Copyright Protection and Trusted Transactions for 3D Models Based on Smart Contracts and Zero-Watermarking

Abstract

With the widespread application of 3D models derived from oblique photography, the need for copyright protection and trusted transactions has risen significantly. Traditional transactions often depend on third parties, making it difficult to balance copyright protection with transaction credibility and to safeguard the rights and interests of both parties. To address these challenges, this paper proposes a novel trusted-transaction scheme that integrates smart contracts with zero-watermarking technology. Firstly, the skewness of the oblique-photography 3D model data is employed to construct a zero-watermark identifier, which is stored in the InterPlanetary File System (IPFS) alongside encrypted data for trading. Secondly, smart contracts are designed and deployed. Lightweight information, such as IPFS data addresses, is uploaded to the blockchain by invoking these contracts, and transactions are conducted accordingly. Finally, the blockchain system automatically records the transaction process and results on-chain, providing verifiable transaction evidence. The experimental results show that the proposed zero-watermarking algorithm resists common attacks. The trusted-transaction framework not only ensures the traceability and trustworthiness of the entire transaction process but also safeguards the rights of both parties. This approach effectively protects copyright while ensuring the reliability of the transactions.

1. Introduction

In recent years, the rapid evolution of geographic information technology has propelled urban development into a new phase of intelligent transformation, shifting from conventional digitalization towards smart city ecosystems. Within this context, traditional two-dimensional cartographic representations have demonstrated significant limitations in addressing the multidimensional demands of contemporary socioeconomic development. As a specialized form of three-dimensional geospatial data, oblique photogrammetric 3D models have emerged as infrastructure components critical for the construction of digital-twin systems in virtual environments. These datasets exhibit three distinctive characteristics: high commercial valuation, inherent information sensitivity, and effortless reproducibility [1,2,3]. These attributes necessitate the establishment of comprehensive data-provenance mechanisms throughout transaction processes to protect stakeholder rights and facilitate secure data circulation.

Currently, the prevailing transaction method employs digital watermarking techniques to embed copyright markers within datasets and is supplemented by third-party archival systems for transaction verification [4,5,6]. However, this methodology presents two fundamental vulnerabilities: (1) excessive reliance on the credibility assumptions of transactional entities and (2) susceptibility to malicious manipulation or compromise by any party in the transaction chain [7]. This dual dependency creates systemic risks in maintaining transaction integrity, particularly regarding potential misconduct by data copyright holders (fraudulent attribution), intermediary malpractice (unauthorized redistribution), and purchaser violations (improper utilization). Consequently, the development of a decentralized trust mechanism that ensures transactional equity, prevents data misuse, and maintains auditability has become a critical research imperative in the domain of oblique photogrammetric 3D-model transactions [8].

Blockchain technology is characterized by its decentralization, traceability, tamper-proof nature, and transparency, effectively addressing the shortcomings of traditional transaction methods. By integrating blockchain technology into data transactions, credibility concerns can be significantly mitigated. A key feature of blockchain technology, smart contracts offer programmability, automated execution, and trustless operations, further enhancing the security and reliability of blockchain-enabled transactions [9,10].

Recent studies have demonstrated the potential of blockchain technology in enabling trusted data transactions. Blockchain networks can facilitate secure payments between data buyers and sellers, with the integration of Trusted Execution Environments (TEE) playing a crucial role in establishing a fair and protected transaction environment [11]. In the domain of medical data security, a blockchain-based MCPS network model has been proposed that would analyze medical data storage architecture and utilize bilinear mapping in order to address authentication-related security threats, thereby mitigating trust-related concerns associated with third-party intermediaries [12]. Additionally, blockchain technology has been leveraged to develop privacy protection systems that safeguard sensitive personal information during data sharing, while enhancing efficiency through decentralized mechanisms [13]. The application of blockchain technology for data sharing and retrieval has also been explored, enabling direct interactions between data owners and users while ensuring the automatic retention of transaction records on the blockchain [14]. While these studies provide effective solutions for secure data transactions, challenges persist, particularly in ensuring robust traceability and verification of data ownership throughout the dissemination process.

Recent research has explored the integration of blockchain technology with digital watermarking technology, aiming to enhance trust in data transactions. By embedding copyright information into carrier data and recording it on the blockchain alongside transactional details, this approach ensures the traceability of data sources and verification of ownership. Some studies have focused on the development of blockchain-based digital watermarking algorithms to strengthen copyright protection. For example, a zero-watermark registration scheme integrating Hyperledger Fabric and the InterPlanetary File System (IPFS) has been proposed to securely register watermarks and their associated parameters while enabling lossless copyright protection for high-resolution remote sensing images [15].

Other research has introduced image copyright-protection frameworks that combine blockchain and zero-watermarking technology. These frameworks incorporate keyword search functionalities for images stored in the system, using Ethereum smart contracts while mitigating blockchain data expansion issues through IPFS, thereby achieving copyright traceability [16]. Furthermore, a multi-type decentralized data transaction scheme has been proposed that would integrate smart contracts with digital watermarking technology. By utilizing zero-watermark technology, the scheme ensures copyright protection during data sharing, while transaction-specific algorithms and protocols implemented through smart contracts guarantee fairness in data transactions [17].

Similarly, blockchain technology has also been applied to transactions involving geographic data, where a notarization and copyright-protection model was developed by combining zero-watermarking technology, IPFS, and smart-contract technology [18]. This model eliminates reliance on third-party intellectual property rights (IPR) agencies by establishing a novel zero-watermark registration mechanism. By leveraging smart contracts, transaction information is permanently recorded on the blockchain, alongside the zero-watermark registration, ensuring timestamp-based authentication of the registration process [19]. While these studies have effectively contributed to data copyright protection, they predominantly focus on safeguarding the rights and interests of copyright holders. However, they often overlook the legitimate rights and interests of data buyers, leading to shortcomings in transaction fairness and overall credibility.

To address these limitations, this paper proposes and develops a trusted-transaction attestation system for oblique-photography 3D models by leveraging the advantages of blockchain technology and digital watermarking and integrating smart contracts with zero-watermark technology. The system utilizes the tamper-proof nature of blockchain technology and the copyright-protection capabilities of zero-watermarking to achieve efficient and reliable transaction attestation, while incorporating the Web3.js library to enable seamless interaction between the blockchain and smart contracts, ensuring that all transaction information is securely deployed and permanently stored on the Ethereum blockchain and remains immutable and fully traceable. The main contributions of this paper are as follows:

(1)

A robust zero-watermarking algorithm for oblique-photography 3D models is proposed, leveraging the vertical skewness of oblique 3D models to construct zero-watermark information. This approach ensures data integrity while significantly enhancing robustness against geometric attacks, providing a reliable mechanism for copyright protection and secure transactions involving high-precision 3D models.

(2)

A blockchain-based trusted-transaction scheme is introduced, utilizing smart contracts to automate key processes such as identity verification and transaction record storage. By ensuring the legitimacy of transactions and preventing data tampering or duplication, this scheme enhances the efficiency and transparency of transaction management while safeguarding data security and the credibility of all parties involved.

The remainder of this paper is structured as follows: Section 2 reviews the related work. Section 3 details the implementation of the proposed approach. Section 4 presents the experimental results along with an in-depth analysis. Finally, Section 5 provides concluding remarks.

2. Basic Ideas and Preliminaries

2.1. Basic Ideas

This paper integrates blockchain technology into the transaction process of oblique-photography 3D models and proposes a trusted-transaction attestation scheme that combines smart contracts with zero-watermarking. The proposed approach effectively addresses key challenges such as mutual trust, protection of stakeholders’ rights, and transaction security in the exchange of oblique-photography 3D model data, providing a robust technical foundation for trusted 3D-model transactions.

The scheme consists of three main phases. First, the oblique-photography 3D model data undergoes processing to generate watermark information, for which a zero-watermark is constructed using an XOR operation with an encrypted binary watermark [20]. Second, the zero-watermark identifier, transaction information, and data metadata are collectively stored on the blockchain. Leveraging the blockchain’s immutability and decentralized nature, this ensures the integrity and security of transaction-related information. Finally, both the copyright holder and the buyer can verify the authenticity and integrity of the model data by accessing the blockchain-stored information, thereby safeguarding their respective rights and interests.

By ensuring security, fairness, and mutual trust, this scheme effectively mitigates risks associated with copyright disputes and fraudulent transactions. The transaction attestation process, which integrates blockchain technology and digital watermarking, is illustrated in Figure 1.

2.2. Smart Contract

A smart contract is a self-executing program based on blockchain technology that autonomously enforces predefined rules and conditions. The structure of the smart-contract model is illustrated in Figure 2. Its content and execution logic are embedded in code and deployed on the blockchain. When external data is fed into the contract and specific conditions are met, the smart contract automatically triggers the corresponding actions, without requiring third-party intervention [21].

Once a transaction is verified as legitimate, it is packaged into a new data block, which is subsequently authenticated through the blockchain’s consensus mechanism and appended to the main chain. Throughout the entire process, the blockchain continuously monitors the contract’s status and value in real time, ensuring transparency, security, and immutability [10,22].

Since smart contracts operate within a decentralized blockchain network, they ensure that transactions are transparent, tamper-proof, and fully traceable, significantly enhancing both security and efficiency. By eliminating the need for intermediaries, smart contracts reduce the risk of fraud and disputes, thereby fostering mutual trust between the transacting parties [23].

2.3. IPFS

IPFS is a decentralized, distributed file storage and sharing protocol designed to provide efficient, reliable, and secure data management across a global network of nodes [16]. Unlike traditional file storage systems that rely on location-based addressing (e.g., URLs in HTTP), IPFS employs content-based addressing, which enhances data integrity and availability.

In conventional systems, file storage relies on centralized servers or data centers, which can lead to bandwidth bottlenecks, storage limitations, and risks associated with single points of failure and centralization. In contrast, the IPFS addresses these challenges by dividing files into smaller blocks, each uniquely identified by a cryptographic hash. Both the file’s content and the file’s metadata are hashed, ensuring the immutability and verifiability of the stored data [11,24].

Furthermore, IPFS distributes file content across multiple nodes. When a user requests a file, the system searches for the closest available node containing the requested data, using the file’s hash value. This content-addressing mechanism ensures data consistency and resilience, even when files are duplicated and stored across different locations in the network.

2.4. Skewness Measure

The statistical analysis of data distribution is crucial for identifying patterns and understanding the characteristics of a dataset. Skewness analysis is a widely used statistical method that measures the asymmetry of data distribution, helping to determine whether the data is symmetrically distributed [20,25].

In a perfectly symmetric distribution, the skewness value is zero, meaning the mean, mode, and median are equal. However, in an asymmetric distribution, the skewness can be either positive or negative. A positive skewness indicates that most of the data points are concentrated on the left side of the mean, with a longer tail extending to the right. Conversely, a negative skewness suggests that the data is concentrated on the right side of the mean, with a longer tail extending to the left. Figure 3 illustrates the graphical representation of positive and negative skewness in data distributions.

The calculation of the skewness value $S$ is shown in (1):

$$S={\displaystyle \frac{n}{(n-1)(n-2)}}\sum _{i=1}^n{\left({\displaystyle \frac{z_i-\mu }{\sigma }}\right)}^3$$

(1)

where $z_i(i=\mathrm{1,\; 2},\dots , n)$ represents the sample data of the group; $\mu$ denotes the mean value of the data; $\sigma$ indicates the standard deviation of the data; and $n$ refers to the number of data points in the group.

3. Proposed Method

3.1. Zero-Watermark Method

Leveraging the characteristic stability of oblique-photography 3D model data in the vertical direction [26,27], this paper proposes a zero-watermarking algorithm based on skewness analysis. The core idea of the algorithm is as follows:

First, Singular Value Decomposition (SVD) is performed on the 1-neighborhood points of each vertex in the model to extract the singular value vector corresponding to the vertex. Then, the angle between the last component vector of the singular value vector and the Z-axis normal vector is calculated. Feature points are selected based on the magnitude of this angle.

Next, Euclidean distance is employed to group the remaining vertices into clusters. Finally, the skewness of the Z-coordinates within each group is calculated, and based on whether the skewness is positive or negative, the watermark information is constructed, resulting in the generation of the zero-watermark identifier [28].

3.1.1. Data Process

(1)

Feature point extraction

In this paper, the triangular mesh model of the oblique-photography 3D model is represented as $G=(P,T)$, where $P=\left\{P_1,P_1,\dots ,P_n\right\}{(P_i=(x_i,y_i,z_i)}^T\in R^3,i=\mathrm{1,2},\dots ,N)$ is the set of vertices of the triangular mesh model, $T$ represents the set of triangular mesh containing the vertex index values, and $P_i=(x_i,y_i,z_i)$ denotes the coordinates of the $i$-th vertex.

Step 1: Based on the index set $T$ of the model triangular mesh, the coordinates of the 1-neighborhood points $\{P_1,P_2,\dots ,P_k\}$ corresponding to each vertex $P_i$ are identified. The center coordinates $M_i$ of the 1-neighborhood points of vertex $P_i$ are then calculated using Formula (2).

$$M_i={\displaystyle \frac{1}{k}}\sum _{i=0}^k{P}_i$$

(2)

where $k$ denotes the number of 1-neighborhood vertices.

Step 2: Using (3), the decentralized coordinates ${\tilde{P}}_i$ of the 1-neighborhood points for each vertex are calculated. The decentralized neighborhood matrix is then constructed using (4).

$${\tilde{P}}_i=P_i-M_i$$

(3)

$$Q=\{{\tilde{P}}_1,{\tilde{P}}_2,\dots ,{\tilde{P}}_k\}$$

(4)

where $Q$ represents the neighborhood matrix composed of the decentralized 1-neighborhood matrix.

Step 3: Using (5), SVD is performed on $Q$ to obtain the singular values and the singular vector $V$. Since $V$ is invariant to scaling, it is used as a key to identify stable feature points. SVD is a technique for decomposing a matrix into three component matrices, as shown below.

$$Q=U\mathsf{\Sigma }{V}^T={[u_1 u_2\dots u_m]}_{(m\times m)}{\Sigma }_{(m\times n)}{\left[\begin{array}{c}{v}_1^T\\ v_2^T\\ ⋮\\ v_n^T\end{array}\right]}_{(n\times n)}$$

(5)

where $U$ is the left singular value matrix of $Q$; $\mathsf{\Sigma }=\left\{{\lambda }_1,{\lambda }_2,\dots ,{\lambda }_m\right\}$ is the singular value matrix, where $\lambda$satisfy ${\lambda }_1\ge {\lambda }_2\ge \dots \ge {\lambda }_m\ge 0$; and $V^T$ is the right singular value vector matrix of $Q$.

When the matrix $Q$ is scaled, $Q^{\prime }=\alpha Q$. SVD decomposition is then performed on the scaled matrix $Q^{\prime }$, using (6):

$$Q^{\prime }=\alpha Q=\alpha U\mathsf{\Sigma }{V}^T=U\alpha \mathsf{\Sigma }{V}^T$$

(6)

where $\alpha$ is the scaling factor. When the singular value matrix $Q$ is scaled, only the singular values $\Sigma$ are affected, while $U$ and $V$ remain unchanged. As a result, the singular value vector $V$ is resistant to scaling transformation attacks.

Step 4: Take the last component ${\overrightarrow{V}}_e$ in the singular value vector $V$, the direction of which must align with the normal vector direction of the vertex. Therefore, this paper uses (7) to calculate the difference between the vertex and the center point of its domain, obtaining the vector pointing from the center to the vertex. This vector’s direction is then considered the direction of the normal vector. Next, (8) is used to determine whether ${\overrightarrow{V}}_e$ aligns with the normal vector ${\overrightarrow{P}}_i$ of the vertex, and (9) is employed to correct the direction of ${\overrightarrow{V}}_e$.

$${\overrightarrow{P}}_i=P_i-M_i$$

(7)

$$value={\overrightarrow{V}}_e·{\overrightarrow{P}}_i=\left|{\overrightarrow{V}}_e\right|\ast \left|{\overrightarrow{P}}_i\right|\ast \mathrm{c}\mathrm{o}\mathrm{s}\left(\phi \right)$$

(8)

where ${\overrightarrow{P}}_i$ represents the vector from point $M_i$ to vertex $P_i$; the $value$ is a variable indicating whether the directions of the two vectors are consistent.

$${\overrightarrow{V}}_{e\_n}=\left\{\begin{array}{c}{\overrightarrow{V}}_e if value>0\\ {-\overrightarrow{V}}_e if value<0\end{array}\right.$$

(9)

where ${\overrightarrow{V}}_{e\_n}=[v_1,v_2,v_3]$ represents the corrected singular value vector.

Step 5: According to (10), the dot product of the vector ${\overrightarrow{V}}_{e\_n}$ and the unit normal vector $\overrightarrow{Z}=\left[\mathrm{0,0},1\right]$ of the Z-axis is calculated to obtain $dp$. During the watermark construction process, $dp$ is used to determine the index of the watermark position. Furthermore, (11) is used to compute the angle between the two vectors, resulting in$\theta$.

$$dp=\left|{\overrightarrow{V}}_{e\_n}\cdot \overrightarrow{Z}\right|=\left|[v_1,v_2,v_3]\cdot \left[\mathrm{0,0},1\right]\right|$$

(10)

$$\theta =\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{c}\mathrm{o}\mathrm{s}\left({\displaystyle \frac{{\overrightarrow{V}}_{e\_n}\ast \overrightarrow{Z}}{\left|{\overrightarrow{V}}_{e\_n}\right|\ast \left|\overrightarrow{Z}\right|}}\right)$$

(11)

In the formula, $\left|{\overrightarrow{V}}_{e\_n}\right|$ and $\left|\overrightarrow{Z}\right|$ denote the magnitudes of the vectors ${\overrightarrow{V}}_{e\_n}$ and $\overrightarrow{Z}$, respectively.

Step 6: Set an angle threshold $an\_th$, and select the vertices with an angle $\theta$ greater than the threshold as feature points.

Step 7: Following the steps above, all vertex coordinates are traversed to obtain the feature point sequence $Fe$. Since duplicate points may exist in the vertex coordinates of the model, repeated feature points in $Fe$ are removed using (12).

$$Fe=unique\left(Fe\right)$$

(12)

In this formula, $unique\left(\right)$ is a function used to remove duplicate point coordinates.

(2)

Grouping of vertices

Based on the selected stable feature points and the geometric spatial information of the vertices, the Euclidean distances between each vertex and all feature points are calculated using (13).

$$d\left({Fe}_j,P_i\right)=\sqrt{{(x_i-x_j)}^2+{(y_i-y_j)}^2+{(z_i-z_j)}^2}$$

(13)

where $d$ represents the Euclidean distance; ${Fe}_j=(x_j,y_j,z_j)$ represents the coordinates of the feature point.

Based on the calculated distances, the vertices are grouped according to the minimum distance principle, ensuring that each vertex is assigned to the feature point associated with the shortest distance.

(3)

Vertex Selection

Since data skewness characterizes the asymmetry of a data distribution, including a large number of points located near the geometric center in its calculation leads to a highly concentrated distribution. These central points dominate the dataset, causing the computed skewness to approach zero and thus fail to effectively capture whether the overall shape exhibits a “one-sided” bias. Furthermore, the radii of these central points are relatively small, making them less sensitive to variations in the overall shape and inadequate for the purpose of reflecting shape changes through skewness measurements. Therefore, prior to calculating skewness, it is essential to remove the points in the central region and use the remaining coordinates for the calculation. In this paper, the center point $M_c=(x_c,y_c,z_c)$ of each set of points is determined using (2). Formula (13) is then applied to calculate the distance from all vertices to the center point, yielding the radius $r$. Formula (14) is used to sort the values of $r$, and the points with larger $r$ are selected for the skewness calculation.

$$S{r}_b=sort(r_i,asc)$$

(14)

In the formula, $sort\left(\right)$ represents the sorting function, $r_b$ represents the larger set, and $asc$ denotes ascending order.

3.1.2. Generation Process of Zero Watermark

The process of constructing the zero watermark is illustrated in Figure 4, and the specific embedding steps are as follows.

Step 1: The watermark image $W$ is encrypted using Arnold scrambling based on the key, resulting in the encrypted watermark image $W^{\prime }= \left\{\omega [i,j]\right\}$, where $i,j\in \left[0,H\right]$ and $H$represents the length (width) of the watermark image.

To enhance the security of the watermark, the watermark information $W$ must be encrypted before generating the zero-watermark identifier. The Arnold transform scrambles the image by rearranging the pixel coordinates, effectively achieving a form of encryption. Formula (15) defines the Arnold transform.

$$\left[\begin{array}{c}{a}^{\prime }\\ b^{\prime }\end{array}\right]=\left[\begin{array}{cc}1& {key}_m\times {key}_n\\ {key}_n& 1+{ key}_m\times {key}_n\end{array}\right] \left[\begin{array}{c}a\\ b\end{array}\right](mod H)$$

(15)

where $(a,b)$ represents the pixel position before encryption, $(a^{\prime },b^{\prime })$ represents the pixel position after encryption, and ${key}_m$ and ${key}_n\in N^{+}$; ${key}_m$, ${key}_n$ and $H$ are collectively referred to as the key K.

Step 2: The vertex coordinates $P$ of the oblique-photography 3D model are read, and the feature points of the 3D model are extracted using the above-mentioned method, resulting in a feature point sequence denoted $Fe$.

Step 3: The Euclidean distances between all vertices $P$ and feature points $Fe$ in the 3D model are calculated according to (13).

Step 4: Based on the minimum distance principle, the vertices $P$ are grouped to obtain ${Fe}_p=(f_1,f_2,\dots ,f_{fe})$, where $f_1=\left({fe}_1,p_i,\dots \right),f_2=\left({fe}_2,p_i,\dots \right),\dots ,f_{fe}=\left({fe}_{fe},p_i,\dots \right)$, with $i\in \left\{\mathrm{1,2},,k\right\},j\in \left\{\mathrm{1,2},,g\right\}$; $fe$ represents the number of feature points.

Step 5: According to (2), the coordinates $M_c$ of the center points for all vertices in each group are calculated sequentially, and the distance $r$ between each vertex and the corresponding center point is determined.

Step 6: The values of $r$ are then sorted, and the skewness $S_i$ of the group is computed using (1), based on the Z values of the coordinates of the 2/3 of the points with the largest distances.

Step 7: The dot product $dp$ of the above vector is then scaled by a factor of ${10}^u$, and the index value $index$ is calculated using (16).

$$index=\left[dp\times {10}^u\right] mod l$$

(16)

In the formula, the symbol [] indicates rounding. The $l$ represents the size of the one-dimensional watermark sequence.

Step 8: Due to the one-to-many relationship between the number of feature points and the corresponding watermark positions, the voting principle is employed to construct the watermark information, with the result recorded as $F$. The formula for this is shown in (17).

$$F=\left\{\begin{array}{c}F\left[index\right]+1, S_i>0\\ F\left[index\right]-1, S_i\le 0\end{array}\right.$$

(17)

In this formula, $F\left[index\right]$ represents the index value corresponding to the$index$ bit.

Step 9: After the watermark information is constructed, $F$ is binarized using (18) to obtain the one-dimensional watermark information $F^{\prime }$.

$$F^{\prime }=\left\{\begin{array}{c}1, F\left(i\right)>0\\ 0, F\left(i\right)\le 0\end{array}\right. i\in \left\{\mathrm{1,2},\dots ,l\right\}$$

(18)

where $i$ is the index of the one-dimensional watermark; $F\left(i\right)$ is the value corresponding to the $i$-th bit of the one-dimensional sequence.

Step 10: The extracted one-dimensional feature information $F^{\prime }$ is reshaped into a two-dimensional sequence to obtain the final watermark information $F″$.

Step 11: The $F″$ is bitwise XOR with the encrypted watermark image $W^{\prime }$ according to (19), resulting in the final zero-watermark sequence $W″=(w_{\left(\mathrm{0,0}\right)},w_{\left(\mathrm{0,1}\right)},\dots ,w_{(H,H)})$.

$$W″=F″\oplus W^{\prime }$$

(19)

3.1.3. Zero-Watermark Detection Process

Watermark detection is the reverse of the process of watermark embedding. By comparing the original watermark with the detected watermark, copyright protection is achieved. The specific steps of the detection process are shown in Figure 5.

Step 1: Read the oblique-photography 3D model data to be tested.

Step 2: Repeat Steps 2 to 10 from the generation process of the zero-watermark identifier.

Step 3: Perform an XOR operation between the binary watermark information of the oblique-photography 3D model data obtained in Step 2 and the zero-watermark image stored on the IPFS to generate scrambled copyright information.

Step 4: Perform Arnold descrambling on the detected encrypted copyright information to recover the final watermark image.

3.2. Trusted-Transaction Scheme

3.2.1. Smart-Contract Platform Selection

The development and design of smart contracts require a fully functional smart-contract platform. Remix IDE, as an efficient integrated development environment (IDE), offers comprehensive support for writing, deploying, and debugging smart contracts. It is fully compatible with the Ethereum Virtual Machine (EVM) and seamlessly supports the Solidity smart-contract language. The platform provides a rich set of features, as well as the ability to simulate and test the execution of smart contracts in real time. This functionality helps developers quickly identify potential issues and optimize contract performance [29]. In response to the requirements for trusted transactions relating to oblique-photography 3D models, this research project selected Remix IDE as the development tool to be used to write relevant smart contracts in Solidity.

3.2.2. Smart-Contract Design

In the trusted-transaction scheme, the design of smart contracts is central to ensuring the integrity of trusted transactions. During the design phase of smart contracts, it is essential to incorporate the addresses of authorized users, ensuring that only these users can call the contract, and thus safeguarding the security and privacy of the data. In this paper, to facilitate trusted data transactions and protect the rights and interests of both parties, two smart contracts are designed: the model information notarization contract (Contract 1) and the data transaction information contract (Contract 2). Contract 1 is responsible for data notarization, while Contract 2 handles the execution of the transaction; together, they form a complete and secure transaction process. The pseudocode for the contracts is as follows.

Contract 1: This contract can only be deployed by the copyright owner (DO) of the data. Its primary function is to store essential data information, addresses, transaction conditions, and other relevant details. When creating the contract, six key variables are defined (five of which are collectively referred to as the data summary):

(1)

owner: An Ethereum account address representing the copyright owner’s address.

(2)

dataIpfsAddress: The storage address of the encrypted data on the IPFS network.

(3)

watermarkIpfsAddress: The storage address of the zero-watermark identifier on IPFS.

(4)

MerkleTree: By calculating the Merkle tree root after dividing the data into blocks, this structure enables detection of the approximate location of any tampering and ensures data integrity [30].

(5)

dataId: The hash value of the data, generated using Hash256. This value is stored on the blockchain as the unique identifier of the data, preventing the upload of duplicate or identical data and ensuring its uniqueness [31].

(6)

Conditions: The transaction fee set by the copyright owner for accessing or using the data.

The copyright holder invokes the setDataMessage function to input the data summary into the contract and deploy it to the blockchain.

Contract 1: Oblique photography 3D model information notarization contract

Input: string dataIpfsAddress, MerkleTree, watermarkIpfsAddres, dataId, conditions; Output: address of CopyrightDeposit; Contract CopyrightDeposit{   address public owner;   string dataIpfsAddress, MerkleTree, watermarkIpfsAddres, dataId, conditions;   struct ContractOwner{   string organization, name;   }   mapping(address => ContractOwner) public contractOwners;   address constant ALLOWED_DEPLOYER =“address”;    …… function setDataMessage(string memory _dataIpfsAddress, string memory      _watermarkIpfsAddress, string memory _dataId, string memory _MerkleTree)      public returns (address) { require(bytes(contractowner[msg.sender].name).length != 0 &&        bytes(contractOwners[msg.sender].tissue).length != 0, “Please execute        Owner_message first to set your details.”);   require(owner==msg.sender, “You do not have access.”);   dataIpfsAddress = _dataIpfsAddress;   watermarkIpfsAddress = _watermarkIpfsAddress;   MerkleTree = _MerkleTree;   dataId = _dataId;   owner = msg.sender;   return owner; }

Contract 2: This contract is also deployed by the DO. After deployment, the buyer (DB) initiates a transaction by calling this contract. The contract contains four key variables:

(1)

User: A structure type used to store the basic information of the buyer.

(2)

userAddress: Stores the Ethereum address of the current buyer, which is used for identity verification during subsequent transactions. Only authorized users are allowed to conduct transactions.

(3)

receiver: Stores the recipient’s address for the contract, enabling the withdrawal of the transaction amount after the transaction is completed.

(4)

copyrightDepositContract: An instantiation of the Contract 1 contract within Contract 2. This variable facilitates interaction between the two contracts, allowing access to the stored data information.

Once the buyer registers the necessary information and transaction conditions via this contract, the contract automatically calls the getDataAddresses function to retrieve the storage address, zero-watermark information, and other relevant data stored in Contract 1. Upon completion of the transaction, the data copyright holder receives the corresponding transaction fee.

Contract 2: Data transaction information contract

Input: address of CopyrightDeposit, conditions; Output: dataIpfsAddress, MerkleTree, watermarkIpfsAddres, dataId; Contract DataTransaction {   address public userAddress, receiver;   CopyrightDeposit public copyrightDepositContract;   struct User {   string organization, name;   }   ……   function tip () public payable {   require (msg.value >= conditions, “you should send to use this function”);   require (userAddress == msg.sender, “You do not have permission”);   }   function getDataAddresses () public view returns(string memory dataIpfsAddress,     string memory MerkleTree, string memory dataId, string memory     watermarkIpfsAddress) {   require(userAddress==msg.sender, “You don’t have permission.”);      return(copyrightDepositContract.dataIpfsAddress(), copyrightDeposit      Contract.MerkleTree(), copyrightDepositContract.dataId(), copyright      DepositContract.watermarkIpfsAddress());   }

3.3. Trusted-Transaction Attestation System Architecture

In the system framework proposed in this paper, there are two main entities: DO and DB. The system architecture is shown in Figure 6, where the smart contract is deployed on the Ethereum blockchain.

① DB obtains the data file of interest from the blockchain.

② DB submits identity information to DO and requests a transaction.

③ DO verifies DB’s identity information. Once verified, DO encrypts the original data using DB’s public key, and uploads the encrypted data along with the zero-watermark information file to IPFS.

④ IPFS generates a unique storage address (hash value) and returns it to DO to identify the stored file.

⑤ DO uploads the copyright holder information, buyer information, and data summary to smart Contract 1.

⑥ The data summary and other relevant information are deployed to a blockchain block via smart Contract 1.

⑦ The blockchain returns the address of smart Contract 1 to DO.

⑧ DO sends the address of smart Contract 1 to DB.

⑨ DB calls Contract 2 to initiate the transaction using the address, payment amount, and buyer information of Contract 1.

⑩ Contract 2 verifies whether the buyer information and payment amount meet the required conditions. If valid, Contract 2 retrieves the data information from Contract 1.

⑪ Contract 2 returns the obtained data summary and other relevant information to DB.

⑫ The smart contract transfers the transaction fee to the copyright holder’s account address.

⑬ Contract 2 records the entire transaction process on the blockchain.

⑭ DB uploads the obtained data storage address to IPFS.

⑮ IPFS provides the encrypted data and zero-watermark information to DB, based on the storage address.

⑯ DO submits a query request to the blockchain.

⑰ The blockchain returns the queried information to DO.

⑱ DB submits a query request to the blockchain.

⑲ The blockchain returns the queried information to DB.

The DO is responsible for the initial upload and management of data. Additionally, it is responsible for verifying the identity of the data buyer, providing the contract address, extracting transaction proceeds, and querying transaction records.

The DB is responsible for initiating and executing transactions by retrieving the desired data files from the blockchain, submitting identity verification, and paying transaction fees. Upon transaction completion, it obtains the data storage address and downloads the required data along with the zero-watermark information.

4. Experiments and Analysis

This section presents the experimental verification and analysis of the proposed zero-watermarking method and the blockchain-based trusted-transaction scheme. To verify the practicality of this method, experiments on copyright protection were conducted using Python 3.6 in a Windows 11 environment. For the trusted-transaction scheme experiment, a contract design experiment was conducted using Solidity

4.1. Copyright Protection Experiment and Analysis

In the copyright protection and trusted-transaction schemes mentioned above, the zero-watermark identifier serves as a key copyright identifier. Its information must be preserved over the long term and be resilient against various potential malicious attacks. Therefore, the extracted zero-watermark identifier must exhibit both robustness and uniqueness, further enhancing the credibility and security of the trusted-transaction mechanism in the realm of copyright protection. The following section presents an experimental analysis of the copyright-protection function.

4.1.1. Experimental Data and Parameter Settings

To evaluate the effectiveness of the zero-watermark method, we obtained oblique-photography 3D models reflecting the realistic geometric features of Hong Kong’s urban and natural landscapes from the official website of the Government of the Hong Kong Special Administrative Region of the People’s Republic of China (https://www.gov.hk/ accessed on 4 November 2024). To enhance the representativeness and breadth of the experiment, we selected six sets of oblique-photography 3D models, denoted Models m_1, m_2, m_3, m_4, m_5, and m_6, respectively, with varying regional scope, data content, and scale. The data are illustrated in Figure 7. A 16 × 16 binary image is chosen as the watermark in this experiment. To assess the robustness of the method, the normalized correlation coefficient (NC) between the extracted watermark image and the original watermark image is calculated [32], as shown in (20). This paper sets the threshold for the similarity NC value at 0.75. If the detected NC value exceeds the threshold when suspicious data or model attacks occur, it indicates successful extraction of the copyright information; otherwise, copyright information authentication is considered to have failed. Furthermore, the experimental parameters are configured as follows: the values of the three parameters ${key}_n$, ${key}_m$, and $H$ of the key $K$ are set to 3, 2, and 16, respectively, while $\mu$ and $an\_th$ are set to 4 and 100, respectively.

$$NC={\displaystyle \frac{\sum _{i=1}^{l}w\left(i\right)w^{\prime }\left(i\right)}{\sqrt{\sum _{i=1}^l{w}^2\left(i\right)}\sqrt{\sum _{i=1}^L{w^{\prime }}^2\left(i\right)}}}$$

(20)

where $w\left(i\right)$ represents the value of the $i$-th bit of the original watermark; $w^{\prime }\left(i\right)$ represents the value of the $i$-th bit of the extracted watermark.

4.1.2. Uniqueness Analysis

The uniqueness of the zero-watermark algorithm ensures that the feature information extracted from each oblique-photography 3D model is distinct, thereby guaranteeing that each 3D-model data has unique feature information. To verify the uniqueness of the zero-watermark identifier, six different datasets were selected, and the corresponding zero-watermark information was extracted for each. The uniqueness was then cross-validated across these six sets of zero-watermark information. The results of this verification are presented in

Table 1. Results of uniqueness evaluation.

Model	m_1	m_2	m_3	m_4	m_5	m_6
m_1	1	0.6038	0.5968	0.6044	0.6062	0.5790
m_2	0.6038	1	0.6045	0.6206	0.6128	0.5963
m_3	0.5968	0.6045	1	0.6110	0.5965	0.6046
m_4	0.6044	0.6206	0.6110	1	0.5921	0.5829
m_5	0.6062	0.6128	0.5965	0.5921	1	0.6023
m_6	0.5790	0.5963	0.6046	0.5829	0.6023	1

.

As can be seen in Table 1, the NC values of the cross-validated watermark information all failed to reach the set NC threshold, and the valid watermark information could not be identified. The NC value of the watermark information verified by the same data was 1, indicating that the zero-watermark method in this paper has strong uniqueness and can well distinguish different oblique-photography 3D model data.

4.1.3. Robustness Analysis

This paper uses six different datasets for experimentation, applying various geometric and non-geometric attacks of different scales, including translation, rotation, scaling, cropping, simplification, reordering, and RST combination attacks. The results are then compared with those of the algorithms in [5,26] and [33] to assess and verify the security and robustness of the proposed algorithm.

(1)

Geometry attacks

In practical 3D-model applications, geometric transformation attacks alter the spatial structure without changing visual semantics, significantly affecting geometric features and challenging the robustness of watermarking and detection algorithms. Therefore, this paper uses dataset m_4 as an example and applies rotation, scaling, translation, and RST combination attacks of varying scales to the model data under detection. The experimental results are presented in Figure 8.

(2)

Cropping attack

Cropping attacks, common in oblique-photography 3D model processing, remove parts of the model, causing information loss and structural damage. This compromises watermark integrity and detection accuracy, making resistance to such attacks essential for robust watermark extraction and model security. To assess the resilience of the proposed method against cropping attacks, dataset m_4 is used as an example. Cropping attacks are performed on the 3D model at 5%, 15%, and 25% reductions, respectively. The corresponding NC values are recorded in

Table 2. Experimental results of cropping attack.

Cutting Strength	NC
	Proposed	[26]	[33]	[5]
5%	0.9823	0.9972	0.9993	0.9917
15%	0.9265	0.9509	0.9814	0.9628
25%	0.8321	0.9272	0.9460	0.9204

.

(3)

Simplified attack

This research project takes the dataset m_4 as an example, and conducts a simplified attack experiment, performing a 5% to 25% simplified attack on the model data to be tested. The experimental results of the simplification attack are presented in

Table 3. Experimental results of simplified attack.

Simplified Strength	NC
	Proposed	[26]	[33]	[5]
5%	0.9371	0.9043	0.7563	0.9812
15%	0.8203	0.7567	0.5483	0.9640
25%	0.7519	0.5312	0.5249	0.9162

.

(4)

Reordering attack

When the oblique-photography 3D model is transmitted, edited, or converted, the order of the coordinate indices in the 3D grid may change. Therefore, the proposed method must be robust against reordering attacks. The experimental results are presented in

Table 4. Experimental results of reordering attack.

Model	Proposed	[26]	[33]	[5]
m_1	1.0000	1.0000	1.0000	0.7382
m_2	1.0000	1.0000	1.0000	0.7132
m_3	1.0000	1.0000	1.0000	0.6890
m_4	1.0000	1.0000	1.0000	0.7804
m_5	1.0000	1.0000	1.0000	0.6890
m_6	1.0000	1.0000	1.0000	0.6516

.

The results from the above experiments demonstrate that the constructed zero-watermark algorithm exhibits high robustness against translation, rotation, scaling, and reordering attacks. It is also capable of effectively extracting watermark information, even under simplification and cropping attacks. Therefore, the zero-watermark algorithm proposed in this paper provides both high security and robustness.

4.1.4. Efficiency Assessment

To evaluate the practical applicability of the proposed watermarking scheme, its computational efficiency must be assessed, particularly in terms of watermark construction and detection time. Therefore, using Models m_1, m_2, and m_5 as examples, we designed an experiment to measure and compare the time consumption of the construction and detection processes. We compare the time required for the proposed scheme with the times required by three existing methods. The experimental results are summarized in Figure 9.

As shown in Figure 9, watermark construction and extraction times vary significantly across methods. The approach in [26] achieves the shortest runtime for all three datasets by directly embedding watermark information into the interpolation of adjacent vertices, resulting in a simple algorithm with low computational complexity and minimal overhead. The method in [33] maintains low runtime by grouping vertices to construct a spherical coordinate system and embedding the watermark into the vertex radius parameter, ensuring robustness, while preserving efficiency. In contrast, the method in [5] records the highest runtime, as it traverses all vertices to compute normal vectors and groups them by distance, greatly increasing complexity and computational burden.

The method proposed in this paper falls between these approaches in terms of runtime. This is because it selects stable feature points from each vertex’s 1-neighborhood and groups them according to inter-point distances, a process that requires additional computational time. However, this design ensures the stability of the selected feature points under geometric attacks and enhances robustness against cropping attacks through the grouping strategy. Therefore, although our method runs slightly longer than some lightweight algorithms, it offers substantial advantages in accuracy preservation and robustness, making it well suited for practical applications with high demands for data integrity and security.

4.2. Trusted-Transaction Experiment and Results

4.2.1. System Implementation

Based on the approach proposed in Section 3, this paper implements a trusted-transaction system for oblique-photography 3D models, utilizing the Ethereum blockchain as the underlying platform. The implementation process includes the writing and deployment of smart contracts as well as the development of the front-end system. Firstly, the smart contract is written using the Solidity language within the Remix IDE. Secondly, the smart contract is deployed to the Ethereum network. MetaMask (Ethereum wallet) serves as a bridge between the front end and the blockchain during the deployment process. When deploying the contract, MetaMask uses the user’s private key to digitally sign the transaction and pay the “gas” fees. Upon successful deployment, the contract code is submitted to the Ethereum blockchain, and the corresponding contract address is generated. Finally, the user interaction interface is developed using a front-end framework, enabling users to seamlessly conduct data transactions, view transaction records, and verify attestation information through the system.

As shown in Figure 10, the execution steps of the trusted-transaction system are as follows: In Figure 10a, the data copyright holder enters the copyright holder information, buyer information, and data summary into the transaction system. By invoking smart Contract 1, the copyright holder deploys this information to the blockchain and receives the stored block information along with the address of Contract 1. In Figure 10b, after the data buyer inputs the buyer’s information, the address of Contract 1, and the transaction amount into the system, the system automatically triggers smart Contract 2. Contract 2 then verifies the information that was provided. Once the verification is successful, Contract 2 returns the data summary and other relevant information to the buyer, completing the transaction. Finally, Contract 2 records the entire transaction process on the blockchain for transparency and traceability.

4.2.2. Security and Credibility Analysis

This scheme effectively integrates the decentralized data storage features of blockchain technology, the benefits of IPFS’s decentralized file storage, and the automated management capabilities of smart contracts to enable secure and reliable transactions of oblique-photography 3D model data. Compared to traditional data trading methods, this scheme offers enhanced security and credibility. The following sections provide a detailed analysis of the security and credibility of the solution.

Credibility: Throughout the transaction process, each transaction record is automatically stored and verified via smart contracts, ensuring a solid foundation of trust between the parties involved. This eliminates the reliance on trust in third parties typical of traditional models. Additionally, the designs of the smart contracts enforce strict access control, ensuring that only authorized users can access sensitive data and perform critical operations. This further strengthens the credibility of the transaction process for both parties.

Security: Leveraging the decentralized and tamper-proof characteristics of blockchain technology, data transaction records are permanently stored, making any attempts to modify, forge, or tamper with these records easily detectable and traceable. During the data storage process, the public key of the purchaser is used for encryption, ensuring that only the purchaser and the copyright holder have access to the data. Furthermore, by integrating IPFS’s distributed storage architecture, the risks of single-point failures and data breaches are effectively mitigated, thereby ensuring the security and integrity of the stored data.

4.2.3. Rights Protection for Both Parties to the Transaction

When the rights of the copyright holder are threatened, the system allows for the retrieval of the information associated with the original data, and the corresponding transaction evidence stored on the blockchain. This evidence includes data information, watermark information, timestamps, and more, thus effectively safeguarding the copyright holder’s rights. For example, in cases of data leakage or theft, the copyright holder can leverage zero-watermark technology to verify the authenticity of the data, confirming the ownership of the copyright. In instances where multiple parties claim ownership of the same data, the true copyright holder can be determined by the timestamp of the data stored on the blockchain. Furthermore, if the purchaser modifies the data and claims that the copyright holder has provided false data, and demands compensation, the copyright holder can compare the Merkle tree root stored in the transaction attestation with the values in the suspicious data. This comparison allows for the swift verification of the data’s integrity and the identification of the approximate location of any tampering, thereby ensuring the effective protection of the copyright holder’s legitimate rights and interests.

When the rights and interests of the purchaser are threatened, they can query the relevant information on the blockchain through the transaction address of the data to protect their rights. For example, after the transaction is completed, if the buyer suspects that the copyright holder has used false data in the transaction, they can generate a corresponding data hash value for the suspicious data and compare it with the hash value stored in the smart contract. If the hash values match, the data obtained by the purchaser is authentic; if they do not match, it indicates that the data is fraudulent, and the copyright holder has violated the terms of the agreement. Additionally, if the purchaser suspects that the watermark information provided by the copyright holder is false or maliciously manipulated, they can extract the watermark information stored in the data and XOR it with the zero-watermark identifier stored in the contract. If the obtained copyright information differs from what was provided by the copyright holder, it indicates that the watermark has been tampered with or forged, and the purchaser can file a complaint accordingly.

Based on the above analysis of the rights-protection procedures for both parties, as shown in Figure 11, this scheme can protect the rights of both the copyright holder and the buyer. Relying on blockchain technology and using smart contracts for automatic execution, the two parties can reach a consensus based on this system, achieve mutual trust without third-party supervision, and realize secure and reliable transactions relating to oblique-photography 3D model data under the dual protection of watermarks and on-chain transaction certificates.

4.2.4. Smart-Contract Execution Costs

Because smart contracts on Ethereum are costly to execute and immutable once deployed, it is crucial to experimentally evaluate gas consumption during both contract deployment and transaction execution. To ensure consistency and comparability, the gas-to-Ether conversion used in the experiment is based on market conditions as of August 6, 2025: 1 Ether ≈ 3575 USD, with a fixed gas price of 1 Gwei (1 Gwei = 10⁹ Wei = 10⁻⁹ Ether). Based on this, actual gas consumption was converted to Ether to provide a clearer understanding of the relative costs of each operation.

Table 5. Execution costs of deployment contracts and trading contracts.

Smart Contract	GAS Cost	Ether Cost	USD Cost
Contract 1	756,893	0.000756893	2.7058
Contract 2	886,305	0.000886305	3.1685

summarizes the gas costs associated with executing different functions on the simulated test chain.

The experimental results presented in Table 5 and

Table 6. Smart-contract function execution cost.

Smart-Contract Function	GAS Cost	Ether Cost	USD Cost
Copyright information	50,038	0.000050038	0.1788
Add data address	245,626	0.000245626	0.8781
Buyer information	69,836	0.000069836	0.2496
Withdraw	27,364	0.000027364	0.0978
Acquire data information	34,458	0.000034458	0.1231

indicate that the smart-contract deployment phase incurs relatively high costs, often exceeding USD 1. In contrast, the cost of invoking contract functions is significantly lower, with the vast majority of function executions costing less than USD 1. Overall, the expenses remain stable and within a reasonable range. This outcome highlights the economic advantage of the “deploy once, call many times” model. In practical scenarios, despite the substantial upfront cost of contract deployment, subsequent function calls incur relatively low costs, supporting long-term contract operation and frequent interactions, thereby enhancing the system’s overall economic efficiency and sustainability.

5. Conclusions

This paper addresses the challenge of ensuring trusted transactions for oblique-photography 3D models by proposing a novel trusted-transaction scheme that integrates smart contracts with zero-watermarking technology. This scheme designs a zero-watermarking method suitable for an oblique-photography 3D model; builds a data transaction attestation framework based on an IPFS distributed storage system, smart-contract technology, and Ethereum blockchain framework; and develops an oblique-photography three-dimensional model trusted-transaction system. The blockchain technology realizes the transparency, traceability, and tamper-proof aspects of the transaction information; the zero-watermark technology is used to protect the copyrights in the data during the transaction and use process, so that both parties can quickly confirm the copyright and use rights associated with the product; and the trusted-transaction framework created by the smart contracts not only effectively protects the legitimate rights and interests of both parties, but also verifies the integrity and authenticity of the data. The experimental results show that the zero-watermarking method proposed in this paper can effectively resist attacks such as those associated with geometry, cropping, simplification, and reordering, and has strong robustness. The constructed trusted-transaction attestation framework effectively solves the problem of lack of mutual trust and protects of the rights and interests of both parties in the transaction process relating to oblique-photography 3D model data, ensuring the security of the data transaction process. This scheme provides a new solution for trusted transactions and copyright protection relating to oblique-photography 3D model data, an approach which is of great value, and one that can promote safe circulation, trusted transactions, and wide applications of the data.

While this research has achieved the initial results described, it also has some limitations. For example, the current algorithm’s watermarking capacity is limited, and its robustness against complex attacks needs to be further enhanced. Furthermore, achieving an optimal balance between rapid verification of data copyright identification and overall system performance in trusted data trading systems remains an important direction for future research.