Unstructured data processing method based on additional information embedding and homomorphic encryption

Abstract

The present invention belongs to the field of confidential communication technology, and discloses a method for processing unstructured data based on embedding additional information and homomorphic encryption. The method includes the following: generating feature vectors by preprocessing unstructured data, generating embedded signals by combining feature vectors with additional information coding and modulation technology, and dynamically adjusting the embedding position through an optimization algorithm to minimize the impact on data quality, thereby completing the embedding of additional information; encrypting the data embedded with additional information using an encryption algorithm that supports homomorphic operations; and restoring the embedded additional information through decryption, feature vector extraction, demodulation, and decoding during data use. Compared with the prior art, the present invention can effectively embed additional information in unstructured data, while using homomorphic encryption technology to ensure the security and privacy of data, and is suitable for a variety of data forms and application scenarios.

Description

Technical Field

[0001] The present invention relates to the field of secure communication technology, and in particular to an unstructured data processing method based on additional information embedding and homomorphic encryption. Background Art

[0002] Unstructured data (such as images, audio, video, and text) plays an important role in modern information processing and storage. With the development of big data and cloud computing, the demand for the generation, transmission, and processing of these data has increased dramatically. However, unstructured data usually lacks a fixed format and is difficult to directly manage and protect efficiently. In addition, the sharing and analysis of data may face security risks such as privacy leakage and data tampering. Therefore, how to embed useful information, achieve data protection, and improve its application value in the process of unstructured data processing has become an important research direction.

[0003] Additional information embedding technology embeds identification, copyright information or authentication information into the original data, which is used for data traceability, anti-counterfeiting and copyright protection. However, additional information embedding usually requires a trade-off between the embedding capacity and the impact of data quality, and ensures the extractability of the embedded information in complex environments. In addition, with the increase in data privacy protection requirements, traditional encryption methods face challenges between efficiency and security.

[0004] As an emerging cryptographic technology, homomorphic encryption can directly perform operations on encrypted data in the ciphertext domain and process data without decryption, which has high security and flexibility. In the scenario of combining with additional information embedding, homomorphic encryption provides a new solution for the protection of data throughout its life cycle. However, the existing technology has not fully explored the combined application of additional information embedding and homomorphic encryption, especially in the specific implementation methods and performance optimization in the field of unstructured data. There is still room for improvement. Summary of the invention

[0005] In view of the above defects in the prior art, the present invention proposes an unstructured data processing method based on additional information embedding and homomorphic encryption.

[0006] The technical solution of the present invention is achieved in this way:

[0007] A method for processing unstructured data based on additional information embedding and homomorphic encryption, comprising the following steps:

[0008] 1) Embed additional information in unstructured data:

[0009] 11) Preprocess the unstructured data to generate feature vectors associated with the content of the unstructured data;

[0010] 12) Encode and modulate the additional information, and combine the feature vector to generate an additional information feature signal suitable for embedding;

[0011] 13) Adjust the embedding position of the feature vector and embed the modulated additional information feature signal into it;

[0012] 14) performing a reconstruction operation on the feature vector in which the feature signal of the additional information is embedded, so as to generate unstructured data in which the additional information is embedded;

[0013] 2) Encrypt the unstructured data embedded with the additional information using an encryption algorithm that supports homomorphic operations to generate encrypted unstructured data embedded with the additional information;

[0014] 3) The user extracts additional information from the encrypted unstructured data in which the additional information is embedded by following the steps below:

[0015] 31) Decrypting the encrypted unstructured data embedded with the additional information and recovering the unstructured data embedded with the additional information;

[0016] 32) Perform feature vector extraction operations on unstructured data embedded with additional information to obtain additional information feature signals;

[0017] 33) Demodulate and decode the acquired additional information characteristic signal to restore the original additional information.

[0018] Preferably, the method for preprocessing the unstructured data in step 11) includes but is not limited to the following:

[0019] (1) Performing discrete cosine transform, discrete Fourier transform, or discrete wavelet transform on unstructured data to generate frequency domain feature vectors;

[0020] (2) Divide the unstructured data into blocks and calculate the statistical characteristics of each block;

[0021] (3) Extract encrypted feature information related to the edge or structure of unstructured data.

[0022] Preferably, the method for modulating the additional information in step 12) includes but is not limited to one of amplitude modulation, phase modulation, spread spectrum modulation, and quantization index modulation.

[0023] Preferably, in step 13), the embedding position of the feature vector is dynamically adjusted by an optimization algorithm, and the optimization algorithm includes but is not limited to the following:

[0024] (1) Based on the selection strategy of frequency domain component energy distribution, medium and high frequency coefficients are embedded preferentially to reduce the impact on the quality of unstructured data;

[0025] (2) Embedding position optimization based on specific constraints;

[0026] (3) Adaptively adjust the embedding position based on the statistical distribution of the encrypted feature vector.

[0027] Preferably, the reconstruction operation in step 14) includes performing an inverse frequency domain transform on the feature vector in which the feature signal of the additional information is embedded, so as to restore the unstructured data in which the additional information is embedded.

[0028] Preferably, the encryption algorithm is an encryption scheme that supports fully homomorphic operations, and the encryption scheme includes but is not limited to one of Paillier encryption, BFV encryption, BGV encryption or CKKS encryption.

[0029] Preferably, the encoding of the additional information in step 12) includes but is not limited to one of error correction encoding, data compression encoding, and hash encoding.

[0030] Preferably, the error correction coding includes Hamming coding, BCH coding or LDPC coding, and the data compression coding includes Huffman coding or arithmetic coding.

[0031] Preferably, the statistical feature includes mean, variance or gradient direction.

[0032] Preferably, the optimization algorithm (2) is: minimizing the difference between the unstructured data feature vector after embedding and the original feature vector.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The present invention discloses an unstructured data processing method based on additional information embedding and homomorphic encryption. The present invention embeds identification or authentication information in data through additional information embedding technology, which can be used for data traceability, anti-counterfeiting and copyright protection, and is particularly suitable for complex unstructured data scenarios; the present invention is suitable for various unstructured data such as images, audio, video, text, etc., and has strong universality and flexibility; the optimization algorithm proposed in the present invention dynamically adjusts the embedding position, reduces the impact of additional information embedding on the quality of original data, and improves processing efficiency by optimizing the embedding process and the decoding process; adopts various coding and modulation technologies (such as error correction coding, quantization index modulation, etc.) to enhance the robustness of embedded information, so that it still has high extractability and accuracy under noise interference or compression. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention is a flowchart of the unstructured data processing method based on additional information embedding and homomorphic encryption. DETAILED DESCRIPTION

[0036] The present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0037] like Figure 1 As shown, a method for processing unstructured data based on additional information embedding and homomorphic encryption of the present invention comprises the following steps:

[0038] 1) Embed additional information in unstructured data:

[0039] 11) Preprocess the unstructured data to generate feature vectors associated with the content of the unstructured data;

[0040] 12) Encode and modulate the additional information, and combine the feature vector to generate an additional information feature signal suitable for embedding;

[0041] 13) Dynamically adjust the embedding position of the unstructured data feature vector through the optimization algorithm, embed the modulated additional information feature signal into it, and minimize the impact on the quality of the unstructured data;

[0042] 14) performing a reconstruction operation on the feature vector in which the feature signal of the additional information is embedded, so as to generate unstructured data in which the additional information is embedded;

[0043] 2) Encrypt the unstructured data embedded with the additional information using an encryption algorithm that supports homomorphic operations to generate encrypted unstructured data embedded with the additional information;

[0044] 3) The user extracts additional information from the encrypted unstructured data in which the additional information is embedded by following the steps below:

[0045] 31) Decrypting the encrypted unstructured data embedded with the additional information and recovering the unstructured data embedded with the additional information;

[0046] 32) Perform feature vector extraction operations on unstructured data embedded with additional information to obtain additional information feature signals;

[0047] 33) Apply a demodulation algorithm and a decoding algorithm to the extracted additional information characteristic signal to restore the original additional information.

[0048] The methods for preprocessing unstructured data in step 11) include but are not limited to the following:

[0049] (1) Perform discrete cosine transform (DCT), discrete Fourier transform (DFT), or discrete wavelet transform (DWT) on unstructured data to generate frequency domain feature vectors;

[0050] (2) Divide the unstructured data into blocks and calculate the statistical features of each block, such as mean, variance or gradient direction;

[0051] (3) Extract encrypted feature information related to the edges or other structures of unstructured data.

[0052] The method of modulating the additional information in step 12) includes, but is not limited to, amplitude modulation (AM), phase modulation (PM), spread spectrum modulation, and quantization index modulation (QIM).

[0053] In step 13), the method of dynamically adjusting the embedding position by optimizing the algorithm includes but is not limited to the following methods:

[0054] (1) Based on the selection strategy of frequency domain component energy distribution, medium and high frequency coefficients are embedded preferentially to reduce the impact on the quality of unstructured data;

[0055] (2) Optimizing the embedding position based on specific constraints, such as by minimizing the difference between the feature vector of the embedded unstructured data and the original feature vector;

[0056] (3) Adaptively adjust the embedding position based on the statistical distribution of the encrypted feature vector.

[0057] The reconstruction operation in step 14) includes performing an inverse frequency domain transform on the feature vector in which the feature signal of the additional information is embedded, so as to restore the unstructured data in which the additional information is embedded.

[0058] The encryption algorithm is an encryption scheme that supports fully homomorphic operations, and the encryption scheme includes but is not limited to Paillier encryption, BFV encryption, BGV encryption or CKKS encryption.

[0059] The encoding of the additional information in step 12) includes but is not limited to error correction coding, such as Hamming coding, BCH coding or LDPC coding; data compression coding, such as Huffman coding or arithmetic coding; and hash coding. Embodiment 1:

[0060] Preprocessing: discrete cosine transform (DCT); Modulation: quantization index modulation; Coding: Hamming coding; Optimization algorithm: embedding position optimization based on specific constraints; Encryption scheme: Paillier encryption.

[0061] The user wants to embed additional information "1010" (in binary form) into a grayscale image of size . After the user completes the embedding operation, the user uses the fully homomorphic encryption algorithm Paillier to encrypt the image with the embedded additional information, and then decrypts and extracts the additional information. The detailed steps of the process are as follows:

[0062] Step 1: Embed additional information:

[0063] 1.1 Original image matrix (grayscale value range is 0-255):

[0064] ,

[0065] 1.2 Image data preprocessing:

[0066] For image data Perform discrete cosine transform,

[0067] ,

[0068] Get the frequency domain feature matrix :

[0069] ,

[0070] 1.3 Additional information coding and modulation:

[0071] 1.3.1 Error correction encoding of the additional information "1010" (taking Hamming encoding as an example):

[0072] ,

[0073] 1.3.2 Assuming the use of quantization index modulation Modulation is performed on the frequency domain eigenvalues, and the modulation parameters (quantization step size). For example, for the frequency domain eigenvalue , after embedding bit 1, it is adjusted to:

[0074] ,

[0075] in It means to take the odd number closest to the parameter.

[0076] 1.4 Dynamic embedding optimization:

[0077] 1.4.1 Select mid-high frequency components in the frequency domain as the embedding location to reduce the impact on visual quality.

[0078] 1.4.2 Adjust the embedding position through the optimization algorithm to minimize the difference between the embedded feature vector and the original feature vector.

[0079] 1.5 Feature vector reconstruction:

[0080] The frequency domain feature matrix after embedding additional information Reverse Transformation to generate image data with embedded additional information .

[0081] Step 2: Encrypt the image data embedded with additional information:

[0082] Use the Paillier encryption algorithm to encrypt the image data embedded with additional information pixel by pixel to obtain the encrypted image embedded with additional information .

[0083] Step 3: Extract additional information:

[0084] 3.1 Encrypted images with embedded additional information Decrypt to obtain the image plaintext with embedded additional information .

[0085] 3.2 pairs Perform feature vector extraction to obtain a feature vector containing additional information.

[0086] 3.3 Apply the demodulation and decoding algorithm to the extracted feature vector to recover the original additional information. Embodiment 2:

[0087] Preprocessing: discrete wavelet transform (DWT); Modulation: quantization index modulation; Coding: Hamming coding; Optimization algorithm: selection strategy based on frequency domain component energy distribution; Encryption scheme: Paillier encryption.

[0088] The user wants to The color picture is embedded with an additional information "1010". After the user completes the embedding operation, the picture with the embedded additional information is encrypted using the fully homomorphic encryption algorithm Paillier.

[0089] Step 1: Embed additional information:

[0090] 1.1 The original image consists of three channels, RGB, and each channel is a The pixel value range is 0-255.

[0091] Red Channel :

[0092] ,

[0093] Green Channel :

[0094] ,

[0095] Blue channel :

[0096] ,

[0097] 1.2 Image data preprocessing:

[0098] Perform discrete wavelet transform (DWT) on the three RGB channels to extract feature vectors, including low-frequency subbands and high frequency subband , , (The wavelet basis is Haar wavelet):

[0099] ,

[0100] For example, the low-frequency subband of the red channel and high frequency subband , , They are:

[0101] ,

[0102] The green channel is processed similarly to the blue channel.

[0103] 1.3 Additional information coding and modulation:

[0104] Additional information Hamming encoding is performed, and then quantization index modulation (QIM) is performed on the signal value. For example, for , select the modulation parameters , after embedding bit 1, it is adjusted to .

[0105] 1.4 Dynamic Embedding Optimization

[0106] Select the high-frequency sub-bands in the three RGB channels (such as , ) to embed and dynamically adjust the embedding position to minimize the impact of embedding on the original feature distribution, that is, to minimize , etc., among which , etc. are wavelet feature matrices that embed additional information.

[0107] 1.5 Feature vector reconstruction:

[0108] right , Perform inverse wavelet transform to generate an encrypted image matrix with embedded additional information .

[0109] Step 2: Encrypt the image data embedded with additional information:

[0110] The user uses the Paillier algorithm to encrypt the three channels separately to generate an encrypted image data matrix .

[0111] From the embodiments of the present invention, it can be seen that the unstructured data processing method based on additional information embedding and homomorphic encryption of the present invention embeds identification or authentication information in the data through additional information embedding technology, which can be used for data traceability, anti-counterfeiting and copyright protection, and is particularly suitable for complex unstructured data scenarios; the present invention is applicable to various unstructured data such as images, audio, video, text, etc., and has strong universality and flexibility; the optimization algorithm proposed in the present invention dynamically adjusts the embedding position, reduces the impact of additional information embedding on the quality of the original data, and improves the processing efficiency by optimizing the embedding process and the decoding process; adopts a variety of coding and modulation technologies (such as error correction coding, quantization index modulation, etc.) to enhance the robustness of the embedded information, so that it still has high extractability and accuracy under noise interference or compression.

Claims

1. A method for processing unstructured data based on additional information embedding and homomorphic encryption, characterized in that: The following steps are involved: 1) Embed additional information in unstructured data: 11) Preprocess the unstructured data to generate feature vectors associated with the content of the unstructured data; 12) Encode and modulate the additional information, and combine the feature vector to generate an additional information feature signal suitable for embedding; 13) Adjust the embedding position of the feature vector and embed the modulated additional information feature signal into it; 14) performing a reconstruction operation on the feature vector in which the feature signal of the additional information is embedded, so as to generate unstructured data in which the additional information is embedded; 2) Encrypt the unstructured data embedded with the additional information using an encryption algorithm that supports homomorphic operations to generate encrypted unstructured data embedded with the additional information; 3) The user extracts additional information from the encrypted unstructured data in which the additional information is embedded by following the steps below: 31) Decrypting the encrypted unstructured data embedded with the additional information and recovering the unstructured data embedded with the additional information; 32) Perform feature vector extraction operations on unstructured data embedded with additional information to obtain additional information feature signals; 33) Demodulate and decode the acquired additional information characteristic signal to restore the original additional information. In step 13), discrete wavelet transform is performed on the three channels of RGB to extract feature vectors, including low-frequency subbands and high frequency subband , , , dynamic embedding optimization, selects high-frequency subbands in the three RGB channels for embedding, and dynamically adjusts the embedding position to minimize the impact of embedding on the original feature distribution, that is, to minimize , ,in , is the wavelet feature matrix with additional information embedded, and the feature vector is reconstructed: , Perform inverse wavelet transform to generate an encrypted image matrix with embedded additional information .

2. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 1 is characterized in that: The methods for preprocessing unstructured data in step 11) include the following: (1) Performing discrete cosine transform, discrete Fourier transform, or discrete wavelet transform on unstructured data to generate frequency domain feature vectors; (2) Divide the unstructured data into blocks and calculate the statistical characteristics of each block; (3) Extract encrypted feature information related to the edge or structure of unstructured data.

3. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 1 is characterized in that: The method for modulating the additional information in step 12) includes one of amplitude modulation, phase modulation, spread spectrum modulation, and quantization index modulation.

4. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 3 is characterized in that: In step 13), the embedding position of the feature vector is dynamically adjusted by an optimization algorithm, and the optimization algorithm includes the following: (1) Based on the selection strategy of frequency domain component energy distribution, mid-high frequency coefficients are embedded to reduce the impact on the quality of unstructured data; (2) Embedding position optimization based on specific constraints; (3) Adaptively adjust the embedding position based on the statistical distribution of the encrypted feature vector.

5. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 3 is characterized in that: The reconstruction operation in step 14) includes performing an inverse frequency domain transform on the feature vector in which the feature signal of the additional information is embedded, so as to restore the unstructured data in which the additional information is embedded.

6. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 1 is characterized in that: The encryption algorithm is an encryption scheme that supports fully homomorphic operations, and the encryption scheme includes one of Paillier encryption, BFV encryption, BGV encryption or CKKS encryption.

7. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 1 is characterized in that: The encoding of the additional information in step 12) includes one of error correction encoding, data compression encoding, and hash encoding.

8. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 7 is characterized in that: The error correction coding includes Hamming coding, BCH coding or LDPC coding, and the data compression coding includes Huffman coding or arithmetic coding.

9. The unstructured data processing method based on additional information embedding and homomorphic encryption according to claim 2 is characterized in that: The statistical features include mean, variance or gradient direction.

Images