features of the above two inappropriate content datasets. This invention has high detection efficiency and accuracy, can detect highly concealed and diverse inappropriate content, and reduces reliance on manual labor.

The invention provides a chaotic digital image encryption system performance analysis method and device based on an intelligent optimization algorithm, and belongs to the field of information security and password engineering. The problems that an existing chaotic image encryption method depends on manual parameter adjustment, parameter optimality is difficult to guarantee, confusion and diffusion intensity is insufficient, and statistical safety is unstable are solved. The method comprises the following steps: encrypting a digital image through a chaotic digital image encryption system; an intelligent optimization algorithm is introduced, a target function related to performance indexes is designed, optimal control parameters are automatically searched in a given Lorenz chaotic system parameter domain, and performance analysis is carried out on the chaotic digital image encryption system through a scientific method. The method and the device are suitable for digital image encryption, content security protection, communication privacy protection and multimedia security application scenes.

A structured specification of a content safety policy is received. The structured specification of the content safety policy conforms to a syntax. Content to be evaluated is received. The received content is evaluated using the structured specification of the content safety policy to determine whether the received content violates the content safety policy.

The invention relates to a material processing method and system based on sensitive content detection and a medium, and belongs to the technical field of multimedia content security. The material processing method comprises the following steps: receiving an input original material, and analyzing the input original material into an image stream, a video stream and a text stream; executing HSV color space conversion, and extracting texture features; performing semantic analysis on the text stream to generate a word segmentation sequence and a named entity recognition result; random frequency noise is injected based on the saturation channel, and local binary pattern features are extracted from the saturation channel after noise injection; generating a semantic vector according to the word segmentation sequence and a corresponding named entity recognition result, and calculating a risk entropy value; and establishing a space-time mapping relationship between the image stream / video stream and the text stream, constructing a cross-modal incidence matrix, executing a grading decision according to an output result, executing a material processing operation, and feeding back a processing result to the cross-modal incidence matrix for weight updating. The sensitive content identification accuracy and timeliness can be improved.

The present application relates to the technical field of data processing, in particular to an intelligent customer service base large model distillation method, system and storage medium, aiming at efficiently and safely migrating the core capabilities of a large intelligent customer service base teacher model to a lightweight student model. In the model distillation process, a composite loss function containing retrieval-aware fidelity loss and safe ethics alignment loss is used to optimize the student model. The retrieval-aware fidelity loss ensures that the student model can effectively learn and use the key retrieval information relied on by the teacher model, while the safe ethics alignment loss constrains the output of the student model to comply with the specifications of specific information security. The present application solves the problem that the traditional distillation method is difficult to fully retain the utilization ability of the intelligent customer service model for external knowledge and ensure the safety of the output content, so that the lightweight model can provide accurate, reliable and safe intelligent services in a resource-limited environment.

The invention discloses a private library local area network medical knowledge security sharing system, and belongs to the technical field of medical health information. The system takes a multi-modal medical knowledge analysis and reconstruction engine as a core, accesses multi-source heterogeneous medical data through interfaces such as DICOM and HL7, and performs semantic analysis and standardized packaging based on a medical ontology knowledge graph to generate a unified medical knowledge object. The system constructs a controlled peer-to-peer private library network, and realizes two-way encryption transmission and fine-grained dynamic access control between nodes based on a digital certificate; classification and desensitization are carried out on the recognizable information of the patient through the content security gateway, and invisible watermarks are embedded to realize knowledge circulation full-life-cycle traceability. Unified semantic analysis and clinical decision rule conversion of multi-source medical knowledge are realized, and the medical knowledge integration efficiency and sharing compliance in a local area network are remarkably improved on the premise of ensuring privacy security.

The application discloses a rule and deep learning model-based illegal content detection method and device and medium, relates to the technical field of content security, and comprises the following steps: obtaining to-be-detected data; analyzing the data content based on a preset rule engine to obtain a rule determination result and a rule confidence; determining a target deep learning model corresponding to the rule confidence and system load, inputting the data content into the target deep learning model, and obtaining a model determination result output by the target deep learning model; performing weighted summation processing on the rule determination result, the model determination result and user information to obtain a final determination result, and generating indicating content indicating that the to-be-detected data is illegal in the case that the final determination result represents illegality. The application is used to solve the problems of low accuracy and slow detection efficiency in the prior art when a single mode is used for illegal content detection, and realizes fast and accurate illegal content detection.

This application discloses a data review method, device, and storage medium based on AI vision, relating to the fields of artificial intelligence and network information security technology. The method includes: responding to a data review instruction, inputting the text and image features of the video to be played into an AI recognition model to determine illegal content; classifying the illegal content of the video to be played and the video to be played according to the regional permissions of the target screen's location, and processing video content outside the permissions and illegal content within the permissions according to the permission content processing strategy to obtain a video processing result; adjusting the video to be played based on the video processing result to obtain a target video, so that the target screen can play the target video. This application solves the problem of low efficiency in illegal content control by using lightweight AI on the edge to identify illegal content, combined with regional permission classification and hierarchical processing to generate compliant videos, improving the review response speed and accuracy, and achieving full-cycle traceable content security protection.

The invention relates to the technical field of artificial intelligence security, discloses an AI content security interception method and system based on an external mounting type mandatory access, and aims to solve the problems that the security protection of an existing AI model is fragmented, non-mandatory and easy to bypass. The core of the technical scheme is to deploy a plug-in type mandatory access independent of a model body at an AI model service entry, all input data is forced to flow through the access to complete feature extraction and security decision, and key innovation points comprise a plug-in type architecture, a mandatory access and a permission isolation mechanism. According to the method, unified security interception of the multi-source heterogeneous AI model is realized, a perceptible but non-intervened defense system is formed, an infrastructure-level solution is provided for AI security governance, and the method can be used as the lowest industrial security standard in the field of artificial intelligence.

The application provides an enterprise document synchronization method and system, an electronic device and a storage medium. The enterprise document synchronization method is used for a server and includes the following steps: receiving an operation request for a target document from a first client, wherein the operation request includes adding, editing or deleting; performing permission verification on the operation request according to the first client; performing content compliance verification on the operation request based on the permission verification; performing version consistency verification on the operation request based on the content compliance verification; and sending consent information to the first client and distributing the operation request to a plurality of second clients based on the version consistency verification. Through real-time synchronization, permission verification and content compliance verification, the application realizes safe and efficient collaboration of documents in a multi-terminal and multi-user environment, and overcomes the problem of lack of fine permission control and content security review mechanism in related technologies.

This invention proposes a content security review method, system, device, and medium based on a multimodal large model. The method includes: acquiring data to be reviewed, wherein the data to be reviewed is an image and / or text; performing image encoding and / or text encoding on the data to be reviewed using an image encoding module and / or a text encoding module to obtain a first target feature; acquiring user-defined review standards, performing text encoding on the review standards using the text encoding module to obtain a second target feature; performing review inference on the first target feature and the second target feature based on a pre-built large model module, and then outputting the prediction result through an output layer module. This invention allows users to customize review standards, solving the challenges of diverse review standards and dynamic content, resulting in higher accuracy, speed, and flexibility in the review inference results.

The invention discloses an outdoor large screen content management and control method and system. Comprising the following steps: firstly, identifying a content publishing mode of an outdoor large screen, and distinguishing a far-end content publishing mode and a local terminal content publishing mode; aiming at a far-end mode, executing a network management and control process, modifying an IP address of a content publishing server through a content display box, performing matching verification in combination with an equipment identifier-server IP mapping relation table pre-stored in a centralized firewall, presetting a content updating window period, only allowing networking updating in the window period, and realizing unnecessary non-networking; for a local mode, executing a remote power supply management and control process, establishing connection with the management and control platform through a remote power supply control module to monitor an abnormal signal, and remotely cutting off a display screen power supply when an abnormity is confirmed; and finally, recording a management and control log containing information such as triggering reasons and execution time. According to the invention, accurate and differentiated management and control of different publishing modes are realized, and the content security guarantee level of the outdoor large screen is improved.

The invention provides a multimedia transmission method based on priority and security auditing. The multimedia transmission method comprises the steps that a second multimedia terminal initiates a multimedia transmission session application to a first multimedia terminal; the first multimedia terminal senses the network topology and selects a content security auditing service according to the session application; calculating a transmission path transmitted from the first multimedia terminal to the second multimedia terminal; the first multimedia terminal generates a flow table rule according to the transmission path and the network topology, and sends the flow table rule to the communication equipment; the first multimedia terminal performs priority distribution on the multimedia information to be transmitted by adopting a priority weight distribution strategy; transmitting the multimedia information with the allocated priority to a second multimedia terminal through the communication equipment; the second multimedia terminal carries out fusion display on the multimedia data sent by the first multimedia terminal; according to the invention, the multimedia data are transmitted according to different priorities, and the data with high real-time requirements are transmitted preferentially, so that the lagging or delay caused by network congestion can be avoided.

The invention provides a network content security identification method and device, a storage medium and equipment, and the method comprises the steps: obtaining image content in target network content, and inputting the image content into an image identification model to obtain an image violation degree; when the image violation degree is lower than a first threshold value, inputting text content in the target network content into a text recognition model to obtain a text violation degree; when the text violation degree is lower than a second threshold value, determining that the security identification result is target network content security; and when the image violation degree reaches a first threshold value or the text violation degree reaches a third threshold value, determining that the security identification result is target network content danger. Through a multi-stage cascade detection mode, the accuracy of the obtained security identification result is ensured, most simple violation contents are quickly intercepted during first-stage image detection, and only a small number of complex contents need to be completely analyzed, so that the average delay of the system is reduced, and the security detection calculation efficiency is greatly improved.

The invention discloses a text generation content security filtering method and system based on semantic detection, and the method comprises the steps: obtaining and preprocessing an original content text generated based on the text to obtain a preprocessed content text, carrying out the sentence segmentation of the preprocessed content text to obtain a plurality of sentences, and determining the overall semantic features of the sentences in the context based on semantic detection; performing word segmentation on the sentence to obtain a plurality of vocabularies, and determining specific semantic features of each vocabulary in the sentence based on semantic detection; and evaluating the content security of each sentence based on the overall semantic feature and the specific semantic feature to obtain a comprehensive security evaluation value, determining the security filtering level of each sentence based on the comprehensive security evaluation value, and performing security filtering on each sentence in the original content text. According to the method, security detection is carried out on the text generation content through a multi-level and fine-grained semantic analysis and evaluation technology, complex harmful content which is difficult to detect through a traditional method can be effectively recognized, and accurate risk grading and intelligent flexible processing are achieved.

The invention discloses a non-intrusive multi-modal content security auditing system, which relates to the technical field of content data security, and comprises a processing module for generating a first label according to a first dominant feature and a first implicit feature of video data to be audited, generating a second label according to a second dominant feature and a second implicit feature of text data to be audited, the auditing module performs auditing according to the first label and the second label, adopts a non-intrusive design to guarantee the integrity of the original content, fuses video and text multi-mode cooperative auditing, extracts video space-time rhythm features and dominant and implicit risk points of abnormal flicker indexes, captures text violation information by optimizing lexicon matching and format abnormity judgment, and finally performs text violation processing. And then accurate analysis of the modal correlation degree is realized by means of timestamp alignment, and a dynamic threshold adjustment mechanism is matched, so that the pain points of missed judgment and misjudgment, one-sided feature mining and insufficient adaptability of traditional single-modal auditing are effectively solved, and the accuracy, comprehensiveness and flexibility of content security auditing are remarkably improved.

The application discloses a system and method for testing the security of content generated by a large language model in a black box scenario. The system includes a jailbreak prompt word library module for storing jailbreak prompt words for testing the security of a large language model; a violation question and answer pair module for storing violation question and answer pairs of different types; a response collection module for obtaining a data request package from query content composed of jailbreak prompt words and query requests; a security analysis module for calculating the similarity between response data corresponding to the query request and an expected violation answer, and inputting the similarity as a security score into an adaptive optimization module; and the adaptive optimization module for optimizing the jailbreak prompt words output by the jailbreak prompt word library module using a genetic algorithm according to the security score output by the security analysis module. The application can effectively test the security of content generated by a large language model.

The invention provides an automatic auditing method and system for online multi-form content and a medium, and belongs to the field of content security and information processing.The method comprises the steps that to-be-audited content and context are obtained and preprocessed to generate content abstract fingerprints; according to the content and context loading auditing strategy, determining that an auditing path is only rule matching, only semantic understanding or mixed collaboration; a composite cache key is constructed by combining fingerprints and versions to query cache, and processing is performed according to a hit result; when rule matching is included, detection is executed, and a disposal action is directly generated and skipped when a high risk is hit; only during semantic understanding, health scores are calculated through the health monitoring module, and the module is called or backspacing is performed according to the scores; under the mixed path, the fusion device aligns and counts conflicts between rules and semantic output, and generates a comprehensive auditing level, confidence coefficient and evidence abstract; generating a disposal action according to the fusion result and strategy mapping; the action and reason abstract is returned, and the problem of balance among accuracy, low time delay and stability of multi-form content auditing under high concurrency is solved.