A distributed disaster recovery drill data early warning method and system based on artificial intelligence

Abstract

The present invention discloses a distributed disaster recovery drill data early warning method and system based on artificial intelligence. If the local edge node has not trained the model, the central early warning model is uploaded to the local area for training; after receiving the central early warning model, the local edge node server performs correlation data analysis and classification aggregation; after calculating the local prior probability, local conditional probability, local adjustment factor, and local posterior probability, the central early warning model is input for training to obtain the optimal disaster recovery drill alarm probability; the optimal disaster recovery drill alarm probability data of various places is submitted to the central server to update the central early warning model, and the headquarters drill result database is updated to store the updated indicators after each training, and the central server program re-executes a new round of model training, thereby completing the entire process. The present invention takes disaster recovery drill as the main line, historical drill data as the analysis basis, and introduces artificial intelligence to replace manual experience to replace the early warning model, thereby greatly reducing labor costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of disaster recovery drill data processing, and in particular relates to a distributed disaster recovery drill data early warning method and system based on artificial intelligence. Background Art

[0002] Usually, after the disaster recovery system is built, the disasters faced are no more than the data level, application system level and disaster recovery center level. Therefore, all disaster recovery drills are based on a specific scenario at one of these three levels. Different disaster scenarios require different technical processes for the drills. Generally speaking, the initial versions of disaster recovery plans for almost all disaster recovery systems are proposed based on the scenarios assumed by the construction goals. The effectiveness of such disaster recovery plans and whether the processes can meet the needs of the enterprise IT department to deal with disasters require the enterprise IT department to conduct multiple multi-scenario drills to verify the personnel technical reserves, various resource coordination, disaster recovery process organization, etc.

[0003] With the gradual deepening of digital development, network security has become an important guarantee for social development, making disaster recovery drills more valuable for reference. The collection and processing of disaster recovery data information is an important part of disaster recovery drills. Comprehensive information collection and accurate data can ensure the normal execution of various tasks in disaster recovery drills.

[0004] The process of disaster recovery drill is highly close to the process of real disaster, which can ensure that the disaster recovery drill can play a role in work, so that the automatic disaster recovery drill can serve as a reference for data maintenance. However, the disaster recovery drill of existing IT business systems has problems such as incomplete business process information and the impact of privacy data of edge nodes in various places on the training of central models. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and to provide a distributed disaster recovery drill data warning method and system based on artificial intelligence. Through distributed training of the central warning model, the problems of incomplete disaster recovery drill business process information representation of the existing IT business system and the impact of privacy data of edge nodes in various places on the training of the central model are solved.

[0006] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0007] A distributed disaster recovery drill data early warning method based on artificial intelligence, comprising:

[0008] Step 1: The central server program at the headquarters executes the first round of model training instructions and sends the instructions to the local edge node server. After receiving the instructions, the local edge node server performs training model initialization judgment. If the local edge node has not trained the model, the central warning model is uploaded to the local for training. Otherwise, the training continues;

[0009] Step 2: After receiving the central warning model, the local edge node server uses the local exercise privacy data and exercise association database to perform association data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data, so as to form real and effective privacy analysis data;

[0010] Step 3: Calculate the local prior probability, local conditional probability, and local adjustment factor parameters based on valid privacy analysis data and input them into the central early warning model for training. The local disaster recovery drill alarm probability is obtained through training, and then refer to the local posterior probability to obtain the optimal disaster recovery drill alarm probability after balancing various indicators.

[0011] Step 4: Submit the alarm probability data of the optimal disaster recovery drills in various locations to the central server to update the central early warning model, and update the headquarters drill result database to store the updated indicators after each training. The central server program then re-executes a new round of model training to complete the entire process.

[0012] To optimize the above technical solutions, the specific measures taken also include:

[0013] The above step 1 specifically includes:

[0014] Step 101: Execute the first round of central model training start instruction through the central server program deployed at the headquarters, and send the instruction to the local disaster recovery drill edge node server, referred to as the local edge node server;

[0015] Step 102: After receiving the instruction, the local edge node server accesses the model training identification field flag deployed in the local exercise database;

[0016] If the flag is not 0, it means that this is not the first training, and the privacy data analysis and calculation will continue;

[0017] If the flag is 0, it means that the local node has not trained the model, and sends a command to the central server to issue the central early warning model. After receiving the request command, the central server sends the central early warning model to the local edge node server.

[0018] The above step 2 is specifically as follows:

[0019] Through the execution of the program deployed on the edge node servers in various locations, all disaster recovery drill data in the local drill database are classified according to the drill steps, and then the drill association database of the central server is accessed based on the drill steps as the query conditions to obtain the same drill step data associated with the local and other edge node servers in different locations, access the associated remote drill database to obtain the associated privacy data, and merge the privacy data according to the drill steps.

[0020] The above-mentioned drill steps include incident reporting, disaster assessment, and disaster declaration classification.

[0021] In the above step 301, the local prior probability, the local conditional probability, the local adjustment factor, and the local posterior probability are calculated based on the valid privacy analysis data;

[0022] Step 302: Input the local prior probability, local conditional probability, local adjustment factor, and local posterior probability data obtained in step 301 into the central early warning model for training. The alarm probability of the local disaster recovery drill is obtained through training, and then the optimal disaster probability indicator after balancing various indicators is obtained by referring to the local posterior probability.

[0023] The local prior probability in step 301 above = the total number of privacy data in the current exercise step;

[0024] Local conditional probability = the current drill step privacy data occurrence cycle, disaster assessment level, drill content, disaster declaration keywords, disaster recovery drill association table and other dimensions, and the data statistics are used to obtain the conditional probability of this training;

[0025] Local adjustment factor = number of false positives of privacy data in the current drill step / local prior probability;

[0026] Local posterior probability = local prior probability x local adjustment factor.

[0027] The above step 302 obtains the optimal disaster occurrence probability index through the central warning model. The central warning model formula is:

[0028] ZY(D|+)=ZY(+|D)ZY(D) / (ZY(+|D)ZY(D)+ZY(+|N)ZY(N))

[0029] ZY(D|+) is the probability that the disaster recovery drill is close to reality considering the false alarm rate;

[0030] ZY(+|D) is the accuracy rate when the disaster recovery drill occurs = 1-false alarm rate;

[0031] ZY(+|N) is the false alarm rate of disaster recovery drill data = number of false alarms in disaster recovery drills / total number of disaster recovery drills

[0032] ZY(N) = represents the probability of no failure in the disaster recovery drill, which is 1-ZY(D).

[0033] A distributed disaster recovery drill data early warning system based on artificial intelligence, including:

[0034] The central warning model module is used by the central server program of the headquarters to execute the first round of model training instructions and send the instructions to the local edge node server. After receiving the instructions, the local edge node server will perform initialization judgment on the training model. If the local edge node has not trained the model, the central warning model will be uploaded to the local for training. Otherwise, the training will continue.

[0035] The data clustering analysis module is used for the local edge node server to perform correlation data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data after receiving the central warning model, using local exercise privacy data and exercise correlation database to form real and effective privacy analysis data;

[0036] The probability calculation module is used to calculate the local prior probability, local conditional probability, local adjustment factor, and local posterior probability parameters based on the effective privacy analysis data, and then input them into the central early warning model for training. The alarm probability of the local disaster recovery drill is obtained through training, and then the optimal disaster recovery drill alarm probability after balancing various indicators is obtained by referring to the local posterior probability;

[0037] The data update module is used to submit the alarm probability data of the optimal disaster recovery drills in various places to the central server to update the central early warning model, and to update the headquarters drill result database, storing the updated indicators after each training. The central server program then re-executes a new round of model training to complete the entire process.

[0038] The present invention has the following beneficial effects:

[0039] The present invention takes disaster recovery drill as the main line and historical drill data as the analysis basis, and introduces artificial intelligence to replace manual experience to replace the early warning model, thereby greatly reducing labor costs; comprehensively adopts distributed data convergence and utilizes edge nodes in local networks to train the central model, thereby optimizing the central training model and greatly reducing the central model training load, that is, highlighting the advantages of artificial intelligence in the disaster recovery drill process, and can achieve the protection of privacy data of distributed edge nodes in the network without affecting the effect of central model training, and greatly reducing the central model training load; comprehensive data training is performed on the disaster recovery drill business process information corresponding to the edge node, thereby obtaining the alarm probability of disaster recovery drill in the edge node area, which can effectively overcome the model deviation problem and reduce the calculation complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Flow chart of the method in the present invention. DETAILED DESCRIPTION

[0041] The embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0042] See also Figure 1 , a distributed disaster recovery drill data early warning method based on artificial intelligence, is described as follows:

[0043] Step 1: The central server program at the headquarters executes the first round of model training instructions and sends the instructions to the local edge node server. After receiving the instructions, the local edge node server performs training model initialization judgment. If the local edge node has not trained the model, the central warning model is uploaded to the local for training. Otherwise, the training continues;

[0044] The present invention relates to a drill result database, a local drill database and a drill association database: a central server deploys a drill result database and a drill association database, and a local server deploys a local drill database.

[0045] Step 1 specifically includes:

[0046] Step 101: Execute the first round of central model training start instruction through the central server program deployed at the headquarters, and send the instruction to the local disaster recovery drill edge node server, referred to as the local edge node server;

[0047] Step 102: After receiving the instruction, the local edge node server accesses the model training identification field flag deployed in the local exercise database;

[0048] If the flag is not 0, it means that this is not the first training, and the privacy data analysis and calculation will continue;

[0049] If the flag is 0, it means that the local node has not trained the model, and sends the [central early warning model] instruction to the central server. After receiving the request instruction, the central server sends the [central early warning model] to the local edge node server.

[0050] Step 2: After receiving the central warning model, the local edge node server uses the local exercise privacy data and exercise association database to perform association data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data, so as to form real and effective privacy analysis data;

[0051] After receiving the [central early warning model], the local edge node server aggregates the privacy data, that is, classifies and aggregates the local drill data and the database related to the drill to form real and effective privacy analysis data.

[0052] The step 2 is specifically as follows:

[0053] Through the execution of the program deployed on the edge node servers in various places, all the disaster recovery drill data in the local drill database are classified according to the drill steps, and then the drill association database of the central server is accessed with the drill steps as the query conditions to obtain the same drill step data associated with the local and other edge node servers in different places, access the related remote drill database to obtain the related privacy data, and merge the privacy data according to the drill steps;

[0054] The drill steps include event reporting, disaster assessment, and disaster declaration classification.

[0055] Step 3: Calculate the [local prior probability], [local conditional probability], and [local adjustment factor] parameters based on valid privacy analysis data and input them into the [central early warning model] for training. The local [disaster recovery drill alarm probability] is obtained through training, and then refer to the [local posterior probability] to obtain the [optimal disaster recovery drill alarm probability] after balancing various indicators;

[0056] include:

[0057] Step 301: Calculate [local prior probability], [local conditional probability], [local adjustment factor], and [local posterior probability] based on valid privacy analysis data;

[0058] Where, [local prior probability] = the total number of private data in the current drill step;

[0059] [Local conditional probability] = the current drill step privacy data occurrence cycle, disaster assessment level, drill content, disaster declaration keywords, disaster recovery drill association table and other dimensions, and the data statistics are used to obtain the conditional probability of this training;

[0060] [Local adjustment factor] = number of false positives of privacy data in the current drill step / [local prior probability];

[0061] [Local posterior probability] = [Local prior probability] x [Local adjustment factor].

[0062] Step 302, [Central Early Warning Module]:

[0063] The [local prior probability], [local conditional probability], and [local adjustment factor] data obtained in step 301 are input into the central early warning model for training. The local [disaster recovery drill alarm probability] is obtained through training, and then the [local posterior probability] is referred to to obtain the [optimal disaster occurrence probability indicator] after balancing various indicators.

[0064] Model mathematical formula:

[0065] ZY(D|+)=ZY(+|D)ZY(D) / (ZY(+|D)ZY(D)+ZY(+|N)ZY(N))

[0066] Obtain the [optimal disaster probability index] through model calculation

[0067] ZY(D|+) is the [local prior probability];

[0068] ZY(+|D) is the [local conditional probability];

[0069] ZY(+|N) is the [local adjustment factor];

[0070] Related parameter description:

[0071] [Local posterior probability] = [Local prior probability] x [Local adjustment factor]; the mathematical formula is described as:

[0072] ZY(+|D)x adjustment factor

[0073]

D

[0074]

/

[0075]

ZY

[0076]

ZY(D|+)

[0077]

ZY(+|D)

[0078] Formula Description:

[0079]

ZY(+|N)

[0080] Disaster recovery drill data false alarm rate = number of disaster recovery drill false alarms / total number of disaster recovery drills

[0081]

ZY(N)

[0082] Step 4: Submit the [optimal disaster recovery drill alarm probability] data from various locations to the central server to update the [central early warning model], and update the headquarters drill result database to store the updated indicators after each training. The central server program will re-execute a new round of model training to complete the entire process.

[0083] A distributed disaster recovery drill data early warning system based on artificial intelligence, characterized by comprising:

[0084] The central warning model module is used by the central server program of the headquarters to execute the first round of model training instructions and send the instructions to the local edge node server. After receiving the instructions, the local edge node server will perform initialization judgment on the training model. If the local edge node has not trained the model, the central warning model will be uploaded to the local for training. Otherwise, the training will continue.

[0085] The data clustering analysis module is used for the local edge node server to perform correlation data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data after receiving the central warning model, using local exercise privacy data and exercise correlation database to form real and effective privacy analysis data;

[0086] The probability calculation module is used to calculate the local prior probability, local conditional probability, local adjustment factor, and local posterior probability parameters based on the effective privacy analysis data, and then input them into the central early warning model for training. The alarm probability of the local disaster recovery drill is obtained through training, and then the optimal disaster recovery drill alarm probability after balancing various indicators is obtained by referring to the local posterior probability;

[0087] The data update module is used to submit the alarm probability data of the optimal disaster recovery drills in various places to the central server to update the central early warning model, and to update the headquarters drill result database, storing the updated indicators after each training. The central server program then re-executes a new round of model training to complete the entire process.

[0088] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A distributed disaster recovery drill data early warning method based on artificial intelligence, characterized in that: include: Step 1: The central server program at the headquarters executes the first round of model training instructions and sends the instructions to the local edge node server. After receiving the instructions, the local edge node server performs training model initialization judgment. If the local edge node has not trained the model, the central warning model is uploaded to the local for training. Otherwise, the training continues; Step 2: After receiving the central warning model, the local edge node server uses the local exercise privacy data and exercise association database to perform association data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data, so as to form real and effective privacy analysis data; Step 3: Calculate the local prior probability, local conditional probability, and local adjustment factor parameters based on valid privacy analysis data and input them into the central early warning model for training. The local disaster recovery drill alarm probability is obtained through training, and then refer to the local posterior probability to obtain the optimal disaster recovery drill alarm probability after balancing various indicators. The local prior probability = the total number of privacy data in the current drill step; Local conditional probability = derive the conditional probability of this training by performing data statistics on the dimensions of the current drill step privacy data occurrence cycle, disaster assessment level, drill content, disaster declaration keywords, and disaster recovery drill association table; Local adjustment factor = number of false positives of privacy data in the current drill step / local prior probability; Local posterior probability = local prior probability x local adjustment factor; The alarm probability of the optimal disaster recovery drill is obtained through the central early warning model. The formula of the central early warning model is: ZY(D|+) = ZY(+|D)ZY(D) / (ZY(+|D)ZY(D)+ZY(+|N)ZY(N)) ZY(D|+) is the local prior probability; ZY (+|D) is the local conditional probability; ZY(+|N) is the local adjustment factor; ZY(N)= represents the probability of no failure in the disaster recovery drill, which is 1-ZY(D); Step 4: Submit the alarm probability data of the optimal disaster recovery drills in various locations to the central server to update the central early warning model, and update the headquarters drill result database to store the updated indicators after each training. The central server program then re-executes a new round of model training to complete the entire process.

2. According to the artificial intelligence-based distributed disaster recovery drill data early warning method of claim 1, it is characterized in that: The step 1 specifically includes: Step 101: Execute the first round of central model training start instruction through the central server program deployed at the headquarters, and send the instruction to the local disaster recovery drill edge node server, referred to as the local edge node server; Step 102: After receiving the instruction, the local edge node server accesses the model training identification field flag deployed in the local exercise database; If the flag is not 0, it means that this is not the first training, and the privacy data analysis and calculation will continue; If the flag is 0, it means that the local node has not trained the model, and sends a command to the central server to issue the central early warning model. After receiving the request command, the central server sends the central early warning model to the local edge node server.

3. The artificial intelligence-based distributed disaster recovery drill data early warning method according to claim 1 is characterized in that: The step 2 is specifically as follows: Through the execution of the program deployed on the edge node servers in various locations, all disaster recovery drill data in the local drill database are classified according to the drill steps, and then the drill association database of the central server is accessed based on the drill steps as the query conditions to obtain the same drill step data associated with the local and other edge node servers in different locations, access the associated remote drill database to obtain the associated privacy data, and merge the privacy data according to the drill steps.

4. The artificial intelligence-based distributed disaster recovery drill data early warning method according to claim 3 is characterized in that: The drill steps include event reporting, disaster assessment, and disaster declaration classification.

5. A distributed disaster recovery drill data early warning system based on artificial intelligence, characterized in that: include: The central warning model module is used by the central server program of the headquarters to execute the first round of model training instructions and send the instructions to the local edge node server. After receiving the instructions, the local edge node server will perform initialization judgment on the training model. If the local edge node has not trained the model, the central warning model will be uploaded to the local for training. Otherwise, the training will continue. The data clustering analysis module is used for the local edge node server to perform correlation data analysis and classification aggregation based on the correlation between local privacy data and remote privacy data after receiving the central warning model, using local exercise privacy data and exercise correlation database to form real and effective privacy analysis data; The probability calculation module is used to calculate the local prior probability, local conditional probability, local adjustment factor, and local posterior probability parameters based on the effective privacy analysis data, and then input them into the central early warning model for training. The alarm probability of the local disaster recovery drill is obtained through training, and then the optimal disaster recovery drill alarm probability after balancing various indicators is obtained by referring to the local posterior probability; The local prior probability = the total number of privacy data in the current drill step; Local conditional probability = the current drill step privacy data occurrence cycle, disaster assessment level, drill content, disaster declaration keywords, disaster recovery drill association table and other dimensions, and the data statistics are used to obtain the conditional probability of this training; Local adjustment factor = number of false positives of privacy data in the current drill step / local prior probability; Local posterior probability = local prior probability x local adjustment factor; The alarm probability of the optimal disaster recovery drill is obtained through the central early warning model. The formula of the central early warning model is: ZY(D|+) = ZY(+|D)ZY(D) / (ZY(+|D)ZY(D)+ZY(+|N)ZY(N)) ZY(D|+) is the local prior probability; ZY (+|D) is the local conditional probability; ZY(+|N) is the local adjustment factor; ZY(N)= represents the probability of no failure in the disaster recovery drill, which is 1-ZY(D); The data update module is used to submit the alarm probability data of the optimal disaster recovery drills in various places to the central server to update the central early warning model, and to update the headquarters drill result database, storing the updated indicators after each training. The central server program then re-executes a new round of model training to complete the entire process.

Images