Search

CN-121981538-A - Intelligent security management method and system for AI artificial intelligence training room

CN121981538ACN 121981538 ACN121981538 ACN 121981538ACN-121981538-A

Abstract

The application provides an AI artificial intelligence practical training room intelligent safety management method and system, relating to the technical field of practical training room safety management, wherein the method comprises the steps of establishing a three-dimensional safety perception matrix, wherein the three-dimensional safety perception matrix comprises a data flow dimension, an energy flow dimension and a personnel flow dimension, carrying out entropy weighting on elements of the three-dimensional safety perception matrix to calculate an initial risk value, inputting the three-dimensional safety perception matrix into a space-time causal reasoning engine, and correcting the initial risk value to obtain a risk entropy value; when the risk entropy value breaks through the dynamic threshold value, an alarm signal is sent out and a prevention and control measure is started. The application can improve the intelligent level of the practical training room management process.

Inventors

  • JIA SIJIE

Assignees

  • 山东九芯科技发展有限公司

Dates

Publication Date
20260505
Application Date
20260116

Claims (10)

  1. 1. An intelligent security management method for an AI artificial intelligence training room is characterized by comprising the following steps: Establishing a three-dimensional safety perception matrix, wherein the three-dimensional safety perception matrix comprises a data flow dimension, an energy flow dimension and a personnel flow dimension, performing entropy weighting on elements of the three-dimensional safety perception matrix to calculate an initial risk value, inputting the three-dimensional safety perception matrix into a space-time causal reasoning engine, and correcting the initial risk value to obtain a risk entropy value; When the risk entropy value breaks through the dynamic threshold value, an alarm signal is sent out and a prevention and control measure is started.
  2. 2. The AI artificial intelligence training room intelligent security management method of claim 1, wherein the prevention and control measures include isolating target devices and freezing associated data, generating an antagonism test sample to verify user intent, reconstructing an access control policy topology.
  3. 3. The AI artificial intelligence training room intelligent security management method of claim 2, wherein the challenge test sample is generated using WASSERSTEIN GAN framework and the generator inputs a historical sequence of operations for the current user.
  4. 4. The AI artificial intelligence training room intelligent security management method of claim 2, wherein reconstructing the access control policy topology comprises: Analyzing the data interaction frequency and the authority use mode of each equipment node in real time; dynamically updating the connection weight among nodes based on the graph rolling network; When an abnormal access path is detected, the high risk connection is automatically cut off and an alternative path is generated.
  5. 5. The AI artificial intelligence training room intelligent security management method of claim 4, wherein upon reconstructing the access control policy topology, the method further comprises: Updating the equipment access right in real time by adopting a dynamic attribute base encryption algorithm; automatically blocking an abnormal data transmission link through network traffic fingerprint identification; a temporary access token of the least privilege principle is generated, and the validity period of the temporary access token is inversely related to the risk entropy value.
  6. 6. The AI artificial intelligence training room intelligent safety management method of claim 1, wherein the spatio-temporal causal reasoning engine employs a causal discovery algorithm to construct a bayesian network, and nodes of the bayesian network include abnormal access frequency in a data flow dimension, harmonic distortion rate in an energy flow dimension, and motion trail entropy in a personnel flow dimension.
  7. 7. The AI artificial intelligence training room intelligent security management method of claim 1, further comprising: Establishing a safety knowledge graph, and carrying out structural storage on the historical risk event according to threat types, treatment measures and effect evaluation; When a new risk mode is detected, similar historical cases are searched in the safety knowledge graph, and Top-K treatment schemes are output.
  8. 8. The AI artificial intelligence training room intelligent security management method of claim 7, wherein searching for similar historical cases in a security knowledge graph comprises: Extracting a risk feature vector from a three-dimensional safety perception matrix, and calculating the similarity between the risk feature vector and a historical case node in a safety knowledge graph; and carrying out self-adaptive sorting on the Top-K treatment schemes according to the similarity, and feeding back sorting results.
  9. 9. The AI artificial intelligence training room intelligent security management method of claim 1, wherein the dynamic threshold determining and updating method comprises: generating a baseline threshold value through a preset BI-LSTM model based on the historical risk entropy value sequence; And monitoring the change rate of each dimension in the three-dimensional safety perception matrix in real time, dynamically reducing the baseline threshold according to a predefined rule when the change rate of any dimension exceeds a preset acceleration, and correcting the dynamic threshold by adopting the reduced baseline threshold.
  10. 10. The AI artificial intelligence practical training room intelligent safety management system is characterized by comprising a processor and a memory which is in communication connection with the processor; a computer readable storage medium is arranged in the memory, and a computer program is stored on the computer readable storage medium; The processor, when processing a computer program stored on the computer readable storage medium, implements the method according to any of claims 1-9.

Description

Intelligent security management method and system for AI artificial intelligence training room Technical Field The application relates to the technical field of practical training room safety management, in particular to an AI artificial intelligent practical training room intelligent safety management method and system. Background With the rapid development of artificial intelligence technology, AI artificial intelligence training rooms are increasingly widely used in universities, scientific research institutions and enterprise research and development departments. These training rooms are not only equipped with a large number of high-value electronic devices, such as high-performance servers, dedicated AI accelerator cards, etc., but also store a large amount of sensitive data concerning intellectual property, business confidentiality, and personal privacy. Meanwhile, the practical training process involves complex electric circuits, high-temperature running equipment and experimental links which possibly generate harmful substances, so that the practical training room faces multiple safety risks such as fire disaster, electric faults, data leakage, personnel accidental injury and the like. The application publication number is CN116777699A, a practical training room intelligent safety management system for campus practical training is provided, the system divides safety management into two aspects of artificial safety and practical training room safety for respective evaluation, thereby improving the comprehensiveness of the practical training room safety evaluation through the combination of people and objects, evaluating the artificial safety, preventing personnel from being mixed, avoiding accidents caused by excessively small number of instructors and incapability of comprehensively supervising the practical training personnel when evaluating the instructors, and dividing the safety management into two angles of equipment safety and monitoring comprehensiveness for management when the practical training room is subjected to safety management, and knowing the danger of the equipment. The system cannot understand the complex causal relationship among risk elements, is difficult to quantify the dynamic evolution of risks, and further lacks the capability of active intervention and intelligent treatment at the time of risk initiation. Disclosure of Invention In order to improve the intelligent level of the practical training room management process, the application provides an AI artificial intelligent practical training room intelligent safety management method and system. In a first aspect, the application provides an intelligent safety management method for an AI artificial intelligent training room, which adopts the following technical scheme: an AI artificial intelligence practical training room intelligent safety management method comprises the following steps: establishing a three-dimensional safety perception matrix, wherein the three-dimensional safety perception matrix comprises a data flow dimension, an energy flow dimension and a personnel flow dimension, performing entropy weighting on elements of the three-dimensional safety perception matrix to calculate an initial risk value, inputting the three-dimensional safety perception matrix into a space-time causal reasoning engine, and correcting the initial risk value to obtain a risk entropy value; When the risk entropy value breaks through the dynamic threshold value, an alarm signal is sent out and a prevention and control measure is started. According to the application, the three-dimensional safety perception matrix comprising the data flow dimension, the energy flow dimension and the personnel flow dimension is established, so that the limitation of traditional single-dimension risk perception is broken through, the comprehensive and multi-level risk perception of a complex system is realized, and the risk missing judgment caused by the fact that the single dimension is considered to be not round is effectively avoided. And then, weighting each dimension element by utilizing an entropy weight method based on the three-dimensional safety perception matrix, and calculating an initial risk value, so that the relative importance of different dimension elements in the overall risk is fully considered. The entropy weight method determines the weight according to the variation degree of the data of each element, can objectively reflect the contribution degree of each element to the risk, and enables the calculation of the initial risk value to be more scientific and reasonable. And then, the three-dimensional safety perception matrix is input into a space-time causal reasoning engine to correct the initial risk value, so that the accuracy of risk assessment is improved. The space-time causal reasoning engine can comprehensively consider the influence of time and space factors on risks, mining causal relations among elements in different dimensio