CoChat
Sign Up
Cited by 9Advanced Steganography and Watermarking TechniquesBlockchain Technology Applications and SecurityLaw, AI, and Intellectual Property

Digital Image Copyright Protection and Management Approach—Based on Artificial Intelligence and Blockchain Technology

Open access
Xu, Jikuan, Zhang, Jiamin, Wang, Junhan·2025·Journal of Theoretical and Applied Electronic Commerce Research
DOI 10.3390/jtaer20020076
Open this paper in CoChat — free accountOpen PDF

Abstract

The issue of image copyright infringement is prevalent in current e-commerce activities. Users employ methods such as image cropping, compression, and noise addition, making it difficult for traditional copyright detection technologies to identify and track infringements. This study proposes an image copyright registration, protection, and management method based on artificial intelligence and blockchain technology, aiming to address the current challenges of low accuracy in digital copyright infringement judgment, the vulnerability of image fingerprints stored on the chain to tampering, the complexity of encryption algorithms and key acquisition methods through contract calls, and the secure storage of image information during data circulation. The research combines artificial intelligence technology with traditional blockchain technology to overcome the inherent technical barriers of blockchain. It introduces an originality detection model based on deep learning technology after conducting both off-chain and on-chain detection of unidentified images, providing triple protection for digital image copyright infringement detection and enabling efficient active defense and passive evidence storage. Additionally, the study improves upon the traditional image perceptual hashing in blockchain, which has poor robustness, by adding chaotic encryption sequences to protect the image data on the chain, and its effectiveness has been verified through experiments. Ultimately, the research hopes to provide e-commerce entities with an effective and feasible digital copyright protection and management solution, safeguarding their intellectual property rights and fostering a legal and reasonable competitive environment in e-commerce.

AI snapshot

Written from the paper’s full text.

TL;DR

This paper proposes a digital image copyright protection and management approach combining improved perceptual hashing, blockchain storage with an MKD tree, chaotic encryption, and MobileNetV2-based originality detection. It reports that the fused triple-detection scheme achieves 93% accuracy, 89% recall, F1 0.91, and AUC 0.95 on its test set.

Key findings

  • The fusion scheme outperformed baselines: accuracy 0.93, recall 0.89, F1 0.91, AUC 0.95, versus traditional perceptual hashing (0.72 accuracy, 0.68 recall, 0.70 F1, 0.75 AUC) and MobileNetV2 alone (0.86 accuracy, 0.82 recall, 0.84 F1, 0.91 AUC).
  • Single-image detection time was 120 ms total (40 ms off-chain, 15 ms on-chain, 65 ms model inference); off-chain prescreening reduced model calls by 75% compared with MobileNetV2 alone (95 ms).
  • Robustness: improved perceptual hash achieved 85% success under mild tampering and 70% under heavy tampering, versus traditional hash 45% under heavy tampering; MobileNetV2 achieved 84% under watermark tampering; fusion achieved 80% under heavy scaling + noise.
  • The paper reports P_h = 0.85 and P_m = 0.89, giving P_total = 0.98, and states the improved perceptual hash nearly doubles accuracy for different image similarity judgments with calibration time increase of 0.000005 s.

Methods

  • Proposed system: copyright owners upload images; off-chain and on-chain originality queries use improved perceptual hash and MKD-tree blockchain storage; chaotic Logistic encryption with an MD5-derived key protects images on IPFS/blockchain; MobileNetV2 extracts features and cosine similarity makes the final infringement determination.
  • Data: COCO dataset with 100,000 training and 10,000 validation images; infringement dataset of 1,000 tampered COCO test images (positive) and 1,000 untampered COCO validation images (negative), with light tampering (random cropping retaining 80%, brightness ±20%, Gaussian noise σ = 0.05) and heavy tampering (50% scaling + interpolation, 5×5 Gaussian blur, text watermark), split 8:1:1.
  • Training/evaluation: MobileNetV2 ImageNet pretrained, first 150 layers frozen, last two convolutional blocks and fully connected layer trained with Adam (initial learning rate 1 × 10⁻⁴, decaying to 0.5 every 10 epochs), weighted cross-entropy (positive weight = 2), 50 epochs with early stopping; metrics include accuracy, recall, F1, AUC, detection time, robustness, and paired t-test (α = 0.05).

Limitations

  • Stated: the paper identifies blockchain energy consumption and transaction costs (PoW ≈ 1000 kWh per transaction; suggests PoS/PBFT), storage limitations (e.g., 100,000 images/day ≈ 11.7 TB/year on-chain; Ethereum main chain ≈ 15 TPS), computational resource requirements/algorithm optimization, and data privacy/legal issues (GDPR, CCPA, cross-border data, legal recognition of blockchain deposits).
  • Stated: data are not publicly available due to privacy and are available on request from the corresponding author.
  • Evident from text: on-chain detection was simulated with a local database rather than an actual blockchain; infringement samples were generated by tampering COCO images rather than drawn from real infringement cases; the evaluation uses a single dataset and does not report deployment or user studies.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/jtaer20020076

Cites

35 of 35 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 35 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The improved perceptual hash algorithm's calibration time only increases by 0.000005 s, which is basically negligible.

    “As can be seen from Table 1, the improved perceptual hash algorithm nearly doubles the accuracy for different image similarity judgments, and the calibration time only increases by 0.000005 s (basically negligible).”
    Quote in the paperNot a substantial claim
  • Findingown data

    The improved perceptual hash algorithm nearly doubles the accuracy for different image similarity judgments.

    “As can be seen from Table 1, the improved perceptual hash algorithm nearly doubles the accuracy for different image similarity judgments, and the calibration time only increases by 0.000005 s (basically negligible).”
    Quote in the paperSubstantial claim
  • Findingown data

    The fusion scheme significantly outperforms traditional perceptual hashing (72% accuracy) and MobileNetV2 alone (86% accuracy).

    “which significantly outperforms traditional perceptual hashing (72% accuracy) and MobileNetV2 alone (86% accuracy).”
    Quote in the paperSubstantial claim
  • Findingown data

    The fusion scheme reduces the number of model calls by 75% through off-chain prescreening.

    “it reduces the number of model calls by 75% through the off-chain prescreening”
    Quote in the paperSubstantial claim
  • Findingown data

    The average daily transaction cost can be reduced from 50 USD to less than 5 USD by batch processing and off-chain computation.

    “As shown in Table 6, the average daily transaction cost can be reduced from 50 USD (for a single submission) to less than 5 USD by batch processing and off-chain computation.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim. 23
  • May combine two claims. 42
  • The quote only partly states this claim. 2

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given

    The spatial sampling rate improvement in the frequency domain enhances the accuracy of originality detection.

    “Improvements are made to the problem of poor robustness of traditional image-aware hashing, and the spatial sampling rate in the frequency domain is improved to enhance the accuracy of originality detection.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Findingnone given

    MobileNetV2 has a single image inference time of 65 ms (Tesla V100) and a theoretical throughput of 15 images/s (single GPU).

    “In terms of inference efficiency, MobileNetV2 has a single image inference time of 65 ms (Tesla V100) and a theoretical throughput of 15 images/s (single GPU).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    For a 224×224 image, the single DCT computation time is about 10 ms (CPU).

    “For a 224 × 224 image, the single DCT computation time is about 10 ms (CPU).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    The experiment measured the misjudgment rate < 10^-6 at t = 5, which meets practical application requirements.

    “the experiment measured the misjudgment rate < 10 -6 at t = 5, which meets the practical application requirements [54].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [54] Swaminathan, 2006, Robust and Secure Image Hashing, IEEE Trans. Inf. Forensics Secur. source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Experimentally measured P_h = 0.85, P_m = 0.89, then P_total = 0.98, close to the theoretical optimum.

    “Experimentally measured P h = 0.85, P m = 0.89, then P total = 0.98 are close to the theoretical optimum.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The perceptual hash and MobileNetV2 detection methods fail independently.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Multi-GPU parallelism can achieve a theoretical throughput of up to 400 sheets/s on four GPU nodes.

    “Multi-GPU parallelism can also be used, with a theoretical throughput of up to 400 sheets/s on four GPU nodes.”
    Quote in the paperNot a substantial claim

  • May combine two claims. The quote only partly states this claim.

    Findingown data

    The integration program achieved a tamper detection success rate of 80% under heavy scaling + noise, significantly better than existing schemes.

    “Experimental data show that the F1 score (0.91) and tamper detection success rate (80%) are significantly better than the existing scheme”
    • Also stated in: § 7. Summary · plain
      “| Heavy Scaling + Noise | 45% | 70% | 80% |”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The improved perceptual hash achieves a detection success rate of 85% under mild tampering and 70% under heavy tampering, significantly better than traditional perceptual hash (45%).

    “In terms of robustness, the improved perceptual hash achieves a detection success rate of 85% under mild tampering (e.g., brightness adjustment, local blurring), and maintains a success rate of 70% under heavy tampering (e.g., scaling + noise), which is significantly better than that of the traditional perceptual hash (45%).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The integration program achieved an F1 score of 0.91, significantly better than existing schemes.

    “Experimental data show that the F1 score (0.91) and tamper detection success rate (80%) are significantly better than the existing scheme”
    • Also stated in: § 7. Summary · plain
      “| Integration Program (this article) | 0.93 | 0.89 | 0.91 | 0.95 | 120 |”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given

    The distributed architecture supports ten million image processing.

    “the distributed architecture supports ten million image processing”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The total detection time of the single image detection fusion scheme is 120 ms, of which 40 ms are off-chain detection, 15 ms on-chain detection, and 65 ms model inference.

    “In terms of detection efficiency, the total detection time of the single image detection fusion scheme is 120 ms, of which 40 ms are accounted for off-chain detection, 15 ms for on-chain detection, and 65 ms for model inference.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The IPFS + blockchain hybrid solution strikes a balance between cost and reliability and is suitable for copyright registration scenarios.

    “As can be seen in Table 7, the IPFS + blockchain hybrid solution strikes a balance between cost and reliability and is suitable for copyright registration scenarios.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    The proposed method has a detection accuracy of 93% for untampered images and 70-85% for tampered images.

    “The image copyright registration protection and management method based on blockchain and deep learning proposed in this paper has a detection accuracy of 93% for untampered images and 70-85% for tampered images.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    The annual on-chain storage demand for 100,000 images per day is about 11.7 TB/year, and the on-chain storage cost is about 14,000 USD.

    “If the infringement detection system processes 100,000 images per day, the annual on-chain storage demand is about: 105 × 256 bits/day × 365 ≈ 11.7 TB/year, and the on-chain storage cost (about 1200 USD/GB/year on the Ethernet main chain) is about 14,000 USD.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The experiment used a paired t-test (α = 0.05) comparing F1 scores of the fusion scheme with baselines; with traditional perceptual hashing p = 1.2e-8 (significant) and with MobileNetV2 alone p = 0.003 (significant).

    “Additionally, the experiment was analyzed for significance using a paired t-test (α = 0.05) comparing the F1 scores of the fusion scheme with the baseline method, and the results obtained were with traditional perceptual hashing: p = 1.2 -8 (significant) and with MobileNetV2 alone: p = 0.003 (significant).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    In high-dimensional feature graphs, the inverted residual structure uses ReLU; in low-dimensional feature graphs, a linear activation function is used to solve information loss caused by ReLU.

    “As shown in Figure 6, in high-dimensional feature graphs, the inverted residual structure uses the ReLU activation function [35], while in low-dimensional feature graphs, in order to solve the problem of information loss caused by the ReLU function when dealing with low-dimensional information, the linear activation function [36] is usually used.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [35] Nair, V., and Geoffrey, E.H. (2010, January 21–24). Rectified Linear Units Improve Restricted Boltzmann Machines. Proceedings of the International Conference on Machine Learning, Haifa, Israel.

      Not checked in the free run.

    • cites [36] Hu, 2021, Handling Vanishing Gradient Problem Using Artificial Derivative, IEEE Access source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The proposed blockchain storage structure based on MKD tree supports multidimensional data query and improves query efficiency.

    “Therefore, this paper, on the basis of the original Merkle Tree validation method, is designing a blockchain storage structure based on MKDT (Merkle K-D Tree), which supports multidimensional data query and improves query efficiency.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Each block contains a collection of multidimensional feature data of one image.

  • May combine two claims.

    Premisecitation

    The high efficiency of MobileNetV2 makes image feature extraction faster, suitable for application scenarios that require a fast response, such as instant infringement detection after user-uploaded images.

    “The high efficiency of MobileNetV2 makes image feature extraction faster, which is suitable for application scenarios that require a fast response (e.g., instant infringement detection after user-uploaded images) [25].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [25] Zhang, X., Zhou, X., Lin, M., and Sun, J. (2018, January 18–23). ShuffleNet: An extremely efficient convolutional neural network for mobile devices. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake, UT, USA. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    In existing blockchain technology, the encryption algorithm and the method of calling the contract to obtain the key are complicated and not suitable for the safe storage of image information in the big data environment.

    “In the existing blockchain technology, the traditional hash algorithm has low accuracy for the infringement detection of images after noise addition and rotation operation, and the encryption algorithm and the method of calling the contract to obtain the key are complicated, which is not suitable for the safe storage of image information in the big data environment [5].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] Jothimani, 2014, A Survey on Image Authentication Techniques, Int. J. Eng. Trends Technol. (IJETT) source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    MobileNetV2 is lightweight and high efficiency, with less computation and a low parameter count, making it suitable for running on resource-constrained devices such as mobile devices or embedded systems.

    “MobileNetV2 model has less computation and a low parameter count, which makes it suitable for running on resource-constrained devices (e.g., mobile devices or embedded systems) [23].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [23] Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., and Chen, L.C. (2018, January 18–23). MobileNetV2: Inverted residuals and linear bottlenecks. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake, UT, USA. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The originality detection algorithms used in current blockchain-based schemes are less robust to rotation, translation, and cropping transformations of images, so modified images can easily avoid detection by the copyright system, resulting in copyright infringement.

    “The originality detection algorithms used in the current blockchain-based schemes to provide protection for digital image copyrights are less robust to rotation, translation, and cropping transformations of images.As a result, modified images can easily avoid detection by the copyright system, resulting in copyright infringement [18].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [18] Wu, 2023, Problems and countermeasures of applying blockchain technology to digital copyright, Front. Soc. Sci.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisereasoning

    The inverted residual structure is known as a lightweight convolutional neural network because it can reduce parameters and operations while retaining feature information.

    “This structure is known as a lightweight convolutional neural network because it can reduce parameters and operations while retaining feature information.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Compared with common complex models such as ResNet and VGG, MobileNetV2 significantly reduces computational overhead while guaranteeing higher accuracy, solving the problem that traditional deep learning models are difficult to deploy in mobile or real-time scenarios.

    “Compared with common complex models such as ResNet and VGG, Mo-bileNetV2 significantly reduces the computational overhead while guaranteeing higher accuracy, which solves the problem that traditional deep learning models are difficult to deploy in mobile or real-time scenarios [24].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [24] Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., and Adam, H. (2017). MobileNets: Efficient convolutional neural networks for mobile vision applications. arXiv.

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Blockchain technology can perform copyright verification by means of timestamping, but this method is time-consuming and laborious, violating the original design intention of the blockchain system.

    “Although blockchain technology can perform copyright verification by means of timestamping, this method is time-consuming and laborious, which violates the original design intention of the blockchain system [19,20].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [19] Yuan, 2016, Current status and outlook of blockchain technology development, J. Autom.

      Not checked in the free run.

    • cites [20] Liu, 2019, Limitations and application path of copyright protection using blockchain, Publ. Distrib. Res.

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The chaotic system has characteristics of traversal, non-periodicity, high sensitivity to initial conditions and control parameters, and pseudo-randomness, making it suitable for image encryption.

    “The chaotic system has the characteristics of traversal, non-periodicity, high sensitivity to the initial conditions and control parameters as well as pseudo-randomness [12], which is suitable to be used for image encryption.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [12] Tong, 2021, VFIRM: Verifiable Fine-Grained Encrypted Image Retrieval in Multi-owner Multi-user Settings, IEEE Trans. Serv. Comput. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Traditional methods may require complex feature matching algorithms such as SIFT, SURF, or additional classifiers such as SVM, whereas the MobileNetV2 network model directly utilizes features extracted by the deep learning model to perform comparison, simplifying the process.

    “Traditional methods may require complex feature matching algorithms (e.g., SIFT, SURF) or additional classifiers (e.g., SVM), whereas the MobileNetV2 network model directly utilizes features extracted by the deep learning model to perform comparison, which simplifies the process [27,28].”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [27] Lowe, 2004, Distinctive image features from scale-invariant keypoints, Int. J. Comput. Vis. source

      Not checked in the free run.

    • cites [28] Cortes, 1995, Support-vector networks, Mach. Learn. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Depth convolution deals with spatial information, and point-by-point convolution deals with cross-channel information, realizing complete decoupling.

    “Among them, depth convolution deals with the spatial information in the aspect direction, and point-by-point convolution deals with the cross-channel information, thus realizing the complete decoupling of the spatial information in the aspect direction and the cross-channel information [32].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32] Wang, 2024, Intelligent extraction of Micro-CT fissures in coal based on deep learning and its application, J. China Coal Soc.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Premisenone given

    When U is closer to 4, the values generated by iteration are pseudo-randomly distributed, and the chaotic sequence encryption effect is better.

    “When U closer the value to 4, the values generated by iteration are pseudo-randomly distributed, and the chaotic sequence encryption effect is better.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Features extracted by MobileNetV2 have high generalization ability, can adapt to multiple types of images, and solve the problem of insufficient feature expression ability of traditional methods in complex scenes.

    “The features extracted by MobileNetV2 have high generalization ability, which can adapt to multiple types of images, solving the problem of insufficient feature expression ability of traditional methods in complex scenes [29].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [29] Simonyan, K., and Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    MobileNetV2 is more lightweight compared to other deep neural network models.

    “MobileNetV2 is more lightweight compared to other deep neural network models [30] as shown in Table 2”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] Liu, 2022, Tracking algorithm based on twin network feature fusion and threshold update, Comput. Technol. Dev.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Depth separable convolution splits traditional convolution into depth convolution and point-by-point convolution.

    “Depth separable convolution splits traditional convolution into two steps, i.e., depth convolution and point-by-point convolution.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [31] Cao, S.L., Zhang, G.P., He, Z.Q., Zhang, Y.Q., and Lv, Q. (2023). Y Application of improved MobileNetv2 model in intelligent waste classification system. Mod. Mach., 87–91.

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    Deep separable convolution significantly reduces computational effort of the network, making MobileNetV2 a lightweight convolutional neural network.

    “In summary, deep separable convolution significantly reduces the computational effort of the network with the advantage of its decomposable convolution operation, which makes MobileNetV2 a lightweight convolutional neural network.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Users usually pay attention to image content rather than specific details, so conventional fingerprinting cannot be purely applied to image copyright registration; it needs to be based on image visual content features for originality detection.

    “Users usually pay attention to the content of the image rather than the specific details of the image, so it cannot be purely applied to the copyright registration of the image, and it needs to be based on the image visual content features for originality detection [40].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [40] Hu, G., Jing, L., and Zhu, L. (2016). Application of machine learning-based image feature extraction technology in image copyright protection. Sci. Technol. Innov. Appl., 14–15.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The MobileNetV2 model used in this method is a pre-trained model in the ImageNet-1K dataset.

    “The MobileNetV2 model used in this method is a pre-trained model in the ImageNet-1K dataset [21].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [21] Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., and Fei-Fei, L. (2009, January 20–25). Imagenet: A large-scale hierarchical image database. Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, USA. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    When detecting the originality of an image, traversing the data of each block in the original Merkle tree of the blockchain and retrieving the approximate image through image multidimensional data is inefficient and does not support multidimensional data querying.

    “When detecting the originality of an image, the system traverses the data of each block in the original Merkle tree of the blockchain and retrieves the approximate image through the image multidimensional data to determine its originality, which is an inefficient way of querying and does not support the multidimensional data querying [10].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [10] Chen, Y.-C., Chou, Y.-P., and Chou, Y.-C. (2019). An Image Authentication Scheme Using Merkle Tree Mechanisms. Future Internet, 11. source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Content-based originality detection methods only capture single features and local information of an image, ignoring the global structure, which causes poor originality detection.

    “First, content-based originality detection methods only capture single features and local information of an image, ignoring the global structure of the image.As a result, only the local information of an image can be compared during originality detection, while the similarity of the global structure cannot be taken into account, which in turn causes poor originality detection.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [6] Jia, 2021, SE-Chain: A Scalable Storage and Efficient Retrieval Model for Blockchain, J. Comput. Sci. Technol. source

      Not checked in the free run.

    • cites [7] Zheng, 2020, Queryability optimization for blockchain systems with hybrid indexing, Comput. Sci.

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The proposed method improves the efficiency and accuracy of infringement detection and provides a more secure and reliable solution for copyright protection.

    “This not only improves the efficiency and accuracy of infringement detection but also provides a more secure and reliable solution for copyright protection.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    The proposed method combines advantages of traditional image processing (perceptual hashing) and deep learning (MobileNetV2), ensures data credibility and traceability through blockchain, supports large-scale data processing, and efficient query.

    “Combines the advantages of traditional image processing (perceptual hashing) and deep learning (MobileNetV2).Ensure data credibility and traceability through blockchain, support large-scale data processing, and efficient query.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    DCT concentrates image energy in the low-frequency region, which is suitable for extracting the main features of the image.

    “DCT concentrates the image energy in the low-frequency region, which is suitable for extracting the main features of the image [51].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [51] Ahmed, 1974, Discrete Cosine Transform, IEEE Trans. Comput. source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    If the blockchain deposit certificate is not legally recognized, the credibility of the system will be greatly reduced.

    “If the blockchain deposit certificate is not legally recognized, the credibility of the system will be greatly reduced.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The current legal validity of blockchain deposits has not been fully clarified in different countries.

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationnone given

    User privacy is protected through chaotic encryption technology while supporting large-scale data processing and efficient queries.

    “User privacy is protected through chaotic encryption technology while supporting large-scale data processing and efficient queries.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    It will support high concurrency scenarios through distributed architecture and GPU acceleration.

    “It will support high concurrency scenarios through distributed architecture and GPU acceleration.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    From experimental data, improved perceptual hashing significantly improves tamper resistance, and the system combines complementary nature of hashing and modeling to perform optimally in complex tampering scenarios.

    “From the experimental data above, it can be seen that the improved perceptual hashing significantly improves the tamper resistance, and the system combines the complementary nature of hashing and modeling to perform optimally in complex tampering scenarios.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The proposed method realizes efficient and accurate infringement detection through a triple detection mechanism.

    “The image copyright registration protection and management method based on blockchain and deep learning proposed in this paper, on the other hand, realizes efficient and accurate infringement detection through a triple detection mechanism.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    Extending the original 8×8 low-frequency region to 16×16 covers more medium and high-frequency components and enhances contour detail-capturing ability.

    “Then, the original 8 × 8 low-frequency region is extended to 16 × 16 to cover more medium and high-frequency components and enhance the contour detail-capturing ability [52].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [52] Tang, 2018, Robust Image Hashing with Dominant DCT Coefficients, Signal Process.

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The AUC of 0.95 is close to the perfect classifier (AUC = 1), indicating that the model performs well under different thresholds.

    “which is close to the perfect classifier (AUC = 1), indicating that the model performs well under different thresholds.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims. The quote only partly states this claim.

    Interpretationreasoning

    The proposed triple detection system supports infringement detection, copyright registration, privacy protection, and large-scale data processing through distributed architecture and storage optimization.

    “It supports functions such as infringement detection, copyright registration, privacy protection, etc.”
    • Also stated in: § System applicability · plain
      “It is applicable to a wide range of scenarios, has high versatility, and supports large-scale data processing through distributed architecture and storage optimization.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Blockchain is utilized to store the hash value to ensure non-tampering of copyright information.

    “Blockchain is utilized to store the hash value to ensure the non-tampering of copyright information.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: All hash values are stored on the chain.

  • May combine two claims.

    Interpretationown data

    The triple detection mechanism proposed in this paper improves perceptual hashing to quickly screen suspected infringing images through off-chain detection, utilizes blockchain to store the hash value to ensure data cannot be tampered with through on-chain detection, and provides high-precision infringement determination through MobileNetV2 model detection.

    “The triple detection mechanism proposed in this paper improves perceptual hashing to quickly screen suspected infringing images through off-chain detection, utilizes blockchain to store the hash value to ensure that the data cannot be tampered with through on-chain detection, and provides high-precision infringement determination through MobileNetV2 model detection.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: All hash values are stored on the chain.

  • May combine two claims.

    Interpretationcitation

    The blockchain registration system stores on-chain query summaries with user identification in an MKD tree structure to ensure tamperability of on-chain query summaries and improve efficiency of on-chain query.

    “The blockchain registration system stores the on-chain query summaries of the images that have passed the originality detection together with the user identification information (e.g., user address) in the blockchain with the structure of the MKD tree to ensure the tamperability of the on-chain query summaries and improve the efficiency of the on-chain query.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [47] Tian, 2021, Design and realization of digital content resources registration service system, J. Harbin Inst. Technol.

      Not checked in the free run.

    • The authors state this as an assumption: All hash values are stored on the chain.

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationcitation

    To ensure accuracy of image similarity detection, it is necessary to break technical barriers of blockchain, combine deep learning, and carry out a secondary comparison of images detected by upper and lower chains.

    “The aforementioned has improved the hash algorithm in the current blockchain technology, and in order to ensure the accuracy of image similarity detection to the greatest extent, it is necessary to break the technical barriers of the blockchain, combined with deep learning technology, and carry out a secondary comparison of the images that have been detected by the upper and lower chains [41].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [41] Zhu, S., Kang, B.N., and Kim, D. (2016, January 9–12). A deep neural network based hashing for efficient image retrieval. Proceedings of the 2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Budapest, Hungary.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationreasoning

    The proposed triple detection system is applicable to a wide range of scenarios, including digital content distribution platforms, social media infringement monitoring, and copyright registration and trading platforms.

    “The triple detection system proposed in this paper is applicable to a wide range of scenarios, including digital content distribution platforms (e.g., music, video, and image libraries), social media infringement monitoring, and copyright registration and trading platforms.”
    • Also stated in: § System applicability · plain
      “It is applicable to a wide range of scenarios, has high versatility, and supports large-scale data processing through distributed architecture and storage optimization.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Therefore, the triple detection mechanism covers the whole process from fast screening to high-precision determination, while the improved perceptual hash enhances detection ability of tampered images, and blockchain storage ensures data's non-tamperability and traceability.

    “Therefore, the triple detection mechanism covers the whole process from fast screening to high-precision determination, while the improved perceptual hash enhances the detection ability of tampered images, and the blockchain storage ensures the data's non-tamperability and traceability.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: All hash values are stored on the chain.

  • Probably not a substantial claim: may not pass S4 (based).

    Study design and conductnone given

    The threshold selection on the optimal Hamming distance threshold is determined by the validation set (e.g., distance ≤ 5 is judged as similar).

    “The threshold selection on the optimal Hamming distance threshold is determined by the validation set (e.g., distance ≤ 5 is judged as similar).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The chaotic encryption sequence protects uplinked image data so that copyright data stored on it will not be maliciously tampered with.

    “The chaotic encryption sequence is utilized to protect the uplinked image data so that the copyright data stored on it will not be maliciously tampered with”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    The system adopts fuzzy commitment-based storage of the off-chain query digest to improve security when querying and ensure tamperability of image fingerprints.

    “It adopts the fuzzy commitment-based storage of the off-chain query digest to improve the security when querying, and to ensure the tamperability of the image fingerprints [46].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [46] Iida, K., Kobayashi, H., and Kiya, H. (September, January 28). Secure identification based on fuzzy commitment scheme for JPEG XR images. Proceedings of the 2016 24th European Signal Processing Conference (EUSIPCO), Budapest, Hungary. source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The fusion scheme significantly reduces overall system load and is suitable for high concurrency scenarios.

    “which significantly reduces the overall system load and is suitable for high concurrency scenarios.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The triple detection mechanism can effectively defend against counterattacks (such as targeted modification of local features to bypass a single detection).

    “The triple detection mechanism can effectively defend against counterattacks (such as targeted modification of local features to bypass a single detection).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    The MKD tree-based block storage structure ensures tamperability of image fingerprints using on-chain storage.

    “A block storage structure based on the MKD tree is proposed to ensure the tamperability of image fingerprints using on-chain storage.”
    Quote in the paperNot a substantial claim
59 of 59

  • The authors state this as an assumption.

    Each block contains a collection of multidimensional feature data of one image.

    “We assume that each block contains a collection of multidimensional feature data of one image v = {< P 1 , L 1 >, < P 2 , L 2 >, < P 3 , L 3 >, . . ., < P n , L n >} and the multidimensional feature data of the first image v i =< P i , L i > of the first image.”

  • The authors state this as an assumption.

    All hash values are stored on the chain.

    “Assuming that all hash values are stored on the chain, use smart contracts to simulate hash matching (simulated by local databases in actual tests).”

  • The paper relies on this without support.

    The perceptual hash and MobileNetV2 detection methods fail independently.

    “where (1 -P h ) is the leakage rate of the sensed hash, (1 -P m ) is the leakage rate of MobileNetV2, and P total denotes the probability that at least one method detects the target.”

  • The paper relies on this without support.

    Existing AI copyright systems have detection accuracy over 90% for untampered images and 50-70% for tampered images.

    “The detection accuracy of existing AI copyright systems is high for untampered images (over 90%) but low for tampered images (50-70%).”

  • The paper relies on this without support.

    Existing AI copyright systems have a single image detection time of about 100 ms.

    “The detection time for a single image is about 100 ms (GPU dependent).”

  • The paper relies on this without support.

    Existing blockchain copyright systems have high on-chain latency (~10 min/block for Bitcoin) and high on-chain storage cost (about 1200 USD/GB/year for Ether main chain).

    “High latency of on-chain operations (~10 min/block for Bitcoin).High on-chain storage cost (about 1200 USD/GB/year for Ether main chain).”

  • The paper relies on this without support.

    Traditional centralized copyright management has a cumbersome process, high cost, difficulty in managing data, and a long period of confirming rights.

    “The traditional centralized copyright management method has many defects in solving the digital infringement problem due to its cumbersome process, high cost, difficulty in managing data, and long period of confirming rights.”

  • The paper relies on this without support.

    Existing AI copyright systems rely on centralized servers, have single-point-of-failure risks, and lack protection of user privacy.

    “they have limited ability to detect tampered images, rely on centralized servers, have single-point-of-failure risks, and lack protection of user privacy.”

  • The paper relies on this without support.

    Existing blockchain copyright systems lack efficient content detection, require manual uploading, have high storage costs, are difficult to support large-scale data, and smart contracts may conflict with the traditional legal framework.

    “lack efficient content detection capabilities, usually require manual uploading of copyright information, have high storage costs, are difficult to support large-scale data, and the automatic execution of smart contracts may conflict with the traditional legal framework.”

  • The paper relies on this without support.

    The IPFS + Filecoin hybrid storage cost is 2.5 USD/GB/year.

    “Using IPFS + Filecoin hybrid storage, the cost is 2.5 USD/GB/year, significantly lower than pure on-chain storage.”

  • The paper relies on this without support.

    Conventional data fingerprinting methods cannot achieve good originality detection for image data, and the definition of uniqueness is particularly precise and strict.

    “However, for image data, conventional data fingerprinting methods cannot achieve good originality detection, and the definition of uniqueness is particularly precise and strict.”

  • The paper relies on this without support.

    Existing AI copyright systems are mostly content distribution platforms and have limited ability to detect tampered images, relying on centralized servers with single-point-of-failure risk.

    “The applicable scenarios of existing AI copyright systems are mostly content distribution platforms (e.g., YouTube, Spotify).It has limited ability to detect tampered images and relies on centralized servers, with the risk of a single point of failure.”

  • The paper relies on this without support.

    Existing blockchain copyright systems are mostly copyright registration and trading platforms and lack efficient content detection, have high storage cost, and difficulty supporting large-scale data.

    “The applicable scenarios of existing blockchain copyright systems, on the other hand, are copyright registration and trading platforms (e.g., KodakOne, Po.et).Lack of efficient content detection capability.High storage cost, difficult to support large-scale data.”

  • The paper relies on this without support.

    Blockchain deposit improves legal validity.

    “Blockchain deposit improves legal validity”

  • The paper relies on this without support.

    The current legal validity of blockchain deposits has not been fully clarified in different countries.

    “The current legal validity of blockchain deposits has not been fully clarified in different countries.”

  • The paper relies on this without support.

    The automatic execution of smart contracts may conflict with traditional legal frameworks.

    “Also note that the automatic execution of smart contracts may conflict with traditional legal frameworks”

  • The paper relies on this without support.

    Copyright laws differ from country to country.

    “as copyright laws differ from country to country”

  • The paper relies on this without support.

    MobileNetV2 model inference relies on high-performance GPUs, requiring expensive hardware for large-scale deployment.

    “In terms of hardware, MobileNetV2 model inference relies on high-performance GPUs (e.g., Tesla V100), which require expensive hardware support for large-scale deployment.”

  • The paper relies on this without support.

    GPU acceleration can significantly improve DCT computation efficiency and increase hardware cost.

    “Although the DCT computation of perceptual hashing can be accomplished by CPU, GPU acceleration can significantly improve efficiency and further increase the hardware cost.”

  • The paper relies on this without support.

    The GDPR requires transparent and lawful processing of personal data and ensures data subject rights.

    “The GDPR requires that the processing of personal data be transparent and lawful, and ensures the rights of data subjects (e.g., access, deletion, correction).”

  • The paper relies on this without support.

    The CCPA gives California residents the right to control their personal data.

    “The CCPA gives California residents the right to control their personal data, including the right to object to the sale of data.”

  • The paper relies on this without support.

    Transparency in blockchain can lead to disclosure of sensitive information.

    “Transparency in blockchain can lead to the disclosure of sensitive information, especially in copyright registration scenarios where users may be reluctant to disclose image hashes.”

  • The paper relies on this without support.

    PoW consensus consumes approximately 1000 kWh per transaction and is not suitable for high-frequency infringement detection.

    “The energy consumption problem of blockchain nodes has not been fully solved, especially since the PoW consensus mechanism consumes ≈ 1000 kWh for a single transaction, which is not suitable for high-frequency infringement detection.”

  • The paper relies on this without support.

    PoS can reduce energy consumption by 99% and is suitable for this system.

    “In this regard, PoS (Ethernet 2.0 single transaction energy consumption ≈ 0.01 kWh) or PBFT can be used, which can effectively reduce energy consumption by 99% and is suitable for this system.”

  • The paper relies on this without support.

    Each node of the blockchain needs to store complete historical data, causing linear storage cost increase.

    “Each node of the blockchain needs to store the complete historical data, resulting in a linear increase in storage costs with the amount of data, and a higher pressure on the storage of the whole node.”

  • The paper relies on this without support.

    Blockchain TPS is limited (Ethernet mainchain approximately 15 TPS) and may congest at high concurrency.

    “At the same time, the number of transactions per second (TPS) of the blockchain is limited (e.g., Ethernet mainchain ≈ 15 TPS), and it may be congested at high concurrency.”

  • The paper relies on this without support.

    The MD5 chaotic encryption algorithm reduces the possibility of users stealing copyright information and ensures image fingerprints are not tampered with.

    “the MD5 chaotic encryption algorithm reduces the possibility of users stealing copyright information and ensures that image fingerprints are not tampered with.”
27 of 27

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Existing works that integrate perceptual hashing into blockchain-based copyright registration strengthen transaction integrity through encryption and smart contract validation, but the similarity layer remains largely unchanged.

    “In these works, the transaction integrity is strengthened through encryption and smart contract validation. But the similarity layer itself remains largely unchanged.”

    From DBpHash: a blockchain-based dual-band perceptual hashing framework for copyright protection of purely chromatic background images (2026) · cites this paper as [29]

    Section 2 (page unavailable) says the paper improves traditional perceptual hashing to address low accuracy in determining image similarity; its Summary reports better tampering-detection results for the improved hash. The paper discusses encryption and smart contracts, but its similarity layer does not remain largely unchanged. (LLM, conf=0.88)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Reported storage cost is inconsistent with the stated storage volume

    Major

    At 1,000 GB per TB, 11.7 TB at 1200 USD/GB/year would cost 14,040,000 USD/year, not about 14,000 USD.

    Reported 14000 · recomputed 14040000

  • Encryption and decryption use different chaotic-map parameters

    Major

    The encryption procedure sets U to 4, but the decryption pseudocode sets it to 3.99. Those parameters generate different chaotic sequences, so the specified decryption cannot reverse the specified encryption.

2 of 2

Cites

Flagged first, then by citations

3 of the 38 references couldn’t be matched to a known paper.

Research works printed but not linked: 14

  1. Peng, 2015, Design and implementation of digital watermarking copyright protection system based on visual password, Comput. Sci. Appl., 5, 285
  2. Zheng, 2020, Queryability optimization for blockchain systems with hybrid indexing, Comput. Sci., 47, 301
  3. Ali, 2021, Blockchain-based secure data storage and sharing framework using IPFS, Future Gener. Comput. Syst., 118, 12
  4. Zhou, 2020, A novel image encryption scheme based on chaotic maps and DNA sequence operations, Signal Process. Image Commun., 85, 115876
  5. Xu, 2020, A chaos-based image encryption scheme with post-processing for enhanced security, Multimed. Tools Appl., 79, 10245
  6. Hua, 2019, A new image encryption algorithm based on dynamic index mapping and chaotic systems, IEEE Access, 7, 123456
  7. Wang, 2021, A secure image encryption algorithm based on chaos and blockchain, IEEE Trans. Multimed., 23, 567
  8. Yuan, 2016, Current status and outlook of blockchain technology development, J. Autom., 42, 481
  9. Liu, 2019, Limitations and application path of copyright protection using blockchain, Publ. Distrib. Res., 9, 48
  10. Liu, 2022, Tracking algorithm based on twin network feature fusion and threshold update, Comput. Technol. Dev., 32, 65
  11. Wang, 2024, Intelligent extraction of Micro-CT fissures in coal based on deep learning and its application, J. China Coal Soc., 49, 3439
  12. Wijaya, 2023, Alvianus Sembiring, Christian Mitro Septiano Hutagalung. Satellite Images Classification using MobileNet V-2 Algorithm, Sink. J. Penelit. Tek. Inform., 7, 2316
  13. Tian, 2021, Design and realization of digital content resources registration service system, J. Harbin Inst. Technol., 26, 119
  14. Kingma, D.P., and Ba, J. (2015, January 7–9). Adam: A method for stochastic optimization. Proceedings of the 3rd International Conference on Learning Representations (ICLR), San Diego, CA, USA.

Cited by

Sorted by citations

Related

Authors
Xu, J., Zhang, J., & Wang, J.
Affiliations
  • China University of Petroleum, East China (CN)
  • University of Chinese Academy of Sciences (CN)
Corresponding author
Xu, Jikuan
ORCID
Zhang, Jiamin
Venue
Journal of Theoretical and Applied Electronic Commerce Research
Publisher
MDPI AG
Published
2025-04-18
Volume
20
Issue
2
Pages
76
ISSN
0718-1876 (ISSN-L)
DOI
10.3390/jtaer20020076
OpenAlex
W4409584896
Semantic Scholar
d754a606165ac66bde13c3f43b344f43f9f2305f
Cited by
9
References
38
Citations by year
2026: 6 · 2025: 3
FWCI
3.18
Type
journal article
Language
en
Open access
Licence
Locations
  • Journal of theoretical and applied electronic commerce research (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Department of Science and Technology of Shandong Province (China) (ZR2024QG075, 2023RW017)
  • Shandong Province Higher Education Excellent Youth Innovation Team ‘Science and Technology Law Collaborative Innovation Mechanism Research Innovation Team’ (ZR2024QG075, 2023RW017)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Steganography and Watermarking Techniques
  • Physical Sciences › Computer Science › Information Systems › Blockchain Technology Applications and Security
  • Physical Sciences › Computer Science › Artificial Intelligence › Law, AI, and Intellectual Property
Keywords
copyright infringement detection, blockchain, artificial intelligence, perceptual hashing, chaos-based cryptography, deep learning, copyright registration

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-08.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-08.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review