Neural network construction method under homomorphic encryption and image processing method and system
A homomorphic encryption and neural network technology, applied in the field of data processing, can solve the problems that two kinds of data cannot be processed at the same time, and the convolutional neural network is rarely involved.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0047] A schematic flow chart of a method for constructing a convolutional neural network under homomorphic encryption provided by the present invention is as follows: figure 1 shown, including:
[0048] Step 1: Obtain the pre-trained convolutional neural network model and output the network parameters of the convolutional neural network model;
[0049] Step 2: Convert the convolutional neural network model under homomorphic encryption according to the network parameters to obtain a convolutional neural network capable of identifying multiple types of data;
[0050] Wherein, the convolutional neural network model is obtained by training a plurality of image data of the user layer and corresponding classification results.
[0051] Specifically, including:
[0052] Step 1: Obtain the pre-trained convolutional neural network model and output the network parameters of the convolutional neural network model;
[0053]Train multiple image data of the user layer and their correspon...
Embodiment 2
[0071] The present invention takes the LeNet-5 convolutional neural network as an example, and uses the method described in the present invention to convert:
[0072] 1. First convert the input. For each pixel value a in the input image, there are:
[0073]
[0074] 2. The convolutional layer is:
[0075]
[0076]
[0077] D. 2 =N
[0078] Among them, W1*H1 is the size of the input image feature, N is the number of convolution kernels, F is the size of the convolution kernel, S is the step size, and P is the convolution operation of zero padding.
[0079] For the weight W in the convolutional layer, perform the following conversion:
[0080]
[0081] Among them, τ is the precision parameter, and ω is the original network parameter.
[0082] 3. The classification function softmax in the convolutional network is as follows:
[0083] softmax(b)=normalize(exp(b))
[0084] Convert it using Taylor series as follows:
[0085]
[0086] Among them, b is the input ...
Embodiment 3
[0103] Based on the same inventive concept, the present invention also provides an image data processing method. Since the processing method is based on a neural network construction method under homomorphic encryption, repeated descriptions will not be repeated here.
[0104] The method, such as image 3 shown, including:
[0105] Obtain image data to be processed;
[0106] Processing the image data to be processed by using a convolutional neural network capable of identifying multiple types of data to obtain a classification result of the image data to be processed;
[0107] Wherein, the convolutional neural network capable of identifying multiple types of data is pre-built using a neural network construction method under homomorphic encryption.
[0108] Specifically include:
[0109] Preferably, the processing of the image data to be processed by using a convolutional neural network capable of identifying multiple types of data to obtain a classification result of the im...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More