Data privacy protection processing method and device using deep learning

Abstract

The invention belongs to the technical field of data privacy protection, and particularly relates to a data privacy protection processing method and device using deep learning. Pictures possibly causing secret leakage are communicated with information of a photographer to serve as a privacy information packet to be backed up, and the privacy information packet is suitable for secrecy departments or normal enterprises which need to confidentialize working addresses when working, so that the pictures which need to be published in the working process, including pictures shot by employees and required by enterprise propaganda, are not required to be published by the employees and are required by enterprise propaganda. The pictures are processed by using the processing method, and part of the pictures possibly causing privacy information leakage are found out, so that the confidentiality requirement of work development is met; especially for picture background parts which are easy to ignore in manual screening, objects with identification features may cause leakage of address privacy information.

Description

Technical Field

[0001] The present invention belongs to the technical field of data privacy protection, and specifically, is a data privacy protection processing method and device using deep learning. Background Art

[0002] In today's life, taking photos has become a common way for us to record daily life and share wonderful moments. Whether it is the beautiful scenery during the journey, the warm scenes in life, or the novel things encountered by chance, everyone will habitually pick up the phone or camera to press the shutter, and then share these photos on social platforms to experience those beautiful moments with relatives and friends. However, many people do not realize that a seemingly ordinary photo often contains a lot of information, and if you are not careful, it may become the source of leakage.

[0003] In scenarios where private information such as addresses needs to be kept confidential, because most people do not have the knowledge to keep information confidential, they usually manually blur the people and text in the photos, while ignoring the buildings, plants and other things with identifiable features in the background, which leads to the leakage of private information. Summary of the invention

[0004] In order to make up for the deficiencies of the prior art and solve the above-mentioned technical problems, the present invention proposes a data privacy protection processing method and device using deep learning.

[0005] The technical solution adopted by the present invention to solve the technical problem is: the present invention proposes a data privacy protection processing method using deep learning, and the specific steps of the processing method are: S1: Obtain a target image through an information input client, and analyze the person image in the target image through a person recognition module to identify the person's identity information; S2: Based on the confidential private person identity information pre-input into the privacy database, compare the person information in the target image, and perform desensitization processing on the person image information in the target image that needs to be kept confidential; S3: Based on the deep learning module training, a background recognition module is obtained. The background recognition module is then used to cut out the background image part other than the facial image of the person in the target image and analyze it separately. The confidential targets contained in the background image are screened out and the confidential targets are desensitized. S4: For the target images containing confidential personal identity information and background images containing confidential targets, the client information, time information of uploading the target image and location information are integrated to obtain a privacy information package, which is transmitted to the management end, and the privacy information package is backed up to the data storage module in the data processing server, and the privacy information package is encrypted and read permissions are set.

[0006] Preferably, the confidential target types in the above step S3 include characteristic buildings, starry sky and characteristic plants.

[0007] Preferably, the training process of the confidential target model in the deep learning module includes: S31: When the confidential object is the address where the target image was taken, collect pictures of the surrounding area of ​​the address where the image was taken, including pictures of surrounding buildings and plants with identification features, and use these pictures as test pictures to form a training set; S32: The images in the training set are transferred to the lightweight deep learning training framework Darknet, and a neural network model capable of screening images with recognition features is obtained through training; S33: randomly download pictures of buildings and plants of the same type but without identification features from the Internet, and mix these pictures with the pictures in the training set as the test set; S34: input the test set images into the trained neural network model, so that it can screen the images with recognition features, select the parts of the screened images that belong to the original training set for calculation, and obtain the recognition accuracy of the neural network model; S35: When the recognition accuracy is higher than 95%, the neural network model can be deployed to the data processing server as the core of the deep learning module to participate in the screening of confidential targets in the target image.

[0008] A data privacy protection processing device using deep learning is applicable to the above-mentioned data privacy protection processing method, comprising a data processing server and a data storage module, wherein the data storage module comprises a data storage device, the data processing server comprises a processing box, the processing box is fixed on the floor of a machine room, the data storage device is mounted on a storage bracket inside the processing box, and is connected to a background recognition module inside the processing box through a data conductor; A transfer port is provided at a position corresponding to the storage bracket at the bottom of the processing box, and the storage bracket is fixed to the upper surface of a closing plate slidably arranged inside the transfer port; a transfer groove is provided on the floor of the machine room at a position on the lower side of the processing box facing the transfer port, and the closing plate at the top of the transfer groove is fixedly connected to the closing plate, and the closing plate is connected to a telescopic device arranged at the bottom of the transfer groove, so that the data storage device is physically cut off by transferring the data storage device to the transfer groove.

[0009] Preferably, the storage bracket includes a top plate and a bottom plate, and a side plate is vertically arranged between the top plate and the bottom plate to realize the connection between the top plate and the bottom plate; Partitions are evenly arranged between the top plate and the bottom plate to evenly divide the area between the top plate and the bottom plate into a plurality of storage cavities, the data storage is fixed inside the storage cavity, and connection grooves are arranged on both sides of the storage cavity, and the data wire passes through the connection groove to realize the connection between the background recognition module and the data storage; The inner wall of the connection groove is provided with a rotary cutter, which is connected to the rotating device on the inner wall of the connection groove to cut off the data wire.

[0010] Preferably, a temperature control cavity is provided at the position of the floor of the machine room on both sides of the storage cavity, and the temperature control cavity is filled with a cooling medium; The partition extends from both sides of the connecting groove to pass through the side plates and protrude outward, and a blocking plate is provided at a position of the inner wall of the storage cavity corresponding to the partition, and the blocking plate is slidably connected to the side wall of the storage cavity; one end of the blocking plate extends into the temperature control cavity and is connected to the propulsion device inside the temperature control cavity, and the other end is embedded in the storage cavity.

[0011] Preferably, the top plate is hollow inside to form an isolation cavity, and the top plate is a double-layer structure, the portion located on the upper side of the isolation cavity is a heat insulation layer made of marble, and the portion located on the lower side of the isolation cavity is a fixed layer made of metal; A fixing hole is provided on the inner wall of the top of the transfer groove, and a fixing pipe is provided inside the fixing hole. The fixing pipe is connected to the output end of the propulsion device, and the fixing pipe extends into the temperature control cavity and contacts with the cooling medium; guide holes are provided at the positions corresponding to the fixing holes on both sides of the top plate.

[0012] Preferably, the interior of the partition is hollow to form a cooling cavity, and a connecting port is provided at the end of the partition corresponding to the end of the blocking plate, and a transmission pipe is provided on the end of the blocking plate located inside the storage cavity. The transmission pipe is part of the gas fire extinguishing system and is responsible for transporting the fire extinguishing gas, and the pipe mouth of the blocking plate extends out from the protruding end facing the partition.

[0013] Preferably, a clamping groove is provided at a portion of the end of the blocking plate corresponding to the protruding end of the partition plate, and the transmission pipe is located on the inner wall of the clamping groove; A closed groove is arranged on the inner wall of the clamping groove, and the cross section of the closed groove is arc-shaped; a closed strip is arranged at a corresponding position of the protruding end of the partition plate, and the closed strip is made of elastic material.

[0014] The beneficial effects of the present invention are as follows: The data privacy protection processing method and device using deep learning described in the present invention are applicable to confidential units or normal enterprises that need to keep their work addresses confidential. Therefore, for pictures that need to be made public during work, including pictures taken by employees and pictures needed for corporate publicity, these pictures are processed using the above-mentioned processing method to identify some pictures that may cause privacy information leakage, which meets the confidentiality needs of the work; especially for the background part of the picture that is easily overlooked in manual screening, where things with identification features may cause the leakage of address privacy information. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings.

[0016] Figure 1 is a flow chart of the data privacy protection processing method in the present invention; Figure 2 It is a flow chart of the confidential target model training of the deep learning module in the present invention; Figure 3 is a three-dimensional diagram of the data privacy protection processing device of the present invention; Figure 4 is a partial cross-sectional view of the data privacy protection processing device of the present invention; Figure 5 yes Figure 4 A partial enlarged view of the middle A; Figure 6 yes Figure 4 A partial enlarged view of point B in the middle; Figure 7 is a three-dimensional diagram of the storage bracket of the present invention; Figure 8 It is a three-dimensional diagram of the blocking plate in the present invention.

[0017] In the figure: data processing server 1, processing box 11, transfer port 12, transfer plate 13, data storage device 2, data wire 21, storage bracket 3, top plate 31, isolation cavity 311, guide hole 312, insulation layer 313, fixing layer 314, bottom plate 32, side plate 33, partition 34, cooling cavity 341, connecting port 342, closing strip 343, storage cavity 35, connecting groove 36, rotating cutter 361, transfer groove 4, closing plate 41, temperature control cavity 42, blocking plate 43, transmission tube 431, clamping groove 432, closing groove 433, fixing hole 44, fixing tube 441. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: As shown in the attached figure of the specification Figure 1-Figure 2 As shown, in some scenarios where private information such as addresses needs to be kept confidential, the people and text in the photos are usually blurred manually, while buildings, plants and other things with identifiable features in the background are ignored, resulting in the leakage of private information. Therefore, this application proposes a data privacy protection processing method using deep learning. The specific steps of the processing method are: S1: Obtain a target image through an information input client, and analyze the person image in the target image through a person recognition module to identify the person's identity information; S2: Based on the confidential private person identity information pre-input into the privacy database, compare the person information in the target image, and perform desensitization processing on the person image information in the target image that needs to be kept confidential; S3: Based on the deep learning module training, a background recognition module is obtained. The background recognition module is then used to cut out the background image part other than the facial image of the person in the target image and analyze it separately. The confidential targets contained in the background image are screened out and the confidential targets are desensitized. S4: For the target image containing confidential person identity information and the target image containing confidential targets in the background image, the client information, the time information of uploading the target image and the location information are integrated to obtain a privacy information package and transmitted to the management end, the privacy information package is backed up to the data storage module in the data processing server 1, and the privacy information package is encrypted and the read permission is set.

[0020] For example, this application is applicable to confidential units or normal enterprises that need to keep their work addresses confidential. Therefore, for pictures that need to be made public during work, including pictures taken by employees and pictures needed for corporate publicity, these pictures are processed using the above processing method to screen out some pictures that may cause privacy information leakage, which meets the confidentiality needs of work; especially for the background part of the picture that is easily overlooked in manual screening, things with identification features may cause the leakage of address privacy information; Therefore, the image information collected by media devices such as employees' mobile phones, computers, and cameras used for promotional photography needs to be uploaded to the data processing server 1 through the client for processing, wherein the video information is converted into pictures, and the target image information in the pictures is extracted. The background recognition module is used to identify and screen possible confidential targets. Pictures that may cause leaks are backed up as privacy information packages together with the photographer's information. At the same time, they are sent to the information security department as the management end to warn or even hold accountable those who may cause confidentiality. If necessary, the confidential targets can be desensitized, for example, by using Gaussian filtering to blur the image information, thereby ensuring the safe use of privacy information.

[0021] Among them, the types of confidential targets in the above step S3 include characteristic buildings, starry sky and characteristic plants. For the starry sky part in the background part of the picture, the photography-related starry sky recognition software included in the background recognition module can recognize the starry sky part and perform blur processing, so that the public cannot locate the shooting location through the background starry sky after the picture is made public; for things such as buildings and plants with unique local characteristics, they are identified through the background recognition module with the deep learning module as the core. Specifically, the training process of the confidential target model in the deep learning module includes: S31: Based on the confidentiality requirement of the shooting address of the target image, collect pictures of the surrounding area of ​​the shooting address, including pictures of surrounding buildings and plants with identification features, and use these pictures as test pictures to form a training set; S32: The images in the training set are transferred to the lightweight deep learning training framework Darknet, and a neural network model capable of screening images with recognition features is obtained through training; S33: randomly download pictures of buildings and plants of the same type but without identification features from the Internet, and mix these pictures with the pictures in the training set as the test set; S34: input the test set images into the trained neural network model, so that it can screen the images with recognition features, select the parts of the screened images that belong to the original training set for calculation, and obtain the recognition accuracy of the neural network model; S35: When the recognition accuracy is higher than 95%, the neural network model can be deployed to the data processing server 1 as the core of the deep learning module to participate in the screening of confidential targets in the target image.

[0022] Furthermore, in addition to the characteristic buildings, starry sky and characteristic plants mentioned above, the recognition objects of the deep learning module can also include the mountain features, geographical features, etc. of the location. By inputting as many things with local identification features as possible according to the actual situation, the confidentiality and security of user privacy information can be improved.

[0023] Embodiment 2: As shown in the attached figure of the specification Figure 3-Figure 8 As shown, based on the above-mentioned data privacy protection processing method, in order to further realize the security of privacy data storage, the present application provides a data privacy protection processing device using deep learning, including a data processing server 1 and a data storage module, the data storage module includes a data storage 2, the data processing server 1 includes a processing box 11, the processing box 11 is fixed on the floor of the machine room, the data storage 2 is installed on a storage bracket 3 inside the processing box 11, and is connected to the background recognition module inside the processing box 11 through a data conductor 21; A transfer port 12 is provided at the bottom of the processing box 11 at a position corresponding to the storage bracket 3, and the storage bracket 3 is fixed on the upper surface of a closing plate 41 slidingly arranged inside the transfer port 12; a transfer slot 4 is provided on the floor of the machine room at a position on the lower side of the processing box 11 facing the transfer port 12, and the closing plate 41 at the top of the transfer slot 4 is fixedly connected to the bottom of the storage bracket 3, and the closing plate 41 is connected to a telescopic device arranged at the bottom of the transfer slot 4; the main body of the telescopic device is located at the bottom of the transfer slot 4, and the output end is connected to the closing plate 41. After startup, the data storage device 2 can be transferred to the transfer slot 4 to achieve physical disconnection of the data storage device 2.

[0024] Specific workflow: Based on the specific workflow in the first embodiment, in the above-mentioned privacy processing of information, the background recognition module installed in the data processing server 1 is mainly used to provide computing support, and then the processed information is formed into an information package and stored in the data storage device 2 hardware in the information storage module; the data storage device 2 is located inside the processing box 11. When there is an abnormal theft of the information package inside the data storage device 2, on the one hand, an alarm is issued to notify the maintenance personnel to track the information thief and prevent the theft of the information package; on the other hand, in order to ensure the security of various privacy information inside the information package, the data storage device 2 can be directly physically cut off; Regarding the physical cut-off method, the present application provides a possible implementation scheme. Specifically, the telescopic device inside the transfer slot 4 is directly started, so that the closing plate 41 on the top opening of the transfer slot 4 drives the connected storage bracket 3 to move downward. As the closing plate 41 reaches the bottom of the transfer slot 4, the top of the storage bracket 3 moves to the top opening position of the transfer slot 4, so as to close the top opening of the transfer slot 4. In this way, the data storage device 2 on the storage bracket 3 directly leaves the processing server and is completely disconnected from the data connection with the background recognition module network, so as to achieve physical disconnection and directly block the theft of privacy information packets. Moreover, because the data storage device 2 is moved to the inside of the transfer slot 4, and the inner wall of the transfer slot 4 is made of solid steel plates and marble, and can only be unlocked by a special password after the movement is completed, the storage bracket 3 is reset. In this way, even if the thief enters the computer room, he cannot access the data storage device 2, so as to more effectively prevent the illegal theft of privacy information packets in the data storage device 2. Furthermore, because the transfer slot 4 is sturdy, when a fire occurs in the computer room and the situation is urgent and personnel need to be transferred first, and the data storage device 2 cannot be transferred in time, the telescopic device can be directly started to move the data storage device 2 and the storage bracket 3 to the inside of the transfer slot 4, so as to isolate the internal environment of the computer room and prevent fire and high temperature from invading the transfer slot 4, thereby better protecting the internal data storage device 2 and properly protecting the privacy information package, thereby avoiding the large-scale loss of privacy information packages due to damage to the data storage device 2.

[0025] Embodiment three: On the basis of the second embodiment, the storage bracket 3 includes a top plate 31 and a bottom plate 32, and a side plate 33 is vertically arranged at the edge of the area between the top plate 31 and the bottom plate 32 to realize the connection between the top plate 31 and the bottom plate 32; A partition plate 34 is evenly and horizontally arranged between the top plate 31 and the bottom plate 32, and the area between the top plate 31 and the bottom plate 32 is evenly divided into a plurality of storage cavities 35. The data storage 2 is fixed inside the storage cavity 35, and connection grooves 36 are arranged on both sides of the storage cavity 35. The data wire 21 passes through the connection grooves 36 to realize the connection between the background recognition module and the data storage 2. The rotary cutter 361 is arranged on the inner wall of the connection groove 36 , and is connected to a rotating device on the inner wall of the connection groove 36 . The rotating device is a micro motor, which drives the rotary cutter 361 to rotate vertically after being started, so as to cut the data wire 21 .

[0026] Specific workflow: Based on the specific workflow in Example 2, the top plate 31 of the storage bracket 3 is made of a high-temperature and impact-resistant material, and can be made of a high-strength metal alloy or marble. In this way, after the storage bracket 3 moves down to the inside of the transfer slot 4, the top plate 31 replaces the closing plate 41 to close the top of the transfer slot 4, providing physical protection for the data storage device 2, and also resisting the explosion impact and flame penetration that may act downward from the upper side when a fire or explosion accident occurs; Furthermore, the internal area of ​​the storage bracket 3 is separated by a partition 34, and the side plate 33 realizes the lateral closure of the storage cavity 35, but retains the connection grooves 36 on both sides. On the one hand, it is convenient for the data wire 21 that undertakes functions such as power supply and data transmission inside the data processing server 1 to pass through and connect with the internal data storage 2 to ensure its normal operation; on the other hand, the connection grooves 36 opened opposite the side plates 33 on both sides are at the same vertical height, so the flow direction of the cooling air generated by the internal cooling device of the data processing server 1 is parallel to the connection line of the connection grooves 36 on both sides. In this way, after the cooling device is started, the cooling air flows from the connection groove 36 on one side to contact the surface of the data storage 2 and then flows from the connection groove 36 on the other side, taking away the heat generated during the operation of the data storage 2, and ensuring the smooth storage process of the privacy data packet; When it is necessary to physically block the electrical connection between the data storage device 2 and the outside world, the rotating cutter 361 inside the connection slot 36 is started. The rotating cutter 361 is made of ceramic material, and the movement trajectory of the rotating process coincides with the position of the data conductor 21. Therefore, as the rotating device is started, the rotating cutter 361 is driven to rotate and cut off the data conductor 21 inside the connection slot 36, thereby achieving physical cutting of the data storage device 2. At the same time, it is also convenient for the subsequent storage bracket 3 to drive the overall movement of the data storage device 2, so as to avoid the problem of damage caused by pulling due to the stable connection of the data conductor 21.

[0027] Embodiment 4: On the basis of the third embodiment, a temperature control cavity 42 is provided at the positions on both sides of the transfer slot 4 on the floor of the machine room, and the temperature control cavity 42 is filled with a cooling medium; The partition 34 extends through the side plate 33 and protrudes outwards near both sides of the connecting groove 36. The inner wall of the storage cavity 35 is provided with a blocking plate 43 at a position corresponding to the partition 34. The blocking plate 43 is horizontally slidably connected to the side wall of the transfer groove 4; one end of the blocking plate 43 extends into the temperature control cavity 42 and is connected to the propulsion device inside the temperature control cavity 42, and the other end is embedded in the transfer groove 4; the propulsion device here is a horizontally arranged electric telescopic device, and a vertical linkage plate is arranged at the output end, and the blocking plates 43 are all connected to the linkage plates to achieve synchronous movement; and the linkage plate is made of metal heat-conducting material, and the bottom of the linkage plate is immersed in the cooling medium to achieve cooling inside the transfer groove 4; Specific workflow: Based on the specific workflow in the second embodiment, the temperature control chambers 42 on both sides of the transfer slot 4 are filled with cooling medium. The cooling medium here can be a common type of cooling liquid used in existing electronic equipment, such as cooling water, and the temperature control chamber 42 is made of metal on one side close to the transfer slot 4, which is convenient for absorbing the heat that penetrates into the transfer slot 4 when a fire occurs, thereby better protecting the internal data storage device 2 and ensuring data security; Further, when the storage bracket 3 moves down to the inside of the transfer slot 4, and the top plate 31 closes the top opening of the transfer slot 4, the protruding end of the partition 34 located on the storage bracket 3 close to the opening of the connecting slot 36 is at the same vertical height as the corresponding blocking plate 43; then the propulsion device can be started to drive the blocking plate 43 to move laterally in the direction close to the data storage device 2, so that the end of the blocking plate 43 coincides with the protruding end of the partition 34, forming a blocking line, preventing the air exchange between the adjacent storage cavities 35 in the vertical direction along the gap between the end of the partition 34 and the transfer slot 4; In this way, even if the explosion causes the top plate 31 to rupture and the high temperature of the flame to penetrate downward, the combination of the protruding end of the partition 34 and the blocking plate 43 also enables the interior of the transfer groove 4 to realize layered blocking lines along the vertical direction, so that the flame and high temperature are confined to the upper area, thereby protecting the data storage device 2 in the lower area; and the high-value privacy data packet with a higher confidentiality priority level is located on the data storage device 2 near the bottom. In this way, even if part of the data storage device 2 is damaged by the flame and high temperature, the survival probability of the high-value privacy data packet with a higher confidentiality priority level can be increased, thereby reducing the loss caused by data loss.

[0028] Embodiment five: On the basis of the fourth embodiment, the top plate 31 is hollow inside to form an isolation chamber 311, and the top plate 31 is a double-layer structure, the portion located on the upper side of the isolation chamber 311 is an insulation layer 313, which is made of marble, and the portion located on the lower side of the isolation chamber 311 is a fixing layer 314, which is made of metal; a fixing hole 44 is provided on the inner wall of the top of the transfer tank 4, and a fixing pipe 441 is provided inside the fixing hole 44, and the fixing pipe 441 is connected to the output end of the propulsion device, and the fixing pipe 441 extends into the temperature control chamber 42 and contacts with the cooling medium; guide holes 312 are provided at the portions on both sides of the top plate 31 corresponding to the fixing holes 44; the portion of the fixing pipe 441 located at the output end of the propulsion device and in the fixing hole 44 is made of a rigid material, which is convenient for horizontal sliding under the pushing action, and the other portions are made of a hose material; Specific workflow: Based on the specific workflow in Example 4, after the storage bracket 3 moves down, the top plate 31 moves to the top opening of the transfer slot 4, and then the guide holes 312 on both sides of the top plate 31 are aligned with the fixing holes 44 on the top opening of the transfer slot 4. As the propulsion device is started, the fixing pipe 441 is pushed to slide out and embed into the guide hole 312 to fix the top plate 31; in this way, when a fire occurs and the temperature inside the machine room rises, in order to prevent heat from being transferred downward to affect the data storage device 2 inside the transfer slot 4, when the temperature sensor inside the transfer slot 4 detects that the temperature is too high, the micro-water pump device at the end of the fixing pipe 441 can be started to pump the cooling medium inside the temperature control cavity 42 on one side into the temperature control cavity 42 on the other side. In this process, the cooling medium passes through the fixing pipe 441 inside the fixing hole 44 and passes through the communicating guide hole 312 to enter the isolation cavity 311 inside the top plate 31, and then flows out from the guide hole 312 on the other side; The flowing cooling medium can absorb the heat transferred to the isolation chamber 311, thereby preventing the high temperature of the machine room from being transferred downward, and better protecting the data storage device 2 at the lower side; and the top plate 31 is a double-layer structure, and the upper heat insulation layer 313 is made of marble with high strength and poor thermal conductivity, which can prevent the heat inside the machine room from being transferred downward; the part located at the lower side of the isolation chamber 311 is a fixed layer 314, which is made of metal material, and can conduct the heat near the lower side of the transfer slot 4 to the cooling medium, thereby improving the cooling effect on the inside of the transfer slot 4; In addition, a layer of metal mesh is adhered to the lower surface of the insulation layer 313, and the metal mesh is in contact with the cooling medium inside the isolation chamber 311. On the one hand, the insulation layer 313 is evenly cooled from bottom to top. On the other hand, when an explosion occurs inside the machine room and a violent impact is transmitted to the insulation layer 313, the insulation layer 313 can absorb the impact vibration by breaking, thereby weakening the impact vibration transmitted downward. The metal mesh on the lower surface can be adhered to limit the fragments after the insulation layer 313 is broken, so that the fragments remain roughly intact and continue to play a heat-insulating role. At this time, the micro-water pump device at the end of the fixed pipe 441 can be closed to avoid the loss of the cooling medium. If the temperature sensor detects that too much heat is transmitted to the transfer slot 4, resulting in excessive heating, the micro-water pump device can be started again to allow the cooling medium flowing into the isolation chamber 311 to flow out from the broken gap, thereby preventing the high temperature from penetrating downward and ensuring the security of the data storage device 2 inside the transfer slot 4 and the privacy data packet stored therein.

[0029] Embodiment six: On the basis of the fifth embodiment, the partition 34 is hollow inside to form a cooling cavity 341, and a connecting port 342 is provided at a position corresponding to the end of the partition 34 and the end of the blocking plate 43, and a transmission pipe 431 is provided inside the blocking plate 43. The transmission pipe 431 is part of a conventional gas fire extinguishing system in a machine room and is responsible for conveying fire extinguishing gas, such as carbon dioxide gas, and the pipe opening of the blocking plate 43 extends out to the protruding end of the partition 34; Specific workflow: On the basis of the specific workflow in Example 5, after the propulsion device is started, the blocking plate 43 is driven to move laterally close to the protruding end of the partition 34, until the end of the blocking plate 43 contacts the protruding end of the partition 34, and the pipe mouth of the transmission pipe 431 at the end of the blocking plate 43 slides and is embedded in the connecting port 342 set on the protruding end of the partition 34; in this way, after the gas fire extinguishing system is started, the liquid fire extinguishing gas released from the high-pressure container is rapidly vaporized and flows along the pipeline, wherein part of the fire extinguishing gas flows into the transmission pipe 431, and then flows into the cooling cavity 341 inside the partition 34 through the connecting port 342 on one side along the transmission pipe 431, and then flows out of the transfer groove 4 through the connecting port 342 on the other side and the combination of the transmission pipe 431, and is supplemented to the fire extinguishing facilities inside the machine room, participating in the fire control inside the machine room; During this process, the temperature of the initially vaporized fire extinguishing gas decreases, so when passing through the partition 34, it can effectively take away the heat of the data storage 2, effectively prevent the high temperature from penetrating downward, and better protect the safety of the data storage 2 inside the storage cavity 35; And because the part of the transmission tube 431 located inside the temperature control chamber 42 is immersed in the cooling medium, the cooling medium can also be cooled, taking away the heat of the cooling medium, and accelerating the vaporization of the fire extinguishing gas, increasing its volume and diffusion range, and reducing the later vaporization time.

[0030] Embodiment seven: On the basis of the sixth embodiment, a clamping groove 432 is provided at a portion corresponding to the end of the blocking plate 43 and the protruding end of the partition plate 34, and the transmission pipe 431 is located on the inner wall of the clamping groove 432; a closed groove 433 is provided on the inner wall of the clamping groove 432, and the cross section of the closed groove 433 is an arc shape; a closed strip 343 is provided at a portion corresponding to the protruding end of the partition plate 34, and the closed strip 343 is made of elastic material; Specific workflow: On the basis of the specific workflow in Example 6, as the storage bracket 3 moves down to the transfer slot 4, the propulsion device is started to push the connected blocking plate 43 to move closer, and the end of the blocking plate 43 contacts the protruding end of the partition 34, and the protruding end of the partition 34 is embedded in the clamping groove 432 on the end of the blocking plate 43. At the same time, the end of the transmission pipe 431 is embedded in the guide hole 312 at the corresponding position on the protruding end of the partition 34. At the same time, the partition 34 and the blocking plate 43 are stably combined to form a blocking line for the high-temperature heat that penetrates downward; at the same time, the closing strip 343 arranged on the protruding end of the partition 34 is embedded in the closing groove 433 on the inner wall of the clamping groove 432, so that the combination between the protruding end of the partition 34 and the blocking plate 43 is more stable, further preventing heat from penetrating downward from the gap between the protruding end of the partition 34 and the blocking plate 43, thereby protecting the information security of the data storage device 2.

[0031] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A data privacy protection processing method using deep learning, characterized in that: The specific steps of the processing method are: S1: Obtain a target image through an information input client, and analyze the person image in the target image through a person recognition module to identify the person's identity information; S2: Based on the confidential private person identity information pre-input into the privacy database, compare the person information in the target image, and perform desensitization processing on the person image information in the target image that needs to be kept confidential; S3: Based on the background recognition module trained by the deep learning module, the background image part other than the facial image of the person in the target image is cut out by the background recognition module and analyzed separately, and the confidential objects contained in the background image are screened out and the confidential objects are desensitized; S4: For the target image containing confidential person identity information and the target image containing confidential targets in the background image, the client information, the time information of uploading the target image and the location information are integrated to obtain a privacy information package and transmitted to the management end. The privacy information package is backed up to the data storage module in the data processing server (1), and the privacy information package is encrypted and the read permission is set.

2. According to claim 1, a data privacy protection processing method using deep learning is characterized by: The types of confidential targets in the above step S3 include buildings, stars and plants with identification features.

3. The data privacy protection processing method using deep learning according to claim 2 is characterized in that: The training process of the confidential target model in the deep learning module includes: S31: Based on the confidentiality requirement of the shooting address of the target image, collect pictures of the surrounding area of ​​the shooting address, including pictures of surrounding buildings and plants with identification features, and use these pictures as test pictures to form a training set; S32: The images in the training set are transferred to the lightweight deep learning training framework Darknet, and a neural network model capable of screening out identification features in the images is obtained through training; S33: randomly download pictures of buildings and plants of the same type but without identification features from the Internet, and mix these pictures with the pictures in the training set as the test set; S34: input the test set images into the trained neural network model, so that it can screen the parts of the images with recognition features, select the parts of the screened images that belong to the original training set for calculation, and obtain the recognition accuracy of the neural network model; S35: When the recognition accuracy is higher than 95%, the neural network model can be deployed to the data processing server (1) as the core of the deep learning module to participate in the screening of confidential targets in the target image.

4. A data privacy protection processing device using deep learning, applicable to the data privacy protection processing method according to any one of claims 1 to 3, comprising a data processing server (1) and a data storage module, wherein the data storage module comprises a data storage device (2), characterized in that: The data processing server (1) comprises a processing box (11), the processing box (11) is fixed on the floor of a machine room, the data storage device (2) is installed on a storage bracket (3) inside the processing box (11), and is connected to a background recognition module inside the processing box (11) via a data wire (21); A transfer opening (12) is provided at a position corresponding to the storage bracket (3) at the bottom of the processing box (11), and the storage bracket (3) is fixed to the upper surface of a closing plate (41) slidably arranged inside the transfer opening (12); a transfer groove (4) is provided on the floor of the machine room at a position located on the lower side of the processing box (11) directly opposite to the transfer opening (12), and the closing plate (41) at the top of the transfer groove (4) is fixedly connected to the bottom of the storage bracket (3), and the closing plate (41) is connected to a telescopic device arranged at the bottom of the transfer groove (4), so that the data storage device (2) is physically cut off by transferring the data storage device (2) to the transfer groove (4).

5. The data privacy protection processing device using deep learning according to claim 4, characterized in that: The storage bracket (3) comprises a top plate (31) and a bottom plate (32), and a side plate (33) is vertically arranged between the top plate (31) and the bottom plate (32) to realize the connection between the top plate (31) and the bottom plate (32); Partitions (34) are evenly arranged between the top plate (31) and the bottom plate (32) to evenly divide the area between the top plate (31) and the bottom plate (32) into a plurality of storage cavities (35); the data storage device (2) is fixed inside the storage cavity (35); and connection grooves (36) are arranged on both sides of the storage cavity (35); the data wire (21) passes through the connection grooves (36) to realize the connection between the background recognition module and the data storage device (2); The inner wall of the connection groove (36) is provided with a rotary cutter (361), and the rotary cutter (361) is connected to a rotating device on the inner wall of the connection groove (36) to cut off the data wire (21).

6. The data privacy protection processing device using deep learning according to claim 5, characterized in that: The floor of the machine room is provided with a temperature control cavity (42) at the two sides of the storage cavity (35), and the temperature control cavity (42) is filled with a cooling medium; The partition (34) extends through the side plate (33) and protrudes outward near both sides of the connecting groove (36); a blocking plate (43) is provided at a position of the inner wall of the transfer groove (4) corresponding to the partition (34); the blocking plate (43) is slidably connected to the side wall of the transfer groove (4); one end of the blocking plate (43) extends into the temperature control cavity (42) and is connected to the propulsion device inside the temperature control cavity (42); the other end is embedded in the transfer groove (4).

7. The data privacy protection processing device using deep learning according to claim 6, characterized in that: The top plate (31) is hollow inside to form an isolation cavity (311), and the top plate (31) is a double-layer structure, wherein the portion located on the upper side of the isolation cavity (311) is a heat insulation layer (313) made of marble; and the portion located on the lower side of the isolation cavity (311) is a fixing layer (314) made of metal; The top opening side wall of the transfer groove (4) is provided with a fixing hole (44), a fixing pipe (441) is slidably provided inside the fixing hole (44), the fixing pipe (441) is connected to the output end of the propulsion device, the fixing pipe (441) extends into the temperature control chamber (42) and contacts with the cooling medium; guide holes (312) are provided at positions corresponding to the fixing holes (44) on both sides of the top plate (31).

8. The data privacy protection processing device using deep learning according to claim 7, characterized in that: The interior of the partition (34) is hollow to form a cooling chamber (341), and a connecting port (342) is provided at a portion of the end of the partition (34) corresponding to the end of the blocking plate (43). A transmission pipe (431) is provided at the end of the blocking plate (43) located inside the storage chamber (35). The transmission pipe (431) is part of the gas fire extinguishing system and is responsible for conveying the fire extinguishing gas, and the pipe mouth of the blocking plate (43) extends out to face the protruding end of the partition (34).

9. The data privacy protection processing device using deep learning according to claim 8, characterized in that: A clamping groove (432) is provided at a portion of the end of the blocking plate (43) corresponding to the protruding end of the partition plate (34), and the pipe mouth end of the transmission pipe (431) is located on the inner wall of the clamping groove (432); A closed groove (433) is provided on the inner wall of the clamping groove (432), and the cross section of the closed groove (433) is arc-shaped; a closed strip (343) is provided at a corresponding position of the protruding end of the partition plate (34), and the closed strip (343) is made of elastic material.

Images