Fault prediction method for cloud encrypted system

Abstract

The invention relates to the field of data security. The invention relates to a cloud encrypted system fault prediction method, which comprises the following steps that: step 1, a user arranges the frequency r of faults and the corresponding frequency f locally, and encrypts the frequency in interlaced data by ci = HeEnc (fi) by using a homomorphic encryption algorithm; and finally obtaining a batch of data and sending the data to the cloud service. Step 2, the cloud service performs expectation calculation on the received encrypted data (r1, c1), (r2, c2)... (rn, cn); the method comprises the following steps: firstly, calculating an expected value Ec = dx after calculating a total fault frequency; 3, the number t of times of user input represents the probability of t times of faults within the future one hour after the user hopes; the system sends t to a cloud service; after the cloud service receives the data, Poisson distribution calculation is carried out; the cloud service transmits a calculation result back to the user; and 4, the user decrypts the calculation result locally after obtaining the calculation result to obtain the probability of fault occurrence, so that subsequent maintenance work is arranged. The intelligent transformation complexity of an enterprise can be reduced, and the enterprise can conveniently apply intelligent achievements.

Description

Technical Field

[0001] The present invention relates to the field of data security, and in particular to a fault prediction method for cloud data processing. Background Art

[0002] For many small and medium-sized companies in traditional industries, using cloud services to deploy and apply intelligent systems is a relatively good option. This is especially true for small and medium-sized industrial units, and even some large industrial sectors. This is because traditional industrial development has a relatively small presence of IT technology, resulting in a lack of IT talent within industrial units and the inability to independently organize computer room construction and IT infrastructure operations and maintenance. Using cloud services can significantly reduce the complexity of intelligent transformation for enterprises and facilitate their application of intelligent technologies.

[0003] When using cloud services, companies need to send their internal data to the cloud. This data, which includes information about the company's production and operations, as well as key intellectual property, is highly valuable. Sending this data directly to the cloud could potentially damage the company's core assets. Summary of the Invention

[0004] The present invention aims to address the above-mentioned shortcomings of the prior art and provides a method for predicting faults in a cloud-based, densely populated system. This method allows enterprises to securely and reliably transfer sensitive data related to their own operations to the cloud for processing.

[0005] The present invention is implemented as follows: a method for predicting faults in a dense cloud system, comprising the following steps:

[0006] Step 1: Data processing: The user organizes the data locally, sorts the number of failures r and the corresponding frequency f, and uses the homomorphic encryption algorithm to encrypt the frequency in each row of data c i =HeEnc(f i ); The final user gets a batch of data (r1, c1), (r2, c2) ... (r n , c n ) and then send this batch of data to the cloud service;

[0007] Step 2: Expected calculation: The cloud service receives the encrypted data (r1, c1), (r2, c2) ... (r n , c n ) to perform expected calculation; first calculate the total number of failures Afterwards, calculate the expected value E c =d\x;

[0008] Step 3: Result prediction: The user enters the number t, which represents the probability of t failures occurring in the next hour. The system sends t to the cloud service. After receiving the data, the cloud service performs Poisson distribution calculations, which are calculated as follows:

[0009]

[0010] In the expression The calculation is completed by performing t cycles of homomorphic multiplication, where t! is the plaintext calculation. The above formula is naturally valid under the fractional operation rules. The cloud service returns the calculation result to the user.

[0011] In step 4, the user obtains the calculation results and decrypts them locally to obtain the probability of failure, thereby arranging subsequent maintenance work.

[0012] In the above-mentioned method for predicting faults of dense systems on the cloud, in the expression of step three, It is the ciphertext power of the plaintext, which is a typical non-polynomial calculation method and is calculated using the polynomial approximation method.

[0013] The cloud-based dense system fault prediction method, the polynomial approximation method calculation includes using the Taylor series expansion method for calculation.

[0014] The method for predicting faults in a dense cloud system, wherein the Taylor series expansion method is:

[0015]

[0016] That is, take the sum of the first six items.

[0017] When performing correlation testing, users of the present invention only need to encrypt the original data locally to generate secret data, then send the secret data to the cloud platform. The cloud platform will perform calculations on the secret data and return the results of the secret calculations to the user. Finally, the user decrypts the calculation results to obtain the final plaintext correlation analysis results. During the entire business process, the cloud platform cannot access the user's original data, allowing users to use the cloud platform for data analysis with confidence, enjoy low-cost computing power, and not worry about data leakage. Based on this technology, when performing correlation testing, users only need to encrypt the original data locally to generate secret data, then send the secret data to the cloud platform. The cloud platform will perform calculations on the secret data and return the results of the secret calculations to the user. Finally, the user decrypts the calculation results to obtain the final plaintext correlation analysis results. During the entire business process, the cloud platform cannot access the user's original data, allowing users to use the cloud platform for data analysis with confidence, enjoy low-cost computing power, and not worry about data leakage. DETAILED DESCRIPTION

[0018] The present invention is implemented as follows: the cloud-based dense system fault prediction method first requires encoding the data:

[0019] (1) Message encoding: For the input message m, amplify it by 10 k times, so that f = m·10 k is an integer, and k is such that f = m·10 k The smallest integer that holds true. Output the encoded result o=(f,10 k ).

[0020] (2) Ciphertext encoding: The encryption algorithm input is o=(h, 10 k ), encrypt h and get c f , output ciphertext c=(c f , 10 k ), where d is equal to 1 by default.

[0021] (3) Decryption and decoding: The decryption algorithm input is c = (c f , c d ), c f and c d Decrypt and get f, calculate and output d In the decryption algorithm, use c d The reason is that in the calculation rules described later, there is a method to perform ciphertext division operation, so that the second element of the ciphertext c becomes ciphertext, which obviously also needs to be decrypted in the decryption algorithm.

[0022] Operation rules for coded data:

[0023] Using the fraction operation rules, c f As a molecule, 10 k Considered as the denominator.

[0024] Program Implementation Process

[0025] The cloud-based dense fault prediction method uses the Poisson distribution to predict the probability of failure within a certain period of time. This method mainly involves three steps: data processing, expectation calculation, and result prediction.

[0026] Data processing:

[0027] The user organizes data locally, sorting out the number of failures and the corresponding frequency. For example, the following data represents the failure status of the equipment within x hours.

[0028] Number of faults (r) Frequency (f) 0 12 1 5 2 6 … … .

[0029] The user uses the homomorphic encryption algorithm locally to encrypt the frequency in each row of data c i =HeEnc(f i ). The final user gets a batch of data (r1, c1), (r2, c2)... (r n , c n ). The user sends this batch of data to the cloud service.

[0030] Expectation Calculation

[0031] If the data on both sides of a binary operator is ciphertext, it means that the operation is a homomorphic operation. In the following text, unless otherwise specified, this expression is still used.

[0032] In this link, the cloud service receives the encrypted data (r1, c1), (r2, c2) ... (r n , c n ) to perform expected calculation. First calculate the total number of failures Afterwards, calculate the expected value E c =d\x.

[0033] Result prediction

[0034] The user enters the number of times t, which represents the probability of r failures occurring within the next hour. The system sends t to the cloud service. After receiving the data, the cloud service performs a Poisson distribution calculation using the following formula:

[0035]

[0036] In this formula is the power of the plaintext of the ciphertext, which can be calculated by homomorphic multiplication for t cycles, and t! is the plaintext calculation, which is naturally established under the fractional operation rules. However It is the ciphertext power of the plaintext, which is a typical non-polynomial calculation method and requires the use of polynomial approximation method for calculation.

[0037] In this scheme, the classic Taylor expansion method is used for approximate calculation, and the calculation method is:

[0038]

[0039] After the calculation is completed, the cloud service sends the encrypted results to the user, who decrypts them locally to obtain the probability of failure and arrange subsequent maintenance work.

Claims

1. A method for predicting faults in a dense cloud system, characterized by: It includes the following steps: Step 1: Data processing: The user organizes the data locally, sorts the number of failures r and the corresponding frequency f, and uses the homomorphic encryption algorithm to encrypt the frequency in each row of data c i =HeEnc(f i ); The final user gets a batch of data (r1, c1), (r2, c2) ... (r n , c n ) and then send this batch of data to the cloud service; Step 2: Expected calculation; The cloud service receives the encrypted data (r1, c1), (r2, c2) ... (r n , c n ) to perform expected calculation; first calculate the total number of failures Afterwards, calculate the expected value E c =d\x; Step 3: Result prediction: The user enters the number t, which represents the probability of t failures occurring in the next hour. The system sends t to the cloud service. After receiving the data, the cloud service performs Poisson distribution calculations, which are calculated as follows: In the expression The calculation is completed by performing t cycles of homomorphic multiplication, where t! is the plaintext calculation. The above formula is naturally valid under the fractional operation rules. The cloud service returns the calculation result to the user. In step 4, the user obtains the calculation results and decrypts them locally to obtain the probability of failure, thereby arranging subsequent maintenance work.

2. The method for predicting faults in a dense cloud system according to claim 1, wherein: In the expression of step 3, It is the ciphertext power of the plaintext, which is a typical non-polynomial calculation method and is calculated using the polynomial approximation method.

3. The method for predicting faults in a dense cloud system according to claim 2, wherein: The polynomial approximation method includes calculating by using a Taylor series expansion method.

4. The method for predicting faults in a dense cloud system according to claim 3, wherein: The method of Taylor series expansion is: That is, take the sum of the first six items.

Images