Cloud computing information security improvement method based on interval training

文档序号：1831443 发布日期：2021-11-12 浏览：11次中文

阅读说明：本技术 一种基于间隔式训练的云计算信息安全的改善方法 (Cloud computing information security improvement method based on interval training ) 是由耿涛于 2021-08-16 设计创作，主要内容包括：本申请揭示了一种基于间隔式训练的云计算信息安全的改善方法,进行第一次训练处理,得到第一参考模型；得到第一验证结果；若验证通过的数量小于通过数量阈值,则将第一图像分类模型的参数进行修改；第一用户端进行参数修改,得到第二图像分类模型；构成第一图像集；输入第二图像分类模型中,得到第一图像分类结果集；将第一图像集和第一图像分类结果集发送给云计算平台；对第一图像集进行处理,得到第二图像分类结果集；若分类结果集相同,则给予暂时访问权限；进行第二次训练处理,得到第二参考模型；得到第二验证结果；若小于通过数量阈值,则进行参数修改；将参数发送给第一用户端,从而动态改善云计算信息安全。(The application discloses a cloud computing information security improving method based on interval training, which comprises the steps of carrying out first training processing to obtain a first reference model; obtaining a first verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; constructing a first set of images; inputting the image into a second image classification model to obtain a first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; processing the first image set to obtain a second image classification result set; if the classification result sets are the same, giving temporary access authority; performing second training treatment to obtain a second reference model; obtaining a second verification result; if the number is smaller than the passing number threshold value, modifying the parameters; and sending the parameters to the first user terminal, thereby dynamically improving the cloud computing information security.)

1. A cloud computing information security improvement method based on interval training is characterized by comprising the following steps:

s1, the cloud computing platform conducts first training processing on a preset neural network model in a first time window to obtain a first reference model; training by using a first amount of first training data in the first training process, wherein the first training data is composed of a first training picture and a first training label which is manually marked on the first training picture, and the corresponding relation between the first training picture and the first training label is wrong;

s2, inputting a preset second amount of verification data into the first reference model by the cloud computing platform for processing, so as to obtain a second amount of first verification results correspondingly output by the first reference model; the second amount of verification data is composed of a first verification picture and a first verification label which manually marks the first verification picture, and the corresponding relation between the first verification picture and the first verification label is correct; the second number is greater than the first number;

s3, the cloud computing platform judges whether the number of passed verification in the second number of first verification results is smaller than a preset passing number threshold value;

s4, if the number of passed verification in the second number of first verification results is smaller than a preset pass number threshold, extracting parameters of each network layer in the first reference model, and modifying the parameters of each network layer of a preset first image classification model into the parameters of each network layer in the first reference model;

s5, the cloud computing platform sends the parameters of each network layer in the first reference model to the first user terminal;

s6, the first user terminal modifies the parameters of each network layer of the preset neural network model into the parameters of each network layer in the first reference model, and therefore a second image classification model is obtained;

s7, the first user end correspondingly acquires a third number of images and a fourth number of images through hand drawing on the touch screen and an image sensor acquisition mode, and the third number of images and the fourth number of images form a first image set;

s8, the first user end inputs the first image set into the second image classification model for processing, so as to obtain a first image classification result set output by the second image classification model;

s9, the first user terminal sends the first image set and the first image classification result set to the cloud computing platform;

s10, inputting the first image set into the first image classification model by the cloud computing platform for processing, so as to obtain a second image classification result set output by the first image classification model;

s11, the cloud computing platform judges whether the first image classification result set is the same as the second image classification result set;

s12, if the first image classification result set is the same as the second image classification result set, giving temporary access authority to the first user side;

s13, performing second training processing on the first reference model by the cloud computing platform in a second time window to obtain a second reference model; a fifth amount of second training data is adopted for training in the second training process, the second training data is composed of a second training picture and a second training label which is used for manually marking the second training picture, and the corresponding relation between the second training picture and the second training label is wrong; the left boundary of the second time window is to the right of the right boundary of the first time window;

s14, inputting a preset sixth amount of verification data into the second reference model by the cloud computing platform for processing, so as to obtain a sixth amount of second verification results correspondingly output by the second reference model; the sixth amount of verification data is composed of a second verification picture and a second verification label which manually labels the second verification picture, and the corresponding relation between the second verification picture and the second verification label is correct; the sixth number is greater than the fifth number;

s15, the cloud computing platform judges whether the verification passing number in the sixth number of second verification results is smaller than a preset passing number threshold value;

s16, if the number of passed verifications in the sixth number of second verification results is less than a preset number-of-passed threshold, extracting parameters of each network layer in the second reference model, and modifying the parameters of each network layer of the first image classification model into the parameters of each network layer in the second reference model;

and S17, the cloud computing platform sends the parameters of each network layer in the second reference model to the first user side, so that the improvement of cloud computing information security is realized.

2. The interval training-based improvement method for the cloud computing information security according to claim 1, wherein the cloud computing platform performs a first training process on a preset neural network model within a first time window to obtain a first reference model; before step S1, in which training is performed by using a first amount of first training data in the first training process, the first training data is composed of a first training picture and a first training label manually labeled to the first training picture, and a correspondence between the first training picture and the first training label is incorrect, the method includes:

s01, the cloud computing platform obtains a sample set corresponding to the trained picture classification model according to the corresponding relation between the preset classification model and the sample set; the recognition accuracy of the trained image classification model is greater than a preset accuracy threshold;

s02, the cloud computing platform extracts a first amount of sample data and a second amount of sample data from the sample set;

s03, modifying the manually labeled label corresponding to each sample picture in the first amount of sample data by the cloud computing platform, so as to obtain a first amount of first training data;

s04, the cloud computing platform records the second amount of sample data as second amount of verification data;

and S05, generating a training and verifying instruction by the cloud computing platform, wherein the training and verifying instruction is used for instructing the preset neural network model to be trained and verified for the first time.

3. The method for improving cloud computing information security based on interval training as claimed in claim 1, wherein the step S7 of the first user end obtaining a third number of images and a fourth number of images by hand-drawing on a touch screen and collecting with an image sensor, and forming the third number of images and the fourth number of images into a first image set comprises:

s701, generating a third number of hand-drawn images by the first user side in a hand-drawing mode on the touch screen;

s702, the first user side carries out image acquisition processing through a preset image sensor to generate a fourth number of environment images;

s703, the first user side generates a seventh number of proliferation images by using the third number of hand-drawn images and the fourth number of environment images according to a preset image proliferation method;

s704, the first user end enables the third quantity of hand-drawn images, the fourth quantity of environment images and the seventh quantity of proliferation images to form a first image set.

4. The interval training-based improvement method for cloud computing information security according to claim 3, wherein the first user end generates a seventh number of proliferation images S703 by using the third number of hand-drawn images and the fourth number of environment images according to a preset image proliferation method, including:

s7031, the first user performs image separation processing on the third number of hand-drawn images to obtain a plurality of first separated portions;

s7032, the first user performs image separation processing on the fourth number of environment images to obtain a plurality of second separated portions;

s7033, the first user side performs a combination process on the plurality of first separated portions and the plurality of second separated portions to obtain a seventh number of proliferation images; wherein a proliferation image is formed by combining at least one first separated part and at least one second separated part.

5. The method for improving cloud computing information security based on interval training as claimed in claim 1, wherein the step S17 of the cloud computing platform sending the parameters of each network layer in the second reference model to the first user end, so as to improve cloud computing information security includes:

s171, the cloud computing platform judges whether the first user side is in communication connection with the cloud computing platform;

s172, if the first user side is not in communication connection with the cloud computing platform, cancelling the temporary access authority of the first user side;

s173, the cloud computing platform judges whether the first user terminal initiates a new communication connection request;

s174, if the first user initiates a new communication connection request, requiring the first user to send a second image set and a classification result set corresponding to the second image set to the cloud computing platform.

6. A cloud computing information security improvement system based on interval training is characterized by comprising:

the first training unit is used for indicating the cloud computing platform to perform first training processing on a preset neural network model in a first time window so as to obtain a first reference model; training by using a first amount of first training data in the first training process, wherein the first training data is composed of a first training picture and a first training label which is manually marked on the first training picture, and the corresponding relation between the first training picture and the first training label is wrong;

the first verification result acquisition unit is used for instructing the cloud computing platform to input a preset second amount of verification data into the first reference model for processing so as to obtain a second amount of first verification results correspondingly output by the first reference model; the second amount of verification data is composed of a first verification picture and a first verification label which manually marks the first verification picture, and the corresponding relation between the first verification picture and the first verification label is correct; the second number is greater than the first number;

the first verification result judging unit is used for indicating the cloud computing platform to judge whether the verification passing number in the second number of first verification results is smaller than a preset passing number threshold value or not;

a first parameter modification unit, configured to instruct, if the number of verification passes in the second number of first verification results is smaller than a preset pass number threshold, to extract parameters of each network layer in the first reference model, and modify the parameters of each network layer of the preset first image classification model into the parameters of each network layer in the first reference model;

the first reference model parameter sending unit is used for indicating the cloud computing platform to send the parameters of each network layer in the first reference model to the first user terminal;

a second image classification model obtaining unit, configured to instruct the first user end to modify parameters of each network layer of a preset neural network model into parameters of each network layer in the first reference model, so as to obtain a second image classification model;

the first image set acquisition unit is used for instructing a first user end to correspondingly acquire a third number of images and a fourth number of images in a manner of hand-drawing on the touch screen and image sensor acquisition, and the third number of images and the fourth number of images form a first image set;

a first image classification result set obtaining unit, configured to instruct a first user to input the first image set into the second image classification model for processing, so as to obtain a first image classification result set output by the second image classification model;

a first image classification result set sending unit, configured to instruct a first user end to send the first image set and the first image classification result set to the cloud computing platform;

a second image classification result set obtaining unit, configured to instruct a cloud computing platform to input the first image set into the first image classification model for processing, so as to obtain a second image classification result set output by the first image classification model;

an image classification result set judgment unit, configured to instruct a cloud computing platform to judge whether the first image classification result set is the same as the second image classification result set;

a temporary access authority giving unit, configured to instruct to give a temporary access authority to the first user side if the first image classification result set is the same as the second image classification result set;

the second training unit is used for indicating the cloud computing platform to perform second training processing on the first reference model in a second time window to obtain a second reference model; a fifth amount of second training data is adopted for training in the second training process, the second training data is composed of a second training picture and a second training label which is used for manually marking the second training picture, and the corresponding relation between the second training picture and the second training label is wrong; the left boundary of the second time window is to the right of the right boundary of the first time window;

the second verification result acquisition unit is used for instructing the cloud computing platform to input a preset sixth amount of verification data into the second reference model for processing, so as to obtain a sixth amount of second verification results correspondingly output by the second reference model; the sixth amount of verification data is composed of a second verification picture and a second verification label which manually labels the second verification picture, and the corresponding relation between the second verification picture and the second verification label is correct; the sixth number is greater than the fifth number;

a second verification result judgment unit, configured to instruct the cloud computing platform to judge whether the number of verification passes in the sixth number of second verification results is smaller than a preset pass number threshold;

a second parameter modification unit, configured to instruct, if the verification number in the sixth number of second verification results is smaller than a preset passing number threshold, to extract parameters of each network layer in the second reference model, and modify the parameters of each network layer of the first image classification model into the parameters of each network layer in the second reference model;

and the second reference model parameter sending unit is used for indicating the cloud computing platform to send the parameters of each network layer in the second reference model to the first user side, so that the improvement of the cloud computing information security is realized.

7. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 5.

Technical Field

The application relates to the field of computers, in particular to a cloud computing information security improvement method based on interval training.

Background

One problem facing cloud computing is that information security is difficult to guarantee. The existing cloud computing platform determines whether a user side has access authority or not through a static account password, the mode is difficult to guarantee information safety, and once the account password of a user is leaked, all information of the user is stolen. Therefore, the existing cloud computing platform lacks a dynamic cloud computing information security improvement scheme.

Disclosure of Invention

The application provides a cloud computing information security improvement method based on interval training, which comprises the following steps:

s3, the cloud computing platform judges whether the number of passed verification in the second number of first verification results is smaller than a preset passing number threshold value;

s5, the cloud computing platform sends the parameters of each network layer in the first reference model to the first user terminal;

s9, the first user terminal sends the first image set and the first image classification result set to the cloud computing platform;

s11, the cloud computing platform judges whether the first image classification result set is the same as the second image classification result set;

s12, if the first image classification result set is the same as the second image classification result set, giving temporary access authority to the first user side;

s15, the cloud computing platform judges whether the verification passing number in the sixth number of second verification results is smaller than a preset passing number threshold value;

Further, the cloud computing platform performs first training processing on a preset neural network model in a first time window to obtain a first reference model; before step S1, in which training is performed by using a first amount of first training data in the first training process, the first training data is composed of a first training picture and a first training label manually labeled to the first training picture, and a correspondence between the first training picture and the first training label is incorrect, the method includes:

s02, the cloud computing platform extracts a first amount of sample data and a second amount of sample data from the sample set;

s04, the cloud computing platform records the second amount of sample data as second amount of verification data;

Further, the step S7 of the first user end obtaining a third number of images and a fourth number of images by hand-drawing on the touch screen and capturing with the image sensor, and forming the third number of images and the fourth number of images into the first image set includes:

s701, generating a third number of hand-drawn images by the first user side in a hand-drawing mode on the touch screen;

s702, the first user side carries out image acquisition processing through a preset image sensor to generate a fourth number of environment images;

s704, the first user end enables the third quantity of hand-drawn images, the fourth quantity of environment images and the seventh quantity of proliferation images to form a first image set.

Further, the generating, by the first user end, a seventh number of proliferation images S703 according to a preset image proliferation method by using the third number of hand-drawn images and the fourth number of environment images includes:

s7031, the first user performs image separation processing on the third number of hand-drawn images to obtain a plurality of first separated portions;

s7032, the first user performs image separation processing on the fourth number of environment images to obtain a plurality of second separated portions;

Further, after the step S17 of sending, by the cloud computing platform, the parameters of each network layer in the second reference model to the first user side, so as to improve the cloud computing information security, the method includes:

s171, the cloud computing platform judges whether the first user side is in communication connection with the cloud computing platform;

s172, if the first user side is not in communication connection with the cloud computing platform, cancelling the temporary access authority of the first user side;

s173, the cloud computing platform judges whether the first user terminal initiates a new communication connection request;

The application provides an improvement system of cloud computing information security based on interval formula training, includes:

The present application provides a computer device comprising a memory storing a computer program and a processor implementing the steps of any of the above methods when the processor executes the computer program.

The present application provides a computer-readable storage medium having stored thereon a computer program which, when being executed by a processor, carries out the steps of the method of any of the above.

According to the method, the system, the computer equipment and the storage medium for improving the cloud computing information security based on interval training, the first training process is carried out to obtain a first reference model; obtaining a second number of first verification results; if the number of passed verification in the first verification result is smaller than the threshold value of the number of passed verification, modifying parameters of each network layer of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; forming a first image set by hand-drawing on the touch screen and acquiring by an image sensor; inputting the image into a second image classification model to obtain an output first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; the first image classification model processes the first image set to obtain a second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority; carrying out second training treatment to obtain a second reference model; obtaining a second verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters; and sending the parameters of each network layer to the first user terminal. Thereby dynamically improving cloud computing information security.

Drawings

Fig. 1-2 are schematic flow charts illustrating a method for improving cloud computing information security based on interval training according to an embodiment of the present application;

fig. 3 is a block diagram illustrating a structure of a computer device according to an embodiment of the present application.

The implementation, functional features and advantages of the objectives of the present application will be further explained with reference to the accompanying drawings.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

Referring to fig. 1-2, an embodiment of the present application provides a method for improving cloud computing information security based on interval training, including the following steps:

s3, the cloud computing platform judges whether the number of passed verification in the second number of first verification results is smaller than a preset passing number threshold value;

s5, the cloud computing platform sends the parameters of each network layer in the first reference model to the first user terminal;

s9, the first user terminal sends the first image set and the first image classification result set to the cloud computing platform;

s11, the cloud computing platform judges whether the first image classification result set is the same as the second image classification result set;

s12, if the first image classification result set is the same as the second image classification result set, giving temporary access authority to the first user side;

s15, the cloud computing platform judges whether the verification passing number in the sixth number of second verification results is smaller than a preset passing number threshold value;

The application discloses a cloud computing information security improvement method based on interval training, wherein the interval training refers to that the first training is performed in a first time window, the second training is performed in a second time window, and the like, and each time window is discontinuous. In addition, the dynamic cloud computing information security improvement can be realized, that is, the temporary access authority of the first user side is valid in a period of time, and the first image set and the first image classification result set for confirming the authority are only valid in a single time.

As described in the above steps S1-S5, the cloud computing platform performs a first training process on the preset neural network model within a first time window to obtain a first reference model; training by using a first amount of first training data in the first training process, wherein the first training data is composed of a first training picture and a first training label which is manually marked on the first training picture, and the corresponding relation between the first training picture and the first training label is wrong; the cloud computing platform inputs a preset second amount of verification data into the first reference model for processing, so that a second amount of first verification results correspondingly output by the first reference model are obtained; the second amount of verification data is composed of a first verification picture and a first verification label which manually marks the first verification picture, and the corresponding relation between the first verification picture and the first verification label is correct; the second number is greater than the first number; the cloud computing platform judges whether the number of passed verification in the second number of first verification results is smaller than a preset number-of-passed threshold value; if the number of passed verification in the second number of first verification results is smaller than a preset pass number threshold, extracting parameters of each network layer in the first reference model, and modifying the parameters of each network layer of a preset first image classification model into the parameters of each network layer in the first reference model; and the cloud computing platform sends the parameters of each network layer in the first reference model to the first user terminal.

The first time window has no excessive limitation, and the training does not need to be carried out for too long time each time, and only the parameters of each layer of network layer of the model need to be changed, so the training speed is high. The training process of the present application is different from a training process of a general neural network model, and is different in that a first amount of first training data is used for training in the first training process, the first training data is composed of a first training picture and a first training label manually labeling the first training picture, and the correspondence between the first training picture and the first training label is wrong; the second amount of verification data is composed of a first verification picture and a first verification label which manually marks the first verification picture, and the corresponding relation between the first verification picture and the first verification label is correct; the second number is greater than the first number. In contrast, the accuracy is pursued by common model training, but the inaccuracy is pursued by the model training of the application, the mechanism is different inaccurate models, the output results are different, and the output results can be ensured to be the same only by the inaccurate models with the same parameters; in addition, in the ordinary model training, the training data and the verification data of the model training must be correct and homologous, while the training data of the application should be wrong and the verification data should be correct; finally, in the ordinary model training, the training data is much larger than the verification data, and in contrast to the present application, the number of training data is smaller than the number of verification data, i.e. the second number is larger than the first number. Through the intermittent training, the chaotic reference model can be generated intermittently, and by using the parameters of the chaotic reference model, other terminals can be prevented from obtaining the same reference model, so that the accuracy of permission giving is ensured, and the information safety is improved. The cloud computing platform judges whether the number of passed verification in the second number of first verification results is smaller than a preset number threshold value of passed verification, so as to determine whether the accuracy of picture classification is low, if the number of passed verification in the second number of first verification results is smaller than the preset number threshold value of passed verification, the accuracy is low, and the accuracy meets the requirements, so that the parameters of each network layer in the first reference model are extracted, and the parameters of each network layer in the preset first image classification model are modified into the parameters of each network layer in the first reference model. And then the parameters of each network layer in the first reference model are sent to the first user terminal. At this time, only the cloud platform and the first user terminal have parameters of each network layer in the first reference model. The neural network model may be any feasible model, such as a convolutional neural network model, and so on.

s02, the cloud computing platform extracts a first amount of sample data and a second amount of sample data from the sample set;

s04, the cloud computing platform records the second amount of sample data as second amount of verification data;

Thereby obtaining a first amount of first training data and a second amount of validation data. The method and the device have the advantage that training data and verification data are obtained as soon as possible by utilizing the existing image classification model after training is completed. Because the recognition accuracy of the trained image classification model is greater than the preset accuracy threshold, the corresponding sample set can be directly used as verification data, and then the manually labeled label corresponding to each sample image in the first amount of sample data is modified, so that the first amount of first training data is obtained. Thereby obtaining a first amount of first training data and a second amount of validation data that are immediately available. Further, the subsequent interval training may adopt the data in the sample set at this time, or may regenerate new training data and verification data.

As described in the above steps S6-S9, the first user modifies the parameters of each network layer of the preset neural network model into the parameters of each network layer in the first reference model, so as to obtain a second image classification model; the method comprises the steps that a first user end correspondingly obtains a third number of images and a fourth number of images through hand drawing on a touch screen and an image sensor acquisition mode, and the third number of images and the fourth number of images form a first image set; the first user end inputs the first image set into the second image classification model for processing, so that a first image classification result set output by the second image classification model is obtained; and the first user terminal sends the first image set and the first image classification result set to the cloud computing platform.

The neural network model on the first user side is the same as the neural network model and the first image classification model on the cloud computing platform in type and has the same network layer, and parameters of each network layer of the preset neural network model are modified into parameters of each network layer in the first reference model, so that the second image classification model is the same as the first image classification model. And correspondingly acquiring a third number of images and a fourth number of images by hand-drawing on the touch screen and acquiring the images by an image sensor, and forming a first image set by the third number of images and the fourth number of images. The method has the advantages that the images in the first image set are not limited, the images comprise hand-drawn images, and the corresponding images can be generated when a user scrawls on the touch screen; and an image sensor is adopted to collect images, so that the types of the images in the first image set are richer, and the subsequent image classification is more facilitated. And then inputting the image into a second image classification model for processing to obtain a first image classification result set. At this time, the first image classification result set can only be obtained by using the second image classification model or the first image classification model as input theoretically. The hand-drawn image input method has the advantages that the hand-drawn mode on the touch screen is adopted, so that the experience of a user can be improved, and the hand-drawn image is input in a game-like mode.

s701, generating a third number of hand-drawn images by the first user side in a hand-drawing mode on the touch screen;

s702, the first user side carries out image acquisition processing through a preset image sensor to generate a fourth number of environment images;

s704, the first user end enables the third quantity of hand-drawn images, the fourth quantity of environment images and the seventh quantity of proliferation images to form a first image set.

According to the method and the device, the third number of images and the fourth number of images are correspondingly acquired in a manner of hand-drawing and image sensor acquisition on the touch screen, the number of the acquired images is small, and the final permission giving judgment is not facilitated. Therefore, the present application performs image multiplication processing using an image multiplication method to increase the number of images in the first image set. In this way, the accuracy of the authority determination of the overall scheme is improved. Among them, since the present application does not excessively restrict the image, it is particularly suitable for image multiplication. The image multiplication method can adopt any feasible method, for example, an image splicing method is adopted to generate a new image.

s7031, the first user performs image separation processing on the third number of hand-drawn images to obtain a plurality of first separated portions;

s7032, the first user performs image separation processing on the fourth number of environment images to obtain a plurality of second separated portions;

The image multiplication is carried out by adopting a mode of image separation and recombination. Since the hand-drawn image and the environment image have different characteristics, the proliferation image obtained by combining at least one first isolated portion and at least one second isolated portion can be used to obtain a new unusual image. The image separation processing method can adopt any feasible method, such as separation by region or main background separation.

As described in the above steps S10-S14, the cloud computing platform inputs the first image set into the first image classification model for processing, so as to obtain a second image classification result set output by the first image classification model; the cloud computing platform judges whether the first image classification result set is the same as the second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority to the first user terminal; the cloud computing platform carries out second training processing on the first reference model in a second time window to obtain a second reference model; a fifth amount of second training data is adopted for training in the second training process, the second training data is composed of a second training picture and a second training label which is used for manually marking the second training picture, and the corresponding relation between the second training picture and the second training label is wrong; the left boundary of the second time window is to the right of the right boundary of the first time window; the cloud computing platform inputs a preset sixth amount of verification data into the second reference model for processing, so that a sixth amount of second verification results correspondingly output by the second reference model are obtained; the sixth amount of verification data is composed of a second verification picture and a second verification label which manually labels the second verification picture, and the corresponding relation between the second verification picture and the second verification label is correct; the sixth number is greater than the fifth number.

Since the first image classification model and the second image classification model have the same parameters, the second image classification result set output by the first image classification model should theoretically be the same as the first image classification result set output by the second image classification model. Therefore, whether the first image classification result set is the same as the second image classification result set is judged, and if the first image classification result set is the same as the second image classification result set, it indicates that the first image set is matched with the first image classification result set (that is, the first image classification result set is obtained by classifying the first image set by the second image classification model on the first user side); on the contrary, if the first image classification result set is different from the second image classification result set, it indicates that the information-stealing party intends to steal data, because different inaccurate image classification models are different for the output of the same input, and only the same inaccurate image classification model is the same for the output of the same input. Judging whether the first image classification result set is the same as the second image classification result set can be realized by adopting any feasible method, for example, comparing results in the two classification result sets in sequence, and if each member is the same, judging that the first image classification result set is the same as the second image classification result set.

Further, the step of the cloud computing platform determining whether the first image classification result set is the same as the second image classification result set includes:

respectively sorting the first image classification result set and the second image classification result set by adopting a first sorting method, so as to obtain a first sorting list and a second sorting list;

comparing the member at the first position in the first sorted list with the member at the first position in the second sorted list to obtain a first comparison result;

judging whether the first comparison results are the same;

if the first comparison result is the same, respectively sorting the first image classification result set and the second image classification result set by adopting a second sorting method, thereby obtaining a third sorting list and a fourth sorting list;

comparing the member at the second position in the third sorted list with the member at the second position in the fourth sorted list to obtain a second comparison result;

judging whether the second comparison results are the same;

and if the second comparison result is the same, judging that the first image classification result set is the same as the second image classification result set.

Therefore, the judgment of whether the two result sets are the same or not is realized in a twice sorting mode without comparing all members in sequence.

And if the first image classification result set is the same as the second image classification result set, giving temporary access authority to the first user terminal. The temporary access right may be set in any feasible form, for example, the temporary access right is given in a manner of having the access right within a certain time, or the temporary access right is defined by a trigger condition, for example, the cloud computing platform is disconnected from the communication connection with the first user terminal. The cloud computing platform carries out second training processing on the first reference model in a second time window to obtain a second reference model; a fifth amount of second training data is adopted for training in the second training process, the second training data is composed of a second training picture and a second training label which is used for manually marking the second training picture, and the corresponding relation between the second training picture and the second training label is wrong; the left boundary of the second time window is to the right of the right boundary of the first time window. The second time window of the second training process is not continuous nor crossed with the first time window, so that the interval type second period training is started. And then, carrying out verification processing to obtain a sixth number of second verification results correspondingly output by the second reference model, wherein the sixth number of second verification results is used as a basis for judging whether the model obtained by the second training is available.

As described in the above steps S15-S17, the cloud computing platform determines whether the number of passed verifications in the sixth number of second verification results is less than a preset threshold number of passed verifications; if the verification passing number in the sixth number of second verification results is smaller than a preset passing number threshold, extracting parameters of each network layer in the second reference model, and modifying the parameters of each network layer of the first image classification model into the parameters of each network layer in the second reference model; and the cloud computing platform sends the parameters of each network layer in the second reference model to the first user terminal, so that the improvement of the cloud computing information security is realized.

If the verification passing number in the sixth number of second verification results is smaller than a preset passing number threshold, indicating that the parameters of the second reference model are available, and modifying the parameters of each network layer of the first image classification model into the parameters of each network layer in the second reference model; and sending the parameters of each network layer in the second reference model to the first user terminal. And after the first user terminal fails in the temporary access authority, the first user terminal can use the parameters of each network layer of the first image classification model to perform another authority application. And the failure time point of the temporary access right is after the step that the cloud computing platform sends the parameters of each network layer in the second reference model to the first user side, so that the improvement of the cloud computing information security is realized.

s171, the cloud computing platform judges whether the first user side is in communication connection with the cloud computing platform;

s172, if the first user side is not in communication connection with the cloud computing platform, cancelling the temporary access authority of the first user side;

s173, the cloud computing platform judges whether the first user terminal initiates a new communication connection request;

Thus, real-time intermittent authority is given to ensure information safety. If the first user side is not in communication connection with the cloud computing platform, namely communication is conducted, in order to prevent information stealing party fake as the first user side from obtaining access authority, the temporary access authority of the first user side is cancelled, and when a new communication connection request is initiated, the first user side is required to send a second image set and a classification result set corresponding to the second image set to the cloud computing platform, so that the inaccurate image classification model is used again to judge whether authority should be given.

According to the cloud computing information safety improvement method based on interval training, first training processing is carried out to obtain a first reference model; obtaining a second number of first verification results; if the number of passed verification in the first verification result is smaller than the threshold value of the number of passed verification, modifying parameters of each network layer of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; forming a first image set by hand-drawing on the touch screen and acquiring by an image sensor; inputting the image into a second image classification model to obtain an output first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; the first image classification model processes the first image set to obtain a second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority; carrying out second training treatment to obtain a second reference model; obtaining a second verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters; and sending the parameters of each network layer to the first user terminal.

The embodiment of the application provides a cloud computing information security's improvement system based on interval formula training, includes:

The operations respectively executed by the above units correspond to the steps of the cloud computing information security improvement method based on interval training in the foregoing embodiment one to one, and are not described herein again.

The cloud computing information safety improving system based on interval training carries out first training processing to obtain a first reference model; obtaining a second number of first verification results; if the number of passed verification in the first verification result is smaller than the threshold value of the number of passed verification, modifying parameters of each network layer of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; forming a first image set by hand-drawing on the touch screen and acquiring by an image sensor; inputting the image into a second image classification model to obtain an output first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; the first image classification model processes the first image set to obtain a second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority; carrying out second training treatment to obtain a second reference model; obtaining a second verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters; and sending the parameters of each network layer to the first user terminal.

Referring to fig. 3, an embodiment of the present invention further provides a computer device, where the computer device may be a server, and an internal structure of the computer device may be as shown in the figure. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the computer designed processor is used to provide computational and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used for storing data used by the cloud computing information security improvement method based on interval training. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an improved method for cloud computing information security based on interval training.

The processor executes the method for improving the cloud computing information security based on interval training, wherein the steps included in the method correspond to the steps of executing the method for improving the cloud computing information security based on interval training in the foregoing embodiment one to one, and are not described herein again.

It will be understood by those skilled in the art that the structures shown in the drawings are only block diagrams of some of the structures associated with the embodiments of the present application and do not constitute a limitation on the computer apparatus to which the embodiments of the present application may be applied.

The computer equipment carries out first training processing to obtain a first reference model; obtaining a second number of first verification results; if the number of passed verification in the first verification result is smaller than the threshold value of the number of passed verification, modifying parameters of each network layer of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; forming a first image set by hand-drawing on the touch screen and acquiring by an image sensor; inputting the image into a second image classification model to obtain an output first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; the first image classification model processes the first image set to obtain a second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority; carrying out second training treatment to obtain a second reference model; obtaining a second verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters; and sending the parameters of each network layer to the first user terminal.

An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored thereon, and when the computer program is executed by a processor, the method for improving cloud computing information security based on interval training is implemented, where steps included in the method correspond to steps of the method for improving cloud computing information security based on interval training in the foregoing embodiment one to one, and are not described herein again.

The computer-readable storage medium of the present application, performs a first training process to obtain a first reference model; obtaining a second number of first verification results; if the number of passed verification in the first verification result is smaller than the threshold value of the number of passed verification, modifying parameters of each network layer of the first image classification model; the first user terminal modifies the parameters to obtain a second image classification model; forming a first image set by hand-drawing on the touch screen and acquiring by an image sensor; inputting the image into a second image classification model to obtain an output first image classification result set; sending the first image set and the first image classification result set to a cloud computing platform; the first image classification model processes the first image set to obtain a second image classification result set; if the first image classification result set is the same as the second image classification result set, giving temporary access authority; carrying out second training treatment to obtain a second reference model; obtaining a second verification result; if the number of the verification passes is smaller than the passing number threshold value, modifying the parameters; and sending the parameters of each network layer to the first user terminal.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware associated with a computer program or instructions, the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can include the processes of the embodiments of the methods described above when executed. Any reference to memory, storage, database, or other medium provided herein and used in the examples may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), double-rate SDRAM (SSRSDRAM), Enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and bus dynamic RAM (RDRAM).

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, system, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, system, article, or method. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, system, article, or method that includes the element.

The above description is only a preferred embodiment of the present application, and not intended to limit the scope of the present application, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the specification and the drawings of the present application, or which are directly or indirectly applied to other related technical fields, are also included in the scope of the present application.

22页详细技术资料下载

Cloud computing information security improvement method based on interval training

相关技术

网友询问留言