Data transmission method and electronic equipment

Abstract

The invention provides a data transmission method and electronic equipment, which are used for carrying out privacy protection on data uploaded to a server and avoiding the problem of privacy disclosure of a large cloud model. The method comprises the following steps: acquiring first target data, and determining privacy information contained in the first target data; performing decryption processing on privacy information in the first target data to obtain second target data; and sending the second target data to the server, so that the server processes the second target data and returns a processing result.

Description

Technical Field

[0001] The present disclosure relates to the technical field of data processing, and in particular to a data transmission method and an electronic device. Background Art

[0002] In recent years, with the development of the Internet, more and more important information is transmitted in the form of online processing, and information security has received great attention and importance. Information storage security loopholes will lead to information leakage, causing great losses to individuals or companies.

[0003] With the development of large models, more and more image generation and text generation tasks are performed using large models. Although large models have superior performance, they are very large in size and have high disk requirements for local deployment. In particular, the reasoning of large models requires hardware support, which is simply not possible on ordinary users' terminal devices. Large models deployed in the cloud are prone to user privacy leaks. Summary of the invention

[0004] The present invention provides a data transmission method and an electronic device for protecting the privacy of data uploaded to a server, thereby avoiding the problem of privacy leakage of a large cloud model.

[0005] In a first aspect, an embodiment of the present disclosure provides a data transmission method, the method comprising:

[0006] Acquire first target data, and determine privacy information contained in the first target data;

[0007] Decrypting the privacy information in the first target data to obtain second target data;

[0008] The second target data is sent to a server, so that the server processes the second target data and returns a processing result.

[0009] In a second aspect, an embodiment of the present disclosure provides an electronic device, including a processor and a memory, wherein the memory is used to store a program executable by the processor, and the processor is used to read the program in the memory and perform the following steps:

[0010] Acquire first target data, and determine privacy information contained in the first target data;

[0011] Decrypting the privacy information in the first target data to obtain second target data;

[0012] The second target data is sent to a server, so that the server processes the second target data and returns a processing result.

[0013] In a third aspect, an embodiment of the present disclosure further provides a data transmission device, the device comprising:

[0014] A privacy determination module, configured to obtain first target data and determine privacy information contained in the first target data;

[0015] A data determination module, configured to decrypt the privacy information in the first target data to obtain second target data;

[0016] The data sending module is used to send the second target data to the server, so that the server processes the second target data and returns the processing result.

[0017] In a fourth aspect, an embodiment of the present disclosure further provides a computer storage medium on which a computer program is stored, and when the program is executed by a processor, it is used to implement the steps of any one of the methods described in the first aspect above.

[0018] In a fifth aspect, the present disclosure provides a computer program product, comprising: a computer program code, and when the computer program code is executed on a computer, the computer executes any one of the methods described in the first aspect.

[0019] These and other aspects of the present disclosure will become more apparent from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 A flowchart of a data transmission method according to an embodiment of the present disclosure;

[0022] Figure 2 A schematic diagram of a human-computer interaction interface provided by an embodiment of the present disclosure;

[0023] Figure 3 A protection transmission flow chart of text privacy information provided by an embodiment of the present disclosure;

[0024] Figure 4 A flowchart of text privacy protection provided by an embodiment of the present disclosure;

[0025] Figure 5 A flowchart of text privacy protection provided by an embodiment of the present disclosure;

[0026] Figure 6 A protection transmission flow chart of image privacy information provided by an embodiment of the present disclosure;

[0027] Figure 7 A flowchart of image privacy protection provided by an embodiment of the present disclosure;

[0028] Figure 8 A schematic diagram of an electronic device provided by an embodiment of the present disclosure;

[0029] Fig. 9 A schematic diagram of a data transmission device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the present disclosure will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present disclosure.

[0031] In the embodiments of the present disclosure, the term "and / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent three situations: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0032] The application scenarios described in the embodiments of the present disclosure are intended to more clearly illustrate the technical solutions of the embodiments of the present disclosure, and do not constitute a limitation on the technical solutions provided by the embodiments of the present disclosure. It is known to those skilled in the art that with the emergence of new application scenarios, the technical solutions provided by the embodiments of the present disclosure are also applicable to similar technical problems. In the description of the present disclosure, unless otherwise specified, the meaning of "multiple" is two or more.

[0033] Before introducing the data transmission method provided by the embodiment of the present disclosure, for ease of understanding, the technical background of the embodiment of the present disclosure is first introduced in detail below.

[0034] Large Model (also called Foundation Model) refers to a machine learning model with large-scale parameters and complex computational structures. These models are usually built with deep neural networks and have billions or even hundreds of billions of parameters. Large models are designed to improve the expressiveness and predictive performance of the model and to handle more complex tasks and data. Large models are widely used in various fields, including natural language processing, computer vision, speech recognition, and recommendation systems. Large models learn complex patterns and features by training massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unseen data.

[0035] There are many large models, such as GPT (Generative Pre-trained Transformer) and ChatGPT, both of which are language models based on the Transformer architecture. Transformer is a neural network architecture based on the attention mechanism, but GPT and ChatGPT differ in design and application. The GPT model is designed to generate natural language text and handle various natural language processing tasks, such as text generation, translation, and summarization. It is usually used in one-way generation, that is, to generate coherent output based on a given text. ChatGPT focuses on dialogue and interactive dialogue. It has been specifically trained to better handle multi-round dialogue and context understanding. ChatGPT is designed to provide a smooth, coherent and interesting dialogue experience to respond to user input and generate appropriate replies. Another example is the Large Language Model, which is usually a natural language processing model with large-scale parameters and computing power, such as OpenAl's GPT-3 model. These models can be trained with a large amount of data and parameters to generate human-like text or answer natural language questions. Large language models are widely used in fields such as natural language processing, text generation, and intelligent dialogue.

[0036] With the development of large models, more and more image generation and text generation tasks are performed using large models. Although large models have superior performance, they are very large in size and have high disk requirements for local deployment. In particular, the reasoning of large models requires hardware support, which is simply not possible on ordinary users' terminal devices. Large models deployed in the cloud are prone to user privacy leaks.

[0037] In order to solve the above technical problems, this embodiment provides a data transmission method. The core idea is to locally protect the privacy of the data of the large model uploaded to the server, replace the privacy information in the data locally and then upload it to the server, and adopt local privacy protection and cloud-based large model reasoning to solve the problem of privacy leakage of large cloud models. The security of data transmission can be protected without deploying the large model locally, thereby improving the security of information transmission without affecting the reasoning performance of the large model.

[0038] like Figure 1 As shown, the data transmission method provided in this embodiment is applied to a local device, and the specific implementation process of the method is as follows:

[0039] Step 100: Acquire first target data and determine the privacy information contained in the first target data;

[0040] In implementation, the first target data may be obtained through a variety of methods such as user input, uploading, downloading, etc. This embodiment does not impose too many restrictions on the method of obtaining the first target data.

[0041] Optionally, the first target data in this embodiment includes but is not limited to text (documents), images (videos), etc. After obtaining the first target data, it is necessary to analyze the privacy information contained in the first target data. The privacy information in this embodiment is used to represent information that the user does not want others to know, or sensitive information and other information that cannot be made public.

[0042] In some embodiments, when the data type of the first target data is text, the privacy information contained in the first target data may be determined in the following manner:

[0043] The semantics of the first target data are analyzed by using a keyword extraction algorithm to identify keywords contained in the first target data; and the privacy information contained in the first target data is determined based on the identified keywords.

[0044] Optionally, the keyword extraction algorithm in this embodiment includes but is not limited to:

[0045] (1) TF-IDF (Term Frequency-Inverse Document Frequency);

[0046] TF-IDF is a statistical method used to evaluate the importance of a word to a document set or one of the documents in a corpus. It is achieved by comparing the term frequency (TF) and the inverse document frequency (IDF). It helps to identify words that appear frequently in only a few documents and is considered to have good discriminability for documents.

[0047] (2) TextRank (graph ranking);

[0048] TextRank is a graph-based ranking algorithm used for text processing, especially in keyword extraction and abstract generation. It builds a graph model of words in the text and then uses the PageRank algorithm to identify important nodes (words) in the graph to extract keywords.

[0049] (3) LDA (Latent Dirichlet Allocation);

[0050] LDA is a topic model algorithm that can be used to identify topics in large-scale text data. By treating each document in a document collection as a topic distribution and each topic as a keyword distribution, LDA can discover hidden topics in a document collection and extract keywords based on them.

[0051] (4) Word2Vec (word to sequence);

[0052] Word2Vec is a method that uses neural networks to express words as vectors. In this way, the similarity between words can be calculated. Although Word2Vec itself is not directly used for keyword extraction, it can assist in identifying semantically related words and thus assist in keyword extraction.

[0053] (5) Methods based on deep learning;

[0054] With the development of deep learning technology, some deep learning-based models such as CNN (Convolutional Neural Networks), RNN (Recurrent Neural Network), BERT (Bidirectional Encoder Representations from Transformers), KeyBert, etc. are used for feature extraction of text data and further for keyword extraction. These models can learn the deep semantic information of text, thereby improving the accuracy and effect of keyword extraction.

[0055] In implementation, this embodiment can use the KeyBert algorithm combined with semantic coding to embed the text and then identify the key information in the text. KeyBERT is a keyword and key phrase extraction tool based on the pre-trained language model BERT. It uses an unsupervised pre-trained model to automatically extract key information from the text to better understand and process text data.

[0056] The basic principle of KeyBERT can be divided into two steps: pre-training and keyword extraction. In the pre-training stage, KeyBERT uses the pre-trained BERT model as its base model. BERT will randomly mask some words in the input text and predict these masked words through other words in the context. Such training can help BERT learn the contextual relationship between words, thereby improving the representation ability of the model. Afterwards, BERT will input two sentences and predict whether the two sentences are continuous. This task can help BERT learn the relationship between sentences and better understand the semantics of sentences in subsequent tasks. Through these two pre-training tasks, BERT can learn rich text representations, including word-level and sentence-level representations. After the pre-training is completed, KeyBERT uses the pre-trained BERT model to extract keywords and key phrases. The process of extracting keywords can be divided into the following steps:

[0057] Step 1) text preprocessing;

[0058] First, KeyBERT preprocesses the input text. It converts the text to lowercase and removes some common stop words and punctuation marks to reduce the impact of noise on keyword extraction.

[0059] Step 2) text encoding;

[0060] KeyBERT uses the pre-trained BERT model to encode the pre-processed text. The BERT model converts each word into a corresponding word vector and encodes the entire sentence into a fixed-length vector representation. This vector representation retains the contextual information between words and can better express the semantics of the sentence.

[0061] Step 3) keyword extraction;

[0062] After encoding, KeyBERT uses the MMR (Maximal Marginal Relevance) algorithm to extract keywords and key phrases. The MMR algorithm selects the most relevant keywords by balancing the relevance and diversity of keywords. Specifically, the MMR algorithm first calculates the similarity score of each word with the text. Then, an initial keyword is selected as a seed, and the relevance score of the seed with other words is calculated. Next, a word with high relevance to the seed and low relevance to other selected keywords is selected as the next keyword. This process is repeated until the desired number of keywords is reached. The number of output keywords is controlled by setting the top_n parameter. Through this algorithm, KeyBERT can select keywords with high relevance and diversity to better represent the important information of the input text.

[0063] Optionally, this embodiment may also utilize synonym generation technology to generate synonyms for privacy vocabulary, and use cosine similarity to find the generated synonyms and words or phrases in the first target data whose similarity is greater than a threshold. Finally, the words or phrases whose similarity is greater than the threshold may be identified as privacy information.

[0064] In some embodiments, when the data type of the first target data is text, this embodiment can also determine the privacy information contained in the first target data through the following steps:

[0065] According to the privacy information stored in the local database, target data having a similarity greater than a threshold with the stored privacy information is determined from the first target data; and the target data having a similarity greater than the threshold is determined as the privacy information included in the first target data.

[0066] In implementation, the local database in this embodiment stores privacy information and replacement information corresponding to the privacy information. The privacy information stored in the local database is used to query whether the first target data contains the privacy information in the local database. If so, the privacy information in the first target data is replaced by the replacement information corresponding to the found privacy information. For example, the local database stores "privacy A" and replacement information "data A" corresponding to "privacy A". When the first target data contains "privacy A", "privacy A" in the first target data is replaced by "data A".

[0067] In some embodiments, when the data type of the first target data is an image, the private information and the second target data are determined in the following manner:

[0068] In response to a target position input by a user, a target image at the target position in the first target data and a background image excluding the target position are extracted using a cutout network; the target image is determined as the privacy information contained in the first target data, and the background image is determined as the second target data.

[0069] Optionally, the target location input by the user in this embodiment includes but is not limited to:

[0070] (1) Text information input by the user;

[0071] In implementation, semantic analysis is performed on text information input by the user to obtain the target position of the private information in the first target data, thereby extracting the target image at the target position.

[0072] In implementation, the target position in the first target data is determined according to the text information, and then the target image at the target position is extracted using the cutout network. For example, if the user inputs a face, the face position in the first target data is determined, and the face image in the first target data is extracted using the cutout network; if the user inputs a hand, the hand position in the first target data is determined, and the hand image in the first target data is extracted using the cutout network.

[0073] (2) user-specified target location;

[0074] In implementation, the first target data is displayed through a display interface, and the user can determine the target position by circling, marking, etc. on the displayed first target data, so as to extract the target image at the target position in the first target data using the cutout network. For example, the user circles a tattoo on the arm on the first target data, takes the circled tattoo position as the target position, and uses the cutout network to extract the tattoo image in the first target data; for another example, the user marks text on the first target data, takes the marked text position as the target position, and uses the cutout network to extract the text image in the first target data.

[0075] In some embodiments, when the data type of the first target data is an image, this embodiment uses a cutout network to extract a target image at the target position in the first target data and a background image other than the target position through the following steps:

[0076] The first target data is processed by using a static image cutout algorithm to obtain a classified image; the first target data and the classified image are input into the cutout network to output the target image and the background image.

[0077] Optionally, before performing image processing on the first target data using a static image cutout algorithm, denoising and normalization processing may be performed on the first target data. The specific implementation steps are as follows:

[0078] First, the first target data is denoised by the Perona-Malik equation to filter out the noise in the first target data and improve the accuracy of image processing; then, the denoised first target data is normalized to obtain the normalized first target data to adapt to the data format requirements of the input end of the cutout network. Among them, the Perona-Malik equation is an anisotropic nonlinear diffusion equation, which realizes anisotropic diffusion by introducing a nonlinear diffusion function, aiming to smooth the regional part of the image while protecting the edge information. This algorithm changes the speed of change of the pixel value by adjusting the gradient information of the pixel point, so that the edge features can be better retained when processing the image, and the regional part of the image can be smoothed at the same time. The Perona-Malik algorithm is particularly suitable for image processing, and can effectively remove noise while maintaining the edges and details of the image.

[0079] During implementation, a static image cutout algorithm is used to perform image processing on the normalized first target data to obtain a classified image. The first target data and the classified image are input into a cutout network together, and a target image and a background image are output. The background image is determined as the second target data, and the target image is determined as privacy information.

[0080] Optionally, the static image cutout algorithm includes but is not limited to: Trimap, Strokes, etc. When processing an image, the static image cutout algorithm usually requires the user to provide some interactive information, such as Trimap and Strokes, to help the algorithm more accurately segment the foreground and background in the image. Trimap requires the user to mark the areas in the image that definitely belong to the foreground, the areas that definitely belong to the background, and the areas that may belong to the foreground or the background (i.e., uncertain areas). This marking method provides the algorithm with a rough segmentation reference, thereby improving the accuracy and efficiency of the cutout.

[0081] In the implementation, the normalized first target data is processed using the Trimap algorithm to obtain a classified image, namely the Trimap image. The Trimap image is a rough classification of a given image. The Trimap image is divided into three parts: foreground (F), background (B), and unknown area (U). The foreground is the cutout area (indicating the area where the privacy information is located, namely the target image), the unknown area is the area where the cutout area and the background are integrated, such as the edge, hair, and other boundary areas between the person and the background, and the background is the area that does not overlap with the cutout area. The cutout network uses the normalized first target data and the Trimap image as input.

[0082] The cutout network in this embodiment can be a trained OpenVINO (Open Visual Inference and Neural Network Optimization) lightweight model (such as MobileNetV3, FBNetV3, etc.), and the first target data and the Trimap image are input into the cutout network to obtain the weight alpha of the cutout. Based on the weight alpha, the background and the first target image are fused, and finally the cutout result is output, that is, the target image and the background image are output.

[0083] Optionally, this embodiment may also perform the following processing on the cutout network:

[0084] 1) Use depthwise separable convolution to extract features of the input image and reduce the computational cost of traditional convolution operations. Among them, depthwise separable convolution reduces the number of parameters required for convolution calculation by splitting the correlation between spatial dimension and channel (depth) dimension, and improves the efficiency of convolution kernel parameters. Depthwise separable convolution calculation is divided into two parts: first, spatial convolution is performed on the channel (depth) separately, and the output is spliced; then, channel convolution is performed using a unit convolution kernel to obtain a feature map. This technology is mainly divided into two processes, namely, channel-wise convolution (Depthwise Convolution) and pointwise convolution (Pointwise Convolution). In channel-wise convolution, one convolution kernel is responsible for one channel, and one channel is convolved by only one convolution kernel. The number of channels of the feature map generated by this process is exactly the same as the number of channels of the input. Pointwise convolution performs a dot product operation on the result of depthwise convolution to further process the feature map.

[0085] 2) Use h-wish as the activation function. The h-wish formula is as follows:

[0086]

[0087] In formula (1), x represents the output of the previous network layer in the cutout network, ReLU represents the ReLU activation function, and h-wish[x] represents the activation function value.

[0088] The cutout network in this embodiment can maximize the accuracy of cutout and increase the cutout speed (to meet the real-time cutout requirements) through the lightweight network + model optimization method.

[0089] Step 101: decrypt the private information in the first target data to obtain second target data;

[0090] Optionally, the data types in this embodiment include but are not limited to text, images, etc. Optionally, the decryption processing in this embodiment is used to remove and / or replace the private information.

[0091] In some embodiments, when the data type of the first target data is text, the second target data is determined in the following manner:

[0092] Generate replacement information corresponding to the private information, wherein the replacement information corresponding to the private information has the same data type as the private information; replace the private information included in the first target data with the replacement information to obtain second target data.

[0093] In implementation, in order to ensure that the second target data after replacing the private information can still be processed by the server's large model, it is necessary to ensure that the data types of the private information and the replacement information are the same, otherwise the large model will not be able to process the second target data.

[0094] In some embodiments, when the data type of the first target data is text, the replacement information corresponding to the private information is generated by any of the following methods:

[0095] Method 1: Generate replacement information corresponding to the private information by using synonym generation technology;

[0096] In practice, synonym generation technology can be used to generate multiple synonyms (ie, replacement information) of private information, and the multiple synonyms of the private information are provided to the user so that the user can select replacement information corresponding to the private information from the multiple synonyms.

[0097] Optionally, the synonym generation technology in this embodiment includes but is not limited to: Wordnet, which is an English vocabulary database that contains a large number of synonyms, near-synonyms and antonyms; word2vec, which is a technology based on deep learning. It learns the semantic relationship between words by training a large amount of text data and can generate words with similar meanings to given words; NLPCDA (NLP Chinese Data Augmentation), etc.

[0098] Optionally, multiple synonyms of the private information can be generated by using the synonym generation technology in the NLPCDA open source tool. NLPCDA is an open source tool for Chinese data enhancement, which aims to enhance the generalization ability of the NLP (Natural Language Processing) model by generating sentences similar to the original text.

[0099] Method 2: Use random word generation technology to generate replacement information corresponding to the private information.

[0100] During implementation, corresponding replacement information may also be randomly generated based on the private information.

[0101] Method three: using a synonym generation technology to generate first replacement information corresponding to the private information, and using a random word generation technology to generate second replacement information corresponding to the private information.

[0102] It should be noted that the “first” and “second” in this embodiment do not indicate an order, but are only used to distinguish different replacement information.

[0103] During implementation, first replacement information and second replacement information corresponding to the private information may be generated and provided to the user for selection, thereby preventing the generated synonyms from directly replacing the private words and causing possible privacy leakage.

[0104] In some embodiments, when the data type of the first target data is text; generating replacement information corresponding to the private information includes:

[0105] According to the correspondence between the privacy information and the replacement information stored in the local database, the replacement information corresponding to the privacy information contained in the first target data is searched from the local database; and the replacement information found is determined as the replacement information corresponding to the privacy information.

[0106] Optionally, the private information includes but is not limited to text information such as words and phrases, and the replacement information includes but is not limited to text information such as words and phrases.

[0107] During implementation, the privacy word in the first target data may be searched in the local database first. When the privacy word is found, the privacy word may be directly replaced with a replacement word corresponding to the privacy word stored in the local database.

[0108] In some embodiments, when the data type of the first target data is an image, the second target data is determined in the following manner:

[0109] Determine the target location of the private information in the first target data; remove the private information contained in the first target data to obtain second target data.

[0110] It should be noted that, when decrypting image data, this embodiment only determines the target position of the private information in the image, thereby removing the image data at the target position to obtain the second target data, and uses the server to replace the image data at the target position with a random image, and merges it with the second target data, that is, fills the second target data with an image at the target position.

[0111] In some embodiments, the second target data is determined by:

[0112] In response to a target position input by a user, a target image at the target position in the first target data and a background image excluding the target position are extracted using a cutout network; and the background image is determined as the second target data.

[0113] In some embodiments, when the data type of the first target data is an image, the server may generate replacement information corresponding to the private information in the following manner:

[0114] According to the characteristic attributes of the private information, a random image that meets the characteristic attributes is randomly generated, wherein different characteristic attributes are used to distinguish different contents included in the first target data; and the random image is determined as replacement information corresponding to the private information.

[0115] Optionally, the characteristic attributes in this embodiment include, but are not limited to, characteristics of different parts of the human body, such as facial features, hand features, etc.; skin features, such as tattoos, etc.; non-human features, such as text, scenery, background, etc. in the image. This embodiment does not impose too many restrictions on the specific information contained in the characteristic attributes, and the corresponding characteristic attributes can be determined according to the privacy information specified by the user.

[0116] In implementation, when the private information determined by the user is a face in the first target data (image), a face image is randomly generated, and the face image is used as replacement information to replace the face in the first target data. When the private information determined by the user is a tattoo on a human body in the first target data, a tattoo image is randomly generated, and the tattoo image is used as replacement information to replace the tattoo on the human body in the first target data.

[0117] Step 102: Send the second target data to a server, so that the server processes the second target data and returns a processing result.

[0118] In some embodiments, when the data type of the first target data is text, the privacy information contained in the first target data is replaced with the replacement information to obtain the second target data. The second target data is sent to the server for processing, and the processing result is returned.

[0119] In some embodiments, when the data type of the first target data is an image, the second target data and the target position may also be sent to a server, so that the server can fill the second target data at the target position with an image using a randomly generated random image.

[0120] In implementation, when the data type of the first target data is an image, and the second target data is a background image in the first target data excluding the privacy information, this embodiment can also send the target positions of the background image and the privacy information in the first target data to the server, so that the server can use a randomly generated random image to fill the target position of the background image.

[0121] During implementation, when the data type of the first target data is an image, the data (background image) in the first target data excluding the privacy information and the target position input by the user are determined as the second target data, and the second target data is sent to the cloud server, which is used for the cloud server to add a random image to the target position of the background image, generate the first target data without the privacy information, perform other image processing, and return the processing result.

[0122] In some embodiments, when the data type of the first target data is text, the following steps may also be performed:

[0123] Storing the corresponding relationship between the private information and the replacement information in a local database;

[0124] Receive the processing results sent by the server;

[0125] According to the correspondence between the private information and the replacement information stored in the local database, the replacement information in the processing result is replaced with the private information.

[0126] In some embodiments, in order to process the results returned by the processor more quickly, the present embodiment can also use the local database to restore the privacy information, that is, store the correspondence between the privacy information and the replacement information in the local database. When the server returns the processing result, the privacy information in the local database can be used to restore the privacy information replaced in the processing result, that is, replace the replacement information of the processing result with the privacy information.

[0127] In some embodiments, when the data type of the first target data is an image, any of the following methods may also be performed:

[0128] Method a) storing the target position and the first target data; receiving a processing result sent by a server, determining a target image according to the stored first target data and the target position, and replacing the image at the target position in the processing result with the target image;

[0129] During implementation, the target position in the first target data can be stored. After receiving the processing result returned by the server, the position of the private information in the processing result can be determined according to the target position, and the private information in the first target data can be used to restore the private information in the processing result.

[0130] Mode b) storing the target position and the target image; receiving a processing result sent by a server, and replacing the image at the target position in the processing result with the target image.

[0131] In implementation, the target position and target image (i.e., privacy information) in the first target data can be stored, so that the image at the target position can be directly replaced with the target image using the stored target image, that is, the privacy information in the processing result can be restored using the stored target image.

[0132] It should be noted that, since the privacy information of the first target data is replaced, when the first target data is text, it can be directly replaced to obtain the second target data; when the first target data is an image, in order to solve the problem of mismatch or even void in edge information, the replacement information and the first target data except the privacy information can be sent to the cloud server for image fusion processing. Similarly, the privacy information of the returned processing result can be restored, and the restored privacy information can be processed using image fusion technology.

[0133] In some embodiments, when the data type of the first target data is an image, after the information at the target position in the processing result is replaced with the target image, the image fusion processing may be performed in the following manner:

[0134] The first target data is processed by using a static image cutout algorithm to obtain a classified image; and the classified image and the processing result are fused by using an image fusion technology to obtain a fused processing result.

[0135] In the implementation, the Trimap algorithm is used to perform image processing on the first target data to obtain a classified image, namely, a Trimap image. The Trimap image is a rough classification of a given image. The Trimap image is divided into three parts: foreground (F), background (B), and unknown area (U). In order to optimize the image details and reflect the edge information of the cutout, the Fusion Module is used to perform image fusion to obtain a more accurate alpha image (processing result). The final accurate alpha image (α p ). The formula for obtaining is as follows:

[0136] α p =F+Uα r Formula (2);

[0137] In formula (2), α p represents the fusion processing result, F represents the foreground of the classified image, U represents the unknown area in the classified image, and αr Indicates the processing result.

[0138] Optionally, in order to ensure that the image style of the stored private information is consistent with the image style of the processed result, style migration may be performed on the image style of the restored processed result to ensure that the image style of the private information and other parts in the restored processed result are consistent.

[0139] In some embodiments, this embodiment may also edit, confirm, and perform other operations on the generated replacement information in any one or more of the following ways, as shown below:

[0140] Mode 1) displaying the privacy information contained in the first target data on a human-computer interaction interface, and in response to a user's editing operation on the privacy information, updating the correspondence between the privacy information and the replacement information, and storing the updated information in a database;

[0141] Mode 2) Displaying replacement information corresponding to the private information included in the first target data on a human-computer interaction interface, and in response to a user's editing operation on the replacement information, updating the correspondence between the private information and the replacement information and storing it in a database.

[0142] Optionally, the editing operations in this embodiment include but are not limited to modification, selection, deletion, addition and the like, and this embodiment does not impose too many limitations on this.

[0143] In implementation, when the data type of the first target data is text, the private information and the corresponding replacement information can be displayed on the human-computer interaction interface, and the user can select, confirm, modify, etc. the replacement information of each private information, and can also add new private information and its corresponding replacement information, etc. When the data type of the first target data is an image, the private information (i.e., the target image) and the background image can be displayed on the human-computer interaction interface, and the user can confirm, modify, etc. the target image.

[0144] like Figure 2 As shown, this embodiment provides a human-computer interaction interface, which displays the private information (a1, a2, a3, a4), the replacement information (b1, b2, b3, b4) generated by the synonym generation technology, and the randomly generated replacement information (c1, c2, c3, c4) to the user. The user can select the corresponding replacement information for each private information, edit the generated replacement information, add private information or replacement information by himself, and edit the extracted private information. After the user confirms the replacement information corresponding to each private information, the private information confirmed by the user and the corresponding replacement information can be saved to the local database.

[0145] In some embodiments, after obtaining the second target data, the second target data may be encrypted and sent to the server to further ensure communication security. The data may be transmitted using https (Hypertext Transfer Protocol Secure) to ensure the security of the communication process. Optionally, data encryption methods such as MD5 (Message Digest Algorithm 5) and SHA256 (Secure Hash Algorithm 256) may be used.

[0146] In some embodiments, the server includes but is not limited to a cloud server or a large server, and the server may be one or more servers in the cloud, and a large model is deployed on the server for model reasoning. The large model includes but is not limited to Qwen1.5-72B, chatglm3-6B, SD (stable diffusion) model, etc., wherein Qwen1.5-72B is a large language model that can perform natural language processing, ChatGLM3-6B is based on ChatGLM, ChatGLM is a generative machine learning method for dialogue systems, specifically built for conversation tasks, and can help the system automatically generate targeted responses. ChatGLM uses the GLM (Generalized Linear Model) algorithm, combined with a neural network model and generative text generation technology, and can understand natural language and infer the correct response from it through a large amount of text learning and imitating the response pattern of the interlocutor. The SD model is used for image processing and is an AI painting generation tool. It is based on the technology of the latent diffusion model (LDM). The encoder in the autoencoder (AE) compresses the image into a latent representation space, and then uses the diffusion model to generate the latent representation of the image. Finally, the generated image is obtained through the decoder module of the AE. This embodiment does not impose too many restrictions on the large model, and the specific model can be determined according to actual usage requirements.

[0147] like Figure 3 As shown, this embodiment provides a protection transmission process of text privacy information, which is specifically as follows:

[0148] Step 300: Acquire first target data;

[0149] Step 301: searching the local database for private information in the first target data and replacement information corresponding to the private information, and replacing the private information in the first target data with the replacement information;

[0150] During implementation, based on the privacy information stored in the local database, target data whose similarity with the stored privacy information is greater than a threshold is determined from the first target data, and the target data is determined as privacy data; based on the correspondence between the privacy information and replacement information stored in the local database, replacement information corresponding to the privacy information contained in the first target data is searched from the local database; and the replacement information found is determined as the replacement information corresponding to the privacy information.

[0151] Step 302: using a keyword extraction algorithm to identify private information in text other than the sentences in which private information has been replaced in the first target data;

[0152] In practice, after the private information is replaced in step 301, when the private information is identified using the key detection and extraction algorithm, it is not necessary to identify the sentence in which the private information has been replaced, and only the text content of the private information that has not been replaced needs to be identified.

[0153] Step 303: Generate first replacement information corresponding to the identified private information using a synonym generation technology, and generate second replacement information corresponding to the identified private information using a random word generation technology;

[0154] Step 304: display the private information, the first replacement information and the second replacement information on the human-computer interaction interface, and determine the replacement information corresponding to each private information in the first target data in response to the user's editing operation on the replacement information.

[0155] During implementation, the user can select one of the first replacement information (synonyms) and the second replacement information (random words) as the replacement information for the private information. In addition, the user can customize the replacement information, or add new unrecognized private information, and / or the replacement information corresponding to the private information.

[0156] Optionally, after the user confirms the replacement information corresponding to each private information, the corresponding relationship between the private information and the replacement information is stored in a local database to facilitate the replacement of the private information next time.

[0157] Step 305: Use the replacement letter corresponding to each private information to replace the private information in the first target data to obtain the second target data;

[0158] Step 306: encrypt the second target data with MD5, and send the encrypted second target data to the cloud server using https;

[0159] The key can be stored in the cloud server in advance for easy use during decryption.

[0160] Step 307: The cloud server decrypts the received data to obtain the second target data, sends the second target data to the large model for processing, encrypts the processing result with MD5, and sends it to the local device using https;

[0161] Taking chatglm3-6B as an example, according to the data length of the input large model, when the data length (token) is greater than 2000, multiple machines and multiple cards (i.e. multiple servers and multiple GPUs) can be used for model reasoning. When the data length (token) is less than 2000, single machine and multiple cards (one server and multiple GPUs) can be used for reasoning. This ensures that data is not blocked and the speed of data processing is guaranteed.

[0162] Step 308: The local device decrypts the received data to obtain a processing result;

[0163] Step 309: query the local database, restore the privacy information in the processing result, obtain the final processing result, and display the final processing result.

[0164] In implementation, according to the correspondence between the private information and the replacement information stored in the local database, the replacement information in the processing result is replaced with the private information, thereby restoring the private information.

[0165] like Figure 4 As shown, this embodiment provides a process of text privacy protection. Taking the first target data including the conference record text as an example, the text privacy protection process provided by this embodiment is described as follows:

[0166] Step 400: The local device obtains the first meeting record text uploaded by the user;

[0167] Step 401: query the local database, and use the local replacement information in the local database to replace the private information related to the company's privacy in the meeting record text to obtain a second meeting record text;

[0168] The privacy information includes but is not limited to the company name, company key content, important data and other information.

[0169] Step 402: Using a keyword recognition algorithm, continue to perform privacy recognition on the text in the second conference record that has not been replaced with privacy information to determine privacy information;

[0170] Step 403: Generate first replacement information and second replacement information of the private information respectively by using a synonym generation technology and a random word generation technology;

[0171] Step 404: display the privacy information, local replacement information, first replacement information and second replacement information on the human-computer interaction interface, and determine corresponding replacement information for each privacy information from the local replacement information, the first replacement information and the second replacement information in response to the user's editing operation on the replacement information.

[0172] Step 405: Use the replacement information corresponding to each piece of private information to replace the private information in the second meeting record text to obtain a third meeting record text;

[0173] Step 406: encrypt the third meeting record text and upload it to the cloud server;

[0174] Step 407: The cloud server decrypts the received data to obtain the third meeting record text, generates a meeting minutes of the third meeting record text using the large model, and encrypts the meeting minutes and sends them to the local device;

[0175] Step 408: The local device decrypts the received data to obtain the meeting minutes, queries the local database, restores the private information in the meeting minutes, and provides the restored meeting minutes to the user.

[0176] like Figure 5 As shown, this embodiment provides a process of text privacy protection. Taking the first target data including the work copy as an example, the text privacy protection process provided by this embodiment is described as follows:

[0177] Step 500: The local device obtains the first work copy uploaded by the user;

[0178] Step 501: query the local database, and use the local replacement information in the local database to replace the privacy information related to the privacy of the work in the work copy, so as to obtain a second work copy;

[0179] The privacy information includes but is not limited to the work structure, work details, important content and other information.

[0180] Step 502: Using a keyword recognition algorithm, continue to perform privacy recognition on the text in the second work that has not been replaced with privacy information to determine the privacy information;

[0181] Step 503: Generate first replacement information and second replacement information of the private information respectively by using a synonym generation technology and a random word generation technology;

[0182] Step 504: display the privacy information, local replacement information, first replacement information and second replacement information on the human-computer interaction interface, and determine corresponding replacement information for each privacy information from the local replacement information, the first replacement information and the second replacement information in response to the user's editing operation on the replacement information.

[0183] Step 505: Use the replacement information corresponding to each piece of private information to replace the private information in the second work copy, to obtain a third work copy;

[0184] Step 506: encrypt the third work text and upload it to the cloud server;

[0185] Step 507: The cloud server decrypts the received data to obtain the third work copy, processes the third work copy using the large model, and encrypts the processing result and sends it to the local device;

[0186] Step 508: The local device decrypts the received data to obtain a processing result, queries a local database, restores the private information in the processing result, and provides the restored processing result to the user.

[0187] like Figure 6 As shown, this embodiment provides a protection transmission process of image privacy information, which is specifically as follows:

[0188] Step 600: Acquire first target data;

[0189] Step 601: Determine the location of the private information of the first target data according to the target location input by the user;

[0190] Step 602: performing denoising and normalization processing on the first target data to obtain processed first target data;

[0191] Step 603: extracting a target image at a target position in the first target data and a background image other than the target position using a cutout network;

[0192] For example, if the target image is a portrait, the text message "Please fill the blank part of the image below with the portrait" can also be saved to fill the blank part of the background image.

[0193] Step 604: displaying the target image and the background image on the human-computer interaction interface, and determining the target image in the first target data in response to the user's editing operation on the target image;

[0194] During implementation, if the user feels that the cutout area (the area where the target image is located) is inappropriate, he or she can modify it by himself or herself, re-capture the area selected by the user as the target image, perform cutout processing on the first target data, and obtain a new target image and background image.

[0195] Optionally, after the user confirms each target image, the target image and the target position of the target image, or the target position of the target image and the first target data are stored in a local database to facilitate restoring the private information next time.

[0196] Step 605: Encrypt the background image and the target location, and send the encrypted second target data to the cloud server using https;

[0197] The key can be stored in the cloud server in advance for easy use during decryption.

[0198] Step 606: The cloud server decrypts the received data to obtain a background image and a target location;

[0199] Step 607: The cloud server generates a random image using the first large model, and adds the random image to a target position of the background image to obtain a privacy-removed image.

[0200] In the implementation, the first large model adds a random image to the position of the missing target image in the background image, and fuses the random image and the background image. When the user enters the target position by entering text, the text entered by the user can also be encrypted and sent to the cloud server together with the encrypted second target data. In the implementation, the first large model randomly generates a random image, adds the random image to the target position in the background image according to the received target position, and fuses the random image and the background image to generate a deprived image.

[0201] Taking the first largest model as Qwen1.5-72B as an example, the prompt of the first largest model can be set as follows: fill the blank area in the following image with a random image. That is, the first largest model is used to fill the target image that is cut out of the background image with a random image.

[0202] Step 608: The cloud server uses the second largest model to perform style conversion, encrypts the style-converted image with MD5, and sends it to the local device using https.

[0203] Taking the second largest model, the SD model, as an example, style transfer is performed on the background image (deprivacy image) filled with random images.

[0204] Step 609: The local device decrypts the received data to obtain an image after style conversion;

[0205] Step 610: Perform image processing on the first target data using a static image cutout algorithm to obtain a classified image, and fuse the classified image and the image after style conversion using an image fusion technique to obtain a fused image;

[0206] In practice, since other processing is done after the cutout, if the random image is directly replaced with the target image, its edge information will be mismatched or even empty. Therefore, image fusion technology can be used to fuse the target image and the background image. In order to optimize the details and reflect the edge information of the cutout, the fusion module is used to fuse the images to obtain a more accurate image.

[0207] Step 611: Use the first target data to perform style migration on the fused image and output a final result.

[0208] In the implementation, the style transfer network is used to transfer the style of the fused image and the first target data, and the image styles of the two are merged. Optionally, a VGG19 network is used as the style transfer network to perform image style transfer processing.

[0209] This embodiment encrypts the user's text and images and uploads them to the cloud to ensure the user's information security. First, for text, two methods of model + manual customization are used to select privacy, replace relevant privacy information, and add the information selected by the user to the local database for storage to accurately replace the next privacy information. Secondly, for images, the same two methods of model customization + manual automatic selection are adopted to replace the image position of the private part. For the security of information transmission between the local machine and the cloud, the information is encrypted using the md5 encryption method, and the transmission uses hhtps to further ensure data security. Decrypt the information processed in the cloud. First decrypt the md5 encrypted data, and then identify whether the data is an image or text. If it is text information corresponding to the local database, replace the corresponding privacy information. If it is image information, the image information replaced by the local machine is partially replaced with the image processed by the large model, and the image can be further optimized through the style transfer network to avoid image distortion caused by replacing privacy information.

[0210] like Figure 7 As shown, this embodiment provides a process of image privacy protection. Taking the first target data as an image as an example, the image privacy protection process provided by this embodiment is described as follows:

[0211] Step 700: Obtain the image to be processed uploaded by the user;

[0212] Step 701: determining the location of the private information of the image to be processed according to the target location input by the user;

[0213] During implementation, the location of the private information of the image is determined using the face location, tattoo location, etc. input by the user.

[0214] Step 702: performing denoising and normalization processing on the image to be processed to obtain a first image;

[0215] Step 703: extracting a target image at a target position in the first image and a background image other than the target position using a cutout network;

[0216] For example, if the target image is a portrait, the text message "Please fill the blank part of the image below with the portrait" can also be saved to fill the blank part of the background image.

[0217] Step 704: displaying the target image and the background image on the human-computer interaction interface, and determining the target image in response to the user's editing operation on the target image;

[0218] During implementation, if the user feels that the cutout area (the area where the target image is located) is inappropriate, he or she can modify it by himself or herself, re-capture the area selected by the user as the target image, perform cutout processing on the first target data, and obtain a new target image and background image.

[0219] Optionally, after the user confirms each target image, the target image and the target position of the target image, or the target position and the first target data are stored in a local database to facilitate restoring the private information next time.

[0220] Step 705: Encrypt the background image and the target location, and send the encrypted image to the cloud server using https;

[0221] The key can be stored in the cloud server in advance for easy use during decryption.

[0222] Step 706: The cloud server decrypts the received encrypted image to obtain a background image and a target position;

[0223] Step 707: The cloud server generates a random image using the first large model, and adds the random image to a target position of the background image to obtain a second image.

[0224] Taking the first largest model as Qwen1.5-72B as an example, the prompt of the first largest model can be set as follows: fill the blank area in the following image with a random image. That is, the first largest model is used to fill the target image that is cut out of the background image with a random image.

[0225] Step 708: The cloud server uses the second large model to perform comic effect processing or photo-editing on the second image to obtain a third image, encrypts the third image with MD5, and sends the third image to the local device using https.

[0226] Taking the second largest model, the SD model, as an example, style transfer is performed on the background image (deprivacy image) filled with random images.

[0227] Step 709: The local device decrypts the received encrypted image to obtain a third image;

[0228] Step 710: Use a static image matting algorithm to process the image to be processed to obtain a classified image, and use an image fusion technology to fuse the classified image and the third image to obtain a fused image;

[0229] Step 711: Output and display the fused image.

[0230] The data transmission method provided in this embodiment can effectively protect the privacy leakage risks brought by large models in the cloud and solve the objective problems of hardware and disk occupancy caused by local deployment of large models. By performing privacy protection locally, compared with the solution of directly uploading to the cloud, the user information of this embodiment is protected to prevent the leakage of user information. Compared with the solution of directly using local large model processing, the data transmission method provided in this embodiment has no special requirements for local hardware and is convenient for users to use. In the case of direct local large model processing, generally a better graphics card is required and it may take a long time. This embodiment can implement one-click privacy protection through a local database and a cutout network in different parts, while paying attention to the user's experience information. Give users the right to customize their own selections to enhance the user's sense of use. This embodiment will also perform image restoration and style synchronization after image restoration to avoid image differences between the cutout area and the non-cutout area due to direct replacement of the privacy area image.

[0231] Based on the same inventive concept, the embodiment of the present disclosure also provides an electronic device. Since the device is the device in the method in the embodiment of the present disclosure, and the principle of solving the problem by the device is similar to that of the method, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0232] like Figure 8 As shown, the electronic device includes a processor 800 and a memory 801, wherein the memory 801 is used to store a program executable by the processor 800, and the processor 800 is used to read the program in the memory 801 and perform the following steps:

[0233] Acquire first target data, and determine privacy information contained in the first target data;

[0234] Decrypting the privacy information in the first target data to obtain second target data;

[0235] The second target data is sent to a server, so that the server processes the second target data and returns a processing result.

[0236] As an optional implementation manner, the data type of the first target data is text; the processor 800 is specifically configured to execute:

[0237] Using a keyword extraction algorithm, the semantics of the first target data is analyzed to identify keywords contained in the first target data;

[0238] The privacy information contained in the first target data is determined according to the identified keywords.

[0239] As an optional implementation manner, the data type of the first target data is text; the processor 800 is specifically configured to execute:

[0240] According to the privacy information stored in the local database, determining, from the first target data, target data whose similarity to the stored privacy information is greater than a threshold;

[0241] The target data whose similarity is greater than a threshold is determined as the privacy information included in the first target data.

[0242] As an optional implementation manner, the data type of the first target data is text; the processor 800 is specifically configured to execute:

[0243] generating replacement information corresponding to the private information, wherein the replacement information corresponding to the private information has the same data type as the private information;

[0244] The private information included in the first target data is replaced with the replacement information to obtain second target data.

[0245] As an optional implementation manner, the processor 800 is specifically configured to execute:

[0246] Generate replacement information corresponding to the private information using synonym generation technology; and / or,

[0247] The replacement information corresponding to the private information is generated by using a random word generation technology.

[0248] As an optional implementation manner, the processor 800 is specifically configured to execute:

[0249] According to the correspondence between the private information and the replacement information stored in the local database, searching the local database for the replacement information corresponding to the private information included in the first target data;

[0250] The replacement information found is determined as the replacement information corresponding to the private information.

[0251] As an optional implementation manner, the processor 800 is further configured to execute:

[0252] Storing the corresponding relationship between the private information and the replacement information in a local database;

[0253] Receive the processing results sent by the server;

[0254] According to the correspondence between the private information and the replacement information stored in the local database, the replacement information in the processing result is replaced with the private information.

[0255] As an optional implementation manner, the data type of the first target data is an image; the processor 800 is specifically configured to execute:

[0256] Determining a target location of the private information in the first target data;

[0257] The privacy information included in the first target data is removed to obtain the second target data.

[0258] As an optional implementation manner, the processor 800 is specifically configured to execute:

[0259] In response to a target position input by a user, extracting a target image at the target position in the first target data and a background image other than the target position by using a cutout network;

[0260] The background image is determined as the second target data.

[0261] As an optional implementation manner, the processor 800 is specifically configured to execute:

[0262] Using a static image cutout algorithm, performing image processing on the first target data to obtain a classified image;

[0263] The first target data and the classified image are input into the cutout network, and the target image and the background image are output.

[0264] As an optional implementation manner, the processor 800 is further configured to execute:

[0265] The second target data and the target position are sent to a server, so that the server uses a randomly generated random image to perform image filling on the second target data at the target position.

[0266] As an optional implementation manner, the processor 800 is further configured to execute:

[0267] storing the target position and the first target data; receiving a processing result sent by a server, determining a target image according to the stored first target data and the target position, and replacing the image at the target position in the processing result with the target image; or,

[0268] The target position and the target image are stored; a processing result sent by a server is received, and the image at the target position in the processing result is replaced with the target image.

[0269] As an optional implementation manner, after replacing the image at the target position in the processing result with the target image, the processor 800 is further configured to execute:

[0270] Using a static image cutout algorithm, performing image processing on the first target data to obtain a classified image;

[0271] The classified image and the processing result are fused by using image fusion technology to obtain a fused processing result.

[0272] As an optional implementation manner, the processor 800 is further configured to execute:

[0273] Displaying the privacy information included in the first target data on a human-computer interaction interface, and in response to a user's editing operation on the privacy information, updating the corresponding relationship between the privacy information and the replacement information, and storing it in a database; and / or,

[0274] The replacement information corresponding to the private information included in the first target data is displayed on the human-computer interaction interface. In response to the user's editing operation on the replacement information, the corresponding relationship between the private information and the replacement information is updated and stored in the database.

[0275] Based on the same inventive concept, the embodiment of the present disclosure also provides a data transmission device. Since the device is the device in the method in the embodiment of the present disclosure, and the principle of solving the problem by the device is similar to that of the method, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0276] like Fig. 9 As shown, the device comprises:

[0277] The privacy determination module 900 is used to obtain first target data and determine the privacy information contained in the first target data;

[0278] A data determination module 901 is used to decrypt the privacy information in the first target data to obtain second target data;

[0279] The data sending module 902 is used to send the second target data to the server, so that the server processes the second target data and returns the processing result.

[0280] Based on the same inventive concept, an embodiment of the present disclosure provides a computer storage medium, the computer storage medium includes: a computer program code, when the computer program code is executed on a computer, the computer executes any of the data transmission methods discussed above. Since the principle of solving the problem by the above-mentioned computer storage medium is similar to that of the data transmission method, the implementation of the above-mentioned computer storage medium can refer to the implementation of the method, and the repeated parts will not be repeated.

[0281] In the specific implementation process, the computer storage medium may include: Universal Serial Bus Flash Drive (USB), mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other storage media that can store program codes.

[0282] Based on the same inventive concept, the embodiment of the present disclosure further provides a computer program product, which includes: computer program code, when the computer program code is run on a computer, the computer executes any of the data transmission methods discussed above. Since the principle of solving the problem by the above computer program product is similar to that of the data transmission method, the implementation of the above computer program product can refer to the implementation of the method, and the repeated parts will not be repeated.

[0283] The computer program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more conductors, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0284] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0285] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that has the functions specified in one or more boxes.

[0286] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0287] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0288] Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalents, the present disclosure is also intended to include these modifications and variations.

Claims

1. A data transmission method, wherein: The method includes: Acquire first target data, and determine privacy information contained in the first target data; Decrypting the privacy information in the first target data to obtain second target data; The second target data is sent to a server, so that the server processes the second target data and returns a processing result.

2. The method according to claim 1, wherein: The data type of the first target data is text; and determining the privacy information contained in the first target data includes: Using a keyword extraction algorithm, the semantics of the first target data is analyzed to identify keywords contained in the first target data; The privacy information contained in the first target data is determined according to the identified keywords.

3. The method according to claim 1, wherein: The data type of the first target data is text; and determining the privacy information contained in the first target data includes: According to the privacy information stored in the local database, determining, from the first target data, target data whose similarity to the stored privacy information is greater than a threshold; The target data whose similarity is greater than a threshold is determined as the privacy information included in the first target data.

4. The method according to claim 1, wherein: The data type of the first target data is text; and decrypting the privacy information in the first target data to obtain the second target data includes: generating replacement information corresponding to the private information, wherein the replacement information corresponding to the private information has the same data type as the private information; The private information included in the first target data is replaced with the replacement information to obtain second target data.

5. The method according to claim 4, wherein: The generating replacement information corresponding to the private information includes: Generate replacement information corresponding to the private information using synonym generation technology; and / or, The replacement information corresponding to the private information is generated by using a random word generation technology.

6. The method according to claim 4, wherein: The generating replacement information corresponding to the private information includes: According to the correspondence between the private information and the replacement information stored in the local database, searching the local database for the replacement information corresponding to the private information included in the first target data; The replacement information found is determined as the replacement information corresponding to the private information.

7. The method according to any one of claims 2 to 6, wherein: The method further includes: Storing the corresponding relationship between the private information and the replacement information in a local database; Receive the processing results sent by the server; According to the correspondence between the private information and the replacement information stored in the local database, the replacement information in the processing result is replaced with the private information.

8. The method according to claim 1, wherein: The data type of the first target data is an image; and the privacy information in the first target data is decrypted to obtain the second target data, including: Determining a target location of the private information in the first target data; The privacy information included in the first target data is removed to obtain the second target data.

9. The method according to claim 8, wherein: The removing the privacy information included in the first target data to obtain the second target data includes: In response to a target position input by a user, extracting a target image at the target position in the first target data and a background image other than the target position by using a cutout network; The background image is determined as the second target data.

10. The method according to claim 9, wherein: The step of extracting the target image at the target position in the first target data and the background image other than the target position by using a cutout network includes: Using a static image cutout algorithm, performing image processing on the first target data to obtain a classified image; The first target data and the classified image are input into the cutout network, and the target image and the background image are output.

11. The method according to claim 8, wherein: The method further comprises: The second target data and the target position are sent to a server, so that the server uses a randomly generated random image to perform image filling on the second target data at the target position.

12. The method according to any one of claims 8 to 11, wherein: The method further comprises: storing the target position and the first target data; receiving a processing result sent by a server, determining a target image according to the stored first target data and the target position, and replacing the image at the target position in the processing result with the target image; or, The target position and the target image are stored; a processing result sent by a server is received, and the image at the target position in the processing result is replaced with the target image.

13. The method according to claim 12, wherein: After replacing the image at the target position in the processing result with the target image, the method further includes: Using a static image cutout algorithm, performing image processing on the first target data to obtain a classified image; The classified image and the processing result are fused by using image fusion technology to obtain a fused processing result.

14. The method according to claim 4, wherein: The method further includes: Displaying the privacy information included in the first target data on a human-computer interaction interface, and in response to a user's editing operation on the privacy information, updating the correspondence between the privacy information and the replacement information, and storing the updated information in a database; and / or, The replacement information corresponding to the private information included in the first target data is displayed on the human-computer interaction interface. In response to the user's editing operation on the replacement information, the corresponding relationship between the private information and the replacement information is updated and stored in the database.

15. An electronic device, wherein: The electronic device comprises a processor and a memory, wherein the memory is used to store a program executable by the processor, and the processor is used to read the program in the memory and execute the steps of any one of the methods described in claims 1 to 14.

16. A computer storage medium having a computer program stored thereon, wherein: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

Images