Remote sensing image trusted transaction method based on digital fingerprint tracing on block chain

Abstract

The invention discloses a remote sensing image trusted transaction method based on digital fingerprint tracing on a block chain. According to the method, the CNN model is adopted for feature extraction, then dimension reduction mapping is achieved through asymmetric transfer hash, a safe encrypted searchable index is generated through ASPE, and safe and efficient remote sensing image retrieval is achieved. And in combination with a block chain technology and a Merkel tree structure, the credibility of storage and retrieval services of the cloud server is ensured. Besides, pixels are marked through pixel prediction errors and Huffman coding, so that space is reserved for digital fingerprint embedding, and flexible tracing of malicious leakage behaviors is realized. According to the method, the authenticity, integrity and copyright protection of the transaction image are ensured while the remote sensing image transaction task is safely and efficiently completed.

Description

Technical Field

[0001] The present invention belongs to the field of spatial data security, and specifically is a remote sensing image trusted transaction method based on digital fingerprint tracing on a blockchain. Background Art

[0002] As earth observation technology continues to mature, the safe and efficient implementation of large-scale remote sensing image storage, processing and trading has become a problem that needs to be solved urgently. The high-quality data services provided by image trading service providers (ITSPs) have become an effective way to solve these problems. In particular, distributed ITSPs have attracted much attention from scholars due to their advantages such as parallel processing, intelligent retrieval, and secure transactions. However, there is a risk of privacy leakage when uploading remote sensing images directly to cloud servers. To protect image privacy, encryption is usually required before uploading, but this will destroy the image content and make encrypted image retrieval and trading difficult. Although existing solutions can achieve encrypted image retrieval, they still face problems such as low search efficiency and low credibility of transaction data. In addition, malicious framing by copyright holders and malicious disclosure by buyers are important factors that endanger the legitimacy, fairness, and credibility of data transactions. Infringement will increase the risk and uncertainty of data transactions, thereby affecting the healthy development of the data trading market.

[0003] Through the summary of the prior art, it is found that there are at least the following technical problems:

[0004] 1) The heterogeneous nature of the feature space of multi-source transaction images will affect the fair execution of tasks, and the high-dimensional features extracted by deep learning will take up a lot of storage and computing resources. In addition, remote sensing images contain rich and sensitive geographic information, and encryption protection will destroy plaintext information, resulting in limited retrieval functions.

[0005] 2) In the current image transaction retrieval services, there is a lack of sufficient attention to the threats posed by potential malicious cloud servers and malicious nodes, which makes it difficult for users to ensure whether the query results returned by the cloud server are credible or meet expectations.

[0006] 3) The current image transactions have mutual trust issues such as malicious framing by copyright holders and malicious disclosure by buyers, which exacerbates the risk of copyright protection. Existing copyright protection methods based on digital watermarks rarely consider the reversibility and separability of image encryption and watermark embedding, making it difficult to ensure high fidelity of remote sensing images and flexibility in data leakage tracing. Summary of the invention

[0007] The present invention proposes a remote sensing image trusted transaction method based on digital fingerprint tracing on blockchain, aiming to solve the threat problems posed by semi-trusted cloud servers and malicious users to the authenticity, integrity and copyright protection of remote sensing image transactions.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] A remote sensing image trusted transaction method based on digital fingerprint tracing on a blockchain, comprising:

[0010] S1: The data owner and the cloud server register the blockchain account address and join the blockchain network;

[0011] S2: The data owner uses the CNN model to extract image features, and reduces the dimension based on asymmetric transfer hashing, and then performs asymmetric dot product to maintain encryption and generate an encrypted searchable index;

[0012] S3: The data owner performs prediction error calculation, image encryption, and Huffman coding marking in sequence to generate an encrypted image with reserved embedding space;

[0013] S4: The data owner stores lightweight information such as encrypted image hash values ​​on the blockchain and transmits encrypted image information such as encrypted searchable index, encrypted image, and encrypted image hash values ​​to the cloud server;

[0014] S5: The data buyer sends a retrieval request to the blockchain, and after identity verification is passed, it is registered as a node on the blockchain;

[0015] S6: The data buyer uses the CNN model to extract image features, and then performs asymmetric transfer hashing to reduce the dimension, and then performs asymmetric dot product to maintain encryption, generate a query trapdoor, and then initiate a transaction through a smart contract. The transaction information includes the query trapdoor and the prepaid principal;

[0016] S7: The cloud server retrieves the secure searchable index according to the query trapdoor on the smart contract, returns the similarity ranking of all feature vectors, generates a Merkle tree based on all images according to the similarity ranking, and then uploads the root hash and similarity ranking of the Merkle tree to the blockchain, and provides a Top-k search image for data buyers to select the data to be traded;

[0017] S8: After the data buyer determines the data to be traded, the cloud server first obtains the Huffman coding rules stored in the reference pixel position of the encrypted image, and then locates the embeddable pixels in the image according to the characteristics of its different-letter codeword, and embeds the digital fingerprint into it by bit replacement, and finally returns the transaction image containing the digital fingerprint to the data buyer;

[0018] S9: The credibility verification of the retrieval results is done by the blockchain generating a Merkle tree based on the hash value of the Top-k retrieval image and the hash value of the encrypted image stored on the chain according to the similarity sorting, and then verifying the authenticity, integrity and correctness of the similarity sorting of the Top-k retrieval image based on the Merkle tree root hash and similarity sorting uploaded to the cloud server;

[0019] S10: Data leakage traceability verification is performed by blockchain to extract and obtain Huffman code at the reference pixel position of the encrypted image, locate excess embeddable pixels, and extract the embedded digital fingerprint. The extracted digital fingerprint is then compared with the digital fingerprint stored in the smart contract to verify the data leakage behavior and determine the copyright liability.

[0020] Compared with the prior art, the present invention has the following beneficial technical effects:

[0021] After using the pre-trained CNN model for feature extraction, asymmetric transfer hashing is used to achieve dimensionality reduction mapping, and the index is protected according to the ASPE algorithm to achieve safe and efficient remote sensing image retrieval. Blockchain technology and Merkle tree structure are used to ensure the credibility of cloud servers in storage and retrieval services. In addition, pixels are marked according to pixel prediction errors and Huffman coding, thereby reserving space for digital fingerprint embedding, so as to achieve flexible tracing of malicious leakage behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0023] Figure 1 A schematic diagram of the overall framework of a remote sensing image trusted transaction method based on digital fingerprint tracing on a blockchain provided by the present invention;

[0024] Figure 2 The image encryption flow chart with reserved embeddable space provided by the present invention;

[0025] Figure 3 The MED predictor provided by the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] See attached Figure 1 The embodiment of the present invention discloses a remote sensing image trusted transaction method based on digital fingerprint tracing on a blockchain, comprising the following steps:

[0028] Step 1: The key generation center generates keys for the data owner, data buyer, and cloud server:

[0029] First, the key generation center randomly generates a d-dimensional binary vector G of length d+1 and two reversible matrices M1 and M2 of size (d+1)×(d+1). Then, the encryption parameters are generated according to the user (data owner or data buyer) identity θ. For the data owner, four reversible matrices are randomly generated: Among them A θ1，1 ·A θ1，2 =M1,B θ1,1 ·B θ1,2 =M2. For data buyers, 4 reversible matrices are randomly generated in K is a set of real numbers encoded in n bits. I，i Represents the key required to encrypt an image using a stream cipher;

[0030] Step 2: The data owner and the cloud server register the blockchain account address to join the blockchain network, and the key generation center will I，i , A θ1,1 , B θ1,1 To the data owner.

[0031] Step 3: Image owner builds an encrypted searchable index.

[0032] Step 1: The image owner uses the ImageNet pre-trained VGG-f model to extract the image from the image collection Extract image features and obtain a set of feature vectors based on asymmetric transfer hashing

[0033] Step 2: For each eigenvector f t , 1≤t≤n, first in f t Add ||f t || 2 Expand it to (d+1) dimensions.

[0034] Step 3: When the binary vector G is equal to 0, let f t，a =f t，b =f t , otherwise f t，a +f t，b =f t , f t Divide into f t，a and f t，b Two vectors. Then according to the random matrix A θ1，1 , B θ1，1 Each expanded eigenvector f t Encrypted to ft ′=(f t， a A θ1,1 ,f t,b B θ1,1 ).

[0035] Step 4: The image owner encrypts the image set and reserves space.

[0036] The image owner performs prediction error (PEs) calculation, image encryption and Huffman coding marking in sequence. The flowchart is as follows Figure 2 As shown, the specific implementation includes:

[0037] Step 1: Calculate image m based on the MED predictor t The prediction error The MED predictor uses the first row and first column of the image as reference pixels and calculates the predicted value of the current pixel based on its left, top, and left top neighbor pixels. Figure 3 As shown, the predicted value of the current pixel x(i, j) Expressed as

[0038]

[0039] Step 2: Set the initial partition threshold, regard PEs with a probability lower than the partition threshold as one case, divide PEs with a probability higher than the partition threshold into s cases, and perform Huffman coding on s+1 prediction errors according to the probability of occurrence. Finally, take the pixels in the first row and the first column as reference pixels, take pixels with a probability of PEs lower than the partition threshold as non-embeddable pixels, and take pixels with a probability of PEs higher than the partition threshold as embeddable pixels.

[0040] Step 3: Using the key K I，i The image set is encrypted by stream cipher Encrypted

[0041] Step 4: Record the reference pixel information as auxiliary information without marking. Non-embeddable pixels are replaced and marked according to their corresponding Huffman codewords, and the replaced information is recorded as auxiliary information. Embeddable pixels are marked according to the Huffman codewords corresponding to their PEs, and the non-marked bits of each embeddable pixel are reserved spaces where auxiliary information can be embedded. After marking, the Huffman coding rules are converted into a binary bit stream and stored in the original reference pixel position in a bit replacement manner. The auxiliary information is embedded in the reserved space of the embeddable pixel in a bit replacement manner to facilitate subsequent information embedding and image recovery.

[0042] Step 5: The data owner stores lightweight information such as the encrypted image hash value on the blockchain, and transmits the encrypted image information such as the encrypted searchable index, the encrypted image, and the encrypted image hash value to the cloud server;

[0043] Step 6: The data buyer first sends a retrieval request to the blockchain. After the identity verification is passed, the buyer will be registered as a node on the blockchain and the key generation center will be notified to transfer the key K I，i , A θ2，1 , B θ2，1 Sent to data buyers.

[0044] Step 7: The data buyer initiates the transaction.

[0045] Step 1: The data buyer uses the ImageNet pre-trained VGG-f model to extract image features from the query image and obtains the query vector f based on asymmetric transfer hashing. q .

[0046] Step 2: f q Multiply it by (-2) and add 1 to the end of the vector to get a (d+1)-dimensional query vector.

[0047] Step 3: When the binary vector G is equal to 0, let f q，a =f q，b =f q , otherwise f q，a +f q，b =f q , f q Divide into f q，a and f q，b Two vectors.

[0048] Step 4: According to the random matrix A θ2,1 , B θ2,1 Each expanded eigenvector f q Encrypted Where γ is a random positive integer used to ensure the randomness and security of the query trapdoor.

[0049] Step 5: Initiate a transaction through a smart contract. The transaction information is {f′ q , C}, C is the prepaid principal, including the transaction payment amount and margin.

[0050] Step 8: The cloud server performs image retrieval and returns the retrieval results and trusted retrieval credentials.

[0051] Step 1: The key generation center will A θ1，2 , B θ1,2 , A θ2，2 , B θ2,2 Leave it to the cloud server.

[0052] Step 2: Cloud server based on A θ1,2 , B θ1,2 Each encrypted feature vector f t 'Convert to Right now According to A θ2,2 , B θ2,2 The query trapdoor f′ q Convert to Right now

[0053] Step 3: Cloud server based on index vector and query vector Calculate f t With f q Euclidean distance between HD t ,Right now The top-k encrypted images are sorted by distance and returned to the data buyer.

[0054] Step 4: Cloud Server Each encrypted image m′ in t And its ID are hashed to get leaf node N t =h(m′ t ||ID(m′ t )). Then, the leaf nodes are sorted by similarity to generate internal nodes and root nodes in turn, thus generating a Merkle tree. Finally, the cloud server will The root hash N root The similarity ranking is uploaded to the blockchain as a trusted retrieval credential.

[0055] Step 9: The image buyer verifies the credibility of the retrieval results through the smart contract.

[0056] Step 1: Smart contract {m′ Top1 , m′ Top2 ,…,m′ Topk}, and hash each encrypted image to obtain the leaf node {N Top1 ,N Top2 …, N Topk}, and sort the leaf nodes {N according to the similarity of the encrypted image hash stored on the blockchain. Top(k+1) , N Top(k+2) …, N Topn}.

[0057] Step 2: Set the leaf node {N Top1 ,N Top2 ,…,N Topk , NTop(k+1) ,…,N Topn} Generate internal nodes and root nodes in sequence according to similarity, thus generating a Merkle tree Its root node is N′ root .

[0058] Step 3: Verify N′ root With N root If they are the same, it means that the Top-k retrieved images are real and the similarity ranking is correct. Otherwise, it means that the encrypted image has been tampered with or the retrieved image is not returned according to the similarity ranking. According to the similarity ranking uploaded by the cloud server, by comparing whether the ID corresponding to each similarity covers the entire image set, it is verified whether the cloud server has completely retrieved the entire image set.

[0059] Step 4: The image buyer verifies the credibility of the retrieved image and determines the image to be traded.

[0060] Step 10: The cloud server embeds a digital fingerprint in the transaction data.

[0061] Step 1: The smart contract generates and stores the GD-PBIBD codeword matrix (v, b, r, k, λ1, λ2). Specifically, v elements are arranged into b subsets of size k, each element is repeated r times, and each pair of elements appears λ1 times within a group and λ2 times between groups. Among them, v is the code length and b is the number of users. The GD-PBIBD code is usually defined as (s p ,s 2(p-1) ,s p-1 , s, 0, 1), S is a prime number, p is a positive integer greater than 2, and it can resist s-1 colluders.

[0062] Step 2: The cloud server first obtains the Huffman coding rules stored at the reference pixel position of the encrypted image, and then locates the embeddable pixels in the image according to the characteristics of its heteronymous codeword, and embeds the digital fingerprint provided by the smart contract into it by bit replacement, and finally returns the transaction image containing the digital fingerprint to the data buyer.

[0063] Step 11: Data breach retrospective verification.

[0064] Step 1: When the data owner finds a suspicious image, the copy is digitally fingerprinted through a smart contract to verify whether it is a data leak.

[0065] Step 2: The smart contract extracts and obtains the Huffman code at the reference pixel position of the encrypted image, locates the redundant embeddable pixels, and extracts the embedded digital fingerprint. The extracted digital fingerprint is then compared with the digital fingerprint stored in the smart contract to verify the data leakage behavior and determine the copyright liability.

[0066] Step 3: The smart contract deducts the deposit as compensation to the data owner.

[0067] Compared with the prior art, the present invention has the following advantages:

[0068] First, use pre-trained CNN to extract image features, and save the feature vectors extracted by each image owner with the same feature space and high-level feature description. Then, use asymmetric transfer hashing to convert high-dimensional feature vectors into low-dimensional binary codes to reduce storage and communication costs. Finally, use non-ASPE algorithms to encrypt image features and generate secure encrypted searchable indexes to achieve privacy protection for feature operations.

[0069] Second, based on the powerful computing and storage capabilities of cloud computing and the decentralized and tamper-proof characteristics of blockchain, the cloud platform is used to store large-capacity data such as encrypted remote sensing images and perform retrieval calculations. The blockchain stores lightweight image hashes to verify the authenticity and integrity of cloud computing, and designs a data interaction mechanism with Merkle trees as the proof medium, which effectively solves the low execution efficiency of image retrieval algorithms in smart contract environments and the privacy protection issues they face.

[0070] Third, the image owner encrypts the image and marks the pixels based on the pixel prediction error and Huffman coding, reserving space for information embedding. The cloud server embeds the digital fingerprint into the reserved space of the marked pixels by bit replacement for image tracking and tracing. This method has the characteristics of separable image encryption and fingerprint embedding, and independent fingerprint extraction and image restoration, which improves the flexibility and practicality of traceable image retrieval.

[0071] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0072] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A remote sensing image trusted transaction method based on digital fingerprint tracing on blockchain, including transaction preprocessing, image transaction and data verification: The transaction preprocessing steps are as follows: S 1: The data owner and the cloud server register the blockchain account address and join the blockchain network; S2: The data owner uses the CNN model to extract image features, and reduces the dimension based on asymmetric transfer hashing, and then performs asymmetric dot product to maintain encryption and generate an encrypted searchable index; S3: The data owner performs prediction error calculation, image encryption, and Huffman coding marking in sequence to generate an encrypted image with reserved embedding space; S4: The data owner stores lightweight information such as encrypted image hash values ​​on the blockchain and transmits encrypted image information such as encrypted searchable index, encrypted image, and encrypted image hash values ​​to the cloud server; The steps of image transaction are as follows: S5: The data buyer sends a retrieval request to the blockchain, and after identity verification is passed, it is registered as a node on the blockchain; S6: The data buyer uses the CNN model to extract image features, and then performs asymmetric transfer hashing to reduce the dimension, and then performs asymmetric dot product to maintain encryption, generate a query trapdoor, and then initiate a transaction through a smart contract. The transaction information includes the query trapdoor and the prepaid principal; S7: The cloud server retrieves the secure searchable index according to the query trapdoor on the smart contract, returns the similarity ranking of all feature vectors, generates a Merkle tree based on all images according to the similarity ranking, and then uploads the root hash and similarity ranking of the Merkle tree to the blockchain, and provides a Top-k search image for data buyers to select the data to be traded; S8: After the data buyer determines the data to be traded, the cloud server first obtains the Huffman coding rules stored in the reference pixel position of the encrypted image, and then locates the embeddable pixels in the image according to the characteristics of its different-letter codeword, and embeds the digital fingerprint into it by bit replacement, and finally returns the transaction image containing the digital fingerprint to the data buyer; The steps for data verification are as follows: S9: The credibility verification of the retrieval results is done by the blockchain generating a Merkle tree based on the hash value of the Top-k retrieval image and the hash value of the encrypted image stored on the chain according to the similarity sorting, and then verifying the authenticity, integrity and correctness of the similarity sorting of the Top-k retrieval image based on the Merkle tree root hash and similarity sorting uploaded to the cloud server; S10: Data leakage traceability verification is performed by blockchain to extract and obtain Huffman code at the reference pixel position of the encrypted image, locate excess embeddable pixels, and extract the embedded digital fingerprint. The extracted digital fingerprint is then compared with the digital fingerprint stored in the smart contract to verify the data leakage behavior and determine the copyright liability.

2. According to claim 1, a remote sensing image trusted transaction method based on digital fingerprint tracing on blockchain is characterized in that: In steps S2 to S3, the data owner first constructs an encrypted searchable index and then generates an encrypted image with reserved embedding space.

3. According to claim 1, a remote sensing image trusted transaction method based on digital fingerprint tracing on blockchain is characterized in that: In S6, the prepaid principal includes the transaction payment amount and the deposit. When the data buyer's leakage behavior is confirmed, the smart contract deducts the deposit as compensation for the data owner.

4. According to claim 1, a remote sensing image trusted transaction method based on digital fingerprint tracing on blockchain is characterized in that: In S9, the credibility verification of the retrieval result is performed before the digital fingerprint is embedded.

Images