Multimedia content traceability system and method based on AI synthesis detection and block chain evidence storage

Abstract

The invention relates to the technical field of communication, in particular to a multimedia content traceability system and method based on AI synthesis detection and block chain evidence storage, and the system comprises a user uploading module, an AI detection module, an abstract data generation module, a block chain evidence storage module, and an information return and query module. The method has the beneficial effects that contents such as uploaded pictures, audios and videos are analyzed through an AI detection technology, whether the contents are synthetic contents or not is judged, credible abstract data are generated based on a detection result, real source recording and on-chain traceability of the contents are realized by means of tampering resistance of the block chain, and false information spreading is prevented.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and specifically to a multimedia content tracing system and method based on AI synthesis detection and blockchain evidence storage. Background Art

[0002] With the widespread application of AI-generated technologies (such as Deepfake and GAN), the threshold for fabricating images, videos, and audio content has been significantly lowered. While this has boosted the creative industry, it has also led to serious problems such as information distortion, the spread of fabricated content, and privacy violations. The public has difficulty discerning whether a video or image is authentic, and the crisis of trust in media content is intensifying.

[0003] At the same time, existing methods for tracing the provenance of media content generally rely on centralized platforms (such as social media or content hosting services), which are susceptible to tampering and lack credibility. While some attempts have been made to use blockchain for content ownership and traceability, it is difficult to integrate AI detection technology to automatically identify synthetic content, and it is impossible to provide a closed-loop solution that integrates content detection, trusted evidence storage, and tracking mechanisms.

[0004] Therefore, there is an urgent need for an innovative system that combines AI detection and blockchain technology, which can identify suspicious content in real time, automatically store evidence, and provide reliable traceability and verification capabilities during subsequent dissemination. Summary of the Invention

[0005] The purpose of the present invention is to provide a multimedia content traceability system and method based on AI synthesis detection and blockchain evidence storage to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a multimedia content traceability system based on AI synthesis detection and blockchain evidence storage, comprising:

[0007] User upload module: used to receive multimedia content to be tested uploaded by users through the front-end interactive interface, supporting multiple formats of images, videos and audio, and performing basic format verification and metadata extraction on the uploaded content, while generating a preliminary hash value for duplicate detection and unique identification;

[0008] AI Detection Module: Built based on a deep learning algorithm, it includes an image / video preprocessing module, a feature extraction module, and a discrimination module. It is used to standardize and extract features from uploaded multimedia content and output a probability value indicating whether the content is AI-generated.

[0009] Summary data generation module: After AI detection is completed, the unique hash value of the multimedia content, the AI ​​synthesis judgment result, the detection timestamp, the content type and basic metadata, the detection device or model identifier, and the user identity are combined into a summary data structure, encapsulated using a unified JSON or binary encoding method, and protected against tampering through a digital signature;

[0010] Blockchain evidence storage module: Deploy smart contracts in a consortium chain or public chain, submit summary data to the blockchain node by calling the contract interface. The contract verifies the data format and signature legitimacy and packages it into the block as transaction data. After the data is successfully written to the blockchain, the block hash and transaction ID identification information are returned;

[0011] Information return and query module: After the on-chain evidence is completed, a complete detection report is returned to the user, including whether the content is AI synthesis, the probability of synthesis, the blockchain evidence result and the on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level, the query interface link or QR code. At the same time, it supports open query interface and verification mechanism. Any dissemination node or recipient can verify the authenticity of the content through hash value comparison, blockchain data reading, API docking and embedding.

[0012] Preferably, the image / video preprocessing module performs standardization on the uploaded multimedia content, including adjusting the resolution, unifying the number of frames, image slicing, and color normalization operations to ensure that the input meets the model requirements.

[0013] Preferably, the feature extraction module uses a pre-trained neural network model to extract image features, texture, frequency domain distribution, and artifact trace information, performs frame-level sampling and time series feature extraction on the video content, and combines optical flow analysis to enhance the recognition ability of AI-generated dynamic content.

[0014] Preferably, the discrimination module is based on a trained discrimination model and combined with the deep feature input of the image or video frame to output the probability value that it is AI synthesized content. The system sets a threshold for the probability value. If it is higher than the threshold, it is preliminarily considered that the content is AI synthesized, and the threshold can be adjusted to adapt to different application scenarios.

[0015] Preferably, the summary data format, verification process, storage structure and query interface are defined in the smart contract, and the contract supports additional fields to facilitate future tracing; the test report can be exported as a PDF file or JSON structure, and can also be called by a third-party platform through an API interface for publishing platforms, social networks or judicial institutions to verify content traceability information.

[0016] A method for a multimedia content traceability system based on AI synthesis detection and blockchain evidence storage, comprising the following steps:

[0017] S1. User uploads multimedia content: Users upload multimedia content to be tested through the system's front-end interactive interface. Multiple formats, including images, videos, and audio, are supported. The uploaded content can be captured, generated, or collected from third-party platforms. Upon receiving the uploaded content, the system performs basic format verification and metadata extraction, and generates a preliminary hash value for duplicate detection and unique identification.

[0018] S2. AI Detection Module Analysis: The AI ​​Detection Module, built based on a deep learning algorithm, analyzes uploaded multimedia content, performing normalization processing through the Image / Video Preprocessing Module, extracting relevant features through the Feature Extraction Module, and finally outputting a probability value for whether the content is AI-generated.

[0019] S3. Generate summary data: After AI detection is completed, the system combines the unique hash value of the multimedia content, the AI ​​synthesis judgment result, the detection timestamp, the content type and basic metadata, the detection device or model identifier, and the user identity into a summary data structure, encapsulates it using a unified encoding method, and protects it against tampering with a digital signature.

[0020] S4. Blockchain Evidence Storage: Through a smart contract pre-deployed in a consortium chain or public chain, the summary data is submitted to the blockchain node. The contract verifies the data format and signature legitimacy and packages it into a block as transaction data. After the data is successfully written to the blockchain, the block hash and transaction ID identification information are returned;

[0021] S5. Return of evidence information and credibility: After the on-chain evidence is stored, the system returns a complete test report to the user, including whether the content is AI-synthesized, the probability of synthesis, the blockchain evidence result and on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level, and the query interface link or QR code;

[0022] S6. Traceability query during content dissemination: The system supports open query interfaces and verification mechanisms. Any dissemination node or recipient can verify the authenticity of the content through hash value comparison, blockchain data reading, API docking and embedding.

[0023] Preferably, in step S2, the image / video preprocessing module performs standardization processing on the uploaded multimedia content, specifically including adjusting the resolution, unifying the number of frames, image slicing, and color normalization operations to ensure that the input meets the model requirements.

[0024] Preferably, the feature extraction module in step S2 adopts a pre-trained neural network model to extract image features, texture, frequency domain distribution, and artifact trace information, performs frame-level sampling and time series feature extraction on the video content, and combines it with optical flow analysis to enhance the recognition ability of AI-generated dynamic content.

[0025] Preferably, the discrimination module in step S2 outputs a probability value of the content being AI-synthesized based on the trained discrimination model and the deep feature input of the image or video frame; the system sets a threshold for the probability value, and if the value is higher than the threshold, the content is preliminarily considered to be AI-synthesized, and the threshold can be adjusted to adapt to different application scenarios.

[0026] Preferably, when the blockchain data is read in step S6, by inputting the hash value, transaction ID or QR code, the system queries the corresponding summary data and AI detection results on the chain, and the visual front-end supports the display of detection records, trust level, and evidence storage time content; when the API is connected and embedded, the platform operator can embed the verification module into its content management system to realize automatic recognition and labeling of AI-generated content, or provide ordinary users with an "authenticity verification" function.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The multimedia content traceability system and method proposed in this paper, based on AI synthesis detection and blockchain evidence storage, uses AI detection technology to analyze uploaded images, audio, video, and other content to determine whether it is synthetic content. Based on the detection results, it generates trusted summary data. Leveraging the immutability of blockchain, it records the true source of the content and traces it on the chain, preventing the spread of false information. To protect personal privacy and portrait rights, the system can automatically identify and mark content that may infringe on personal portrait rights, such as AI face-swapping and deep fakes, as "AI synthesis" and store this information on the chain for evidence, strengthening the evidentiary basis for victims to protect their rights and curbing the spread of illegally generated content. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0030] In order to clearly and completely describe the objectives and technical solutions of the present invention and make the advantages more clearly understood, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, not all of them, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] In the first embodiment, the present invention provides a technical solution: a multimedia content traceability system based on AI synthesis detection and blockchain evidence storage, comprising:

[0032] User upload module: used to receive multimedia content to be detected uploaded by users through the front-end interactive interface, supporting multiple formats of images, videos and audios, and performing basic format verification and metadata extraction on the uploaded content, while generating a preliminary hash value for duplicate detection and unique identification.

[0033] AI detection module: built based on deep learning algorithms, it includes an image / video preprocessing module, a feature extraction module, and a discrimination module, which are used to standardize and extract features of uploaded multimedia content and output a probability value of whether it is AI-synthesized; the image / video preprocessing module standardizes the uploaded multimedia content, including adjusting the resolution, unifying the number of frames, image slicing, and color normalization to ensure that the input meets the model requirements; the feature extraction module uses a pre-trained neural network model to extract image features, texture, frequency domain distribution, and artifact trace information, performs frame-level sampling and time series feature extraction on the video content, and combines optical flow analysis to enhance the ability to recognize AI-generated dynamic content; the discrimination module is based on the trained discrimination model and combines the deep feature input of the image or video frame to output the probability value of it being AI-synthesized content. The system sets a threshold for the probability value. If it is higher than the threshold, it is preliminarily considered that the content is AI-synthesized, and the threshold can be adjusted to adapt to different application scenarios.

[0034] Summary data generation module: After AI detection is completed, the unique hash value of the multimedia content, AI synthetic judgment results, detection timestamp, content type and basic metadata, detection device or model identifier, and user identity identifier are combined into a summary data structure, which is encapsulated using a unified JSON or binary encoding method and protected against tampering through a digital signature.

[0035] Blockchain evidence storage module: deploy smart contracts in consortium chains or public chains, submit summary data to blockchain nodes by calling contract interfaces, and the contract verifies the data format and signature legitimacy before packaging it into blocks as transaction data. After the data is successfully written into the blockchain, the block hash and transaction ID identification information are returned. The smart contract defines the summary data format, verification process, storage structure, and query interface. The contract supports additional fields for easy traceability in the future. The test report can be exported as a PDF file or JSON structure, or can be called by a third-party platform through an API interface for publishing platforms, social networks, or judicial institutions to verify content traceability information.

[0036] Information return and query module: After the on-chain evidence is completed, a complete detection report is returned to the user, including whether the content is AI synthesis, the probability of synthesis, the blockchain evidence result and the on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level, the query interface link or QR code. At the same time, it supports open query interface and verification mechanism. Any dissemination node or recipient can verify the authenticity of the content through hash value comparison, blockchain data reading, API docking and embedding.

[0037] Example 2, based on Example 1, proposes a multimedia content traceability method based on AI synthesis detection and blockchain evidence storage, including the following steps:

[0038] S1, users upload multimedia content;

[0039] S2, the AI ​​detection module analyzes the content and outputs a probability value of whether it is AI synthesis;

[0040] S3. Generate summary data by combining the content hash value and the detection result;

[0041] S4. Write the summary data into the blockchain for evidence storage through smart contracts;

[0042] S5. The system returns the evidence information and credibility mark;

[0043] S6. During the content dissemination process, any node can query content traceability information through hash verification and on-chain information.

[0044] Step S1: The specific steps for users to upload multimedia content are as follows:

[0045] 1.1 Users upload multimedia content to be tested through the front-end interactive interface provided by the system. Supported formats include but are not limited to images (JPG, PNG, etc.), videos (MP4, AVI, etc.), and audio (MP3, WAV, etc.). Uploaded content can be captured, generated, or collected from third-party platforms.

[0046] 1.2 After receiving the uploaded content, the system first performs basic format verification and metadata extraction (such as resolution, frame rate, shooting time, encoding method, etc.). This step must ensure that the input content format is legal and provide the necessary context information for subsequent AI detection.

[0047] 1.3 To prevent malicious uploading or duplicate detection, the system will generate a preliminary hash value (such as SHA-256) for the uploaded content for duplicate detection and unique identification.

[0048] Step S2 uses the AI ​​detection module to analyze the content and output the probability value of whether it is AI synthesis. The steps are as follows:

[0049] The detection module is built based on deep learning algorithms and is divided into the following submodules:

[0050] 2.1 Image / video preprocessing module: This module standardizes the uploaded multimedia content, such as adjusting the resolution, unifying the number of frames, slicing the image, normalizing the color, etc., to ensure that the input meets the model requirements.

[0051] 2.2 Feature Extraction Module: Pre-trained neural network models (such as EfficientNet, ResNet, Xception, etc.) are used to extract information such as image features, texture, frequency domain distribution, and artifact traces. Frame-level sampling and time series feature extraction are performed on video content, combined with optical flow analysis to enhance the recognition ability of AI-generated dynamic content.

[0052] 2.3 Discrimination Module: Based on a trained discriminant model and combined with the deep feature input of an image or video frame, it outputs a probability value for AI-generated content. The model can be constructed based on GAN anti-detection technology, spectral anomaly detection, image artifact analysis, extended edge features, and other methods. The system sets a threshold for the probability value (e.g., 80%). If the threshold is higher than this, the content is preliminarily considered to be AI-generated. This threshold can be adjusted to adapt to different application scenarios.

[0053] Step S3 generates summary data by combining the content hash value and the detection result. The specific steps are as follows:

[0054] 3.1 After AI detection is completed, the system organizes the following data into a summary data structure:

[0055] Unique hash value of multimedia content (using SHA-256 algorithm);

[0056] AI synthesis judgment results (Boolean value and corresponding probability);

[0057] Detection timestamp (based on a trusted time source);

[0058] Content type and basic metadata (such as image / video ID, duration, resolution, etc.);

[0059] Testing equipment or model identification (such as testing model version);

[0060] User identity (if real-name registration is required).

[0061] The summary data structure is encapsulated using a unified JSON or binary encoding method and is protected against tampering by a digital signature. The signature key is generated and kept by the system's central node or smart contract.

[0062] Step S4 writes the summary data into the blockchain through the smart contract. The specific steps are as follows:

[0063] 4.1 Smart Contract Deployment. Smart contracts are pre-deployed in consortium chains or public chains (such as Ethereum, Hyperledger Fabric, FISCO BCOS, etc.). The contract defines the summary data format, verification process, storage structure and query interface.

[0064] 4.2 The system submits the summary data to the blockchain node by calling the contract interface. The contract first verifies the data format and signature legitimacy. If verified, it is packaged as transaction data into the block. The contract also supports additional fields (such as the decentralized storage address to which the content points) for easy traceability.

[0065] 4.3 After the data is successfully written into the blockchain, identification information such as the block hash and transaction ID will be obtained. This information, along with the summary data, will be returned to the user as a certificate for subsequent evidence storage. Each transaction has a consensus record across the entire network and has strong credibility.

[0066] Step S5: The system returns the evidence information and credibility mark. The specific steps are as follows:

[0067] 5.1 After the system completes the on-chain evidence storage, it will return a complete test report to the user, including the following: whether the content is AI synthesis, the synthesis probability, the blockchain evidence result and the on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level (the AE level can be set, based on the comprehensive judgment of the test results), and the query interface link or QR code (for public verification).

[0068] The report can be exported as a PDF file or JSON structure, or can be called by third-party platforms through the API interface for publishing platforms, social networks or judicial institutions to verify content traceability information.

[0069] During the content dissemination process in step S6, any node can query the content traceability information through hash verification and on-chain information. The specific steps are as follows:

[0070] The system supports open query interfaces and verification mechanisms. Any dissemination node or recipient can verify the authenticity of the content as follows:

[0071] 6.1 Hash value comparison: The received multimedia content first calculates its hash value and matches it with the hash value in the blockchain evidence. If they match, it means that the content is the evidenced version.

[0072] 6.2 Blockchain data reading: By entering a hash value, transaction ID, or QR code, the system queries the corresponding summary data and AI detection results on the chain. The visual front-end supports displaying detection records, trust level, storage time, and other content to help the public judge its authenticity and credibility.

[0073] 6.3 API integration and embedding: Platform operators can embed verification modules into their content management systems to automatically identify and label AI-generated content, or provide ordinary users with "authenticity verification" functions to enhance content transparency and trust.

[0074] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multimedia content traceability system based on AI synthesis detection and blockchain evidence storage, characterized by: include: User upload module: used to receive multimedia content to be tested uploaded by users through the front-end interactive interface, supporting multiple formats of images, videos and audio, and performing basic format verification and metadata extraction on the uploaded content, while generating a preliminary hash value for duplicate detection and unique identification; AI Detection Module: Built based on a deep learning algorithm, it includes an image / video preprocessing module, a feature extraction module, and a discrimination module. It is used to standardize and extract features from uploaded multimedia content and output a probability value indicating whether the content is AI-generated. Summary data generation module: After AI detection is completed, the unique hash value of the multimedia content, the AI ​​synthesis judgment result, the detection timestamp, the content type and basic metadata, the detection device or model identifier, and the user identity are combined into a summary data structure, encapsulated using a unified JSON or binary encoding method, and protected against tampering through a digital signature; Blockchain evidence storage module: Deploy smart contracts in a consortium chain or public chain, submit summary data to the blockchain node by calling the contract interface. The contract verifies the data format and signature legitimacy and packages it into the block as transaction data. After the data is successfully written to the blockchain, the block hash and transaction ID identification information are returned; Information return and query module: After the on-chain evidence is completed, a complete detection report is returned to the user, including whether the content is AI synthesis, the probability of synthesis, the blockchain evidence result and the on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level, the query interface link or QR code. At the same time, it supports open query interface and verification mechanism. Any dissemination node or recipient can verify the authenticity of the content through hash value comparison, blockchain data reading, API docking and embedding.

2. The multimedia content traceability system based on AI synthesis detection and blockchain evidence storage according to claim 1 is characterized by: The image / video preprocessing module standardizes the uploaded multimedia content, including adjusting the resolution, unifying the frame rate, image slicing, and color normalization to ensure that the input meets the model requirements.

3. The multimedia content traceability system based on AI synthesis detection and blockchain evidence storage according to claim 2 is characterized by: The feature extraction module uses a pre-trained neural network model to extract image features, texture, frequency domain distribution, and artifact trace information, performs frame-level sampling and time series feature extraction on video content, and combines it with optical flow analysis to enhance the recognition ability of AI-generated dynamic content.

4. The multimedia content traceability system based on AI synthesis detection and blockchain evidence storage according to claim 3 is characterized by: The discrimination module is based on a trained discrimination model and combines the deep feature input of the image or video frame to output the probability value of it being AI-synthesized content. The system sets a threshold for the probability value. If it is higher than the threshold, it is preliminarily considered that the content is AI-synthesized, and the threshold can be adjusted to adapt to different application scenarios.

5. The multimedia content traceability system based on AI synthesis detection and blockchain evidence storage according to claim 4 is characterized by: The smart contract defines the summary data format, verification process, storage structure and query interface. The contract supports additional fields to facilitate future traceability. The test report can be exported as a PDF file or JSON structure, or can be called by a third-party platform through an API interface for publishing platforms, social networks or judicial institutions to verify content traceability information.

6. A method for a multimedia content traceability system based on AI synthesis detection and blockchain evidence storage according to claim 5, characterized in that: The following steps are involved: S1. User uploads multimedia content: Users upload multimedia content to be tested through the system's front-end interactive interface. Multiple formats, including images, videos, and audio, are supported. The uploaded content can be captured, generated, or collected from third-party platforms. Upon receiving the uploaded content, the system performs basic format verification and metadata extraction, and generates a preliminary hash value for duplicate detection and unique identification. S2. AI Detection Module Analysis: The AI ​​Detection Module, built based on a deep learning algorithm, analyzes uploaded multimedia content, performing normalization processing through the Image / Video Preprocessing Module, extracting relevant features through the Feature Extraction Module, and finally outputting a probability value for whether the content is AI-generated. S3. Generate summary data: After AI detection is completed, the system combines the unique hash value of the multimedia content, the AI ​​synthesis judgment result, the detection timestamp, the content type and basic metadata, the detection device or model identifier, and the user identity into a summary data structure, encapsulates it using a unified encoding method, and protects it against tampering with a digital signature. S4. Blockchain Evidence Storage: Through a smart contract pre-deployed in a consortium chain or public chain, the summary data is submitted to the blockchain node. The contract verifies the data format and signature legitimacy and packages it into a block as transaction data. After the data is successfully written to the blockchain, the block hash and transaction ID identification information are returned; S5. Return of evidence information and credibility: After the on-chain evidence is stored, the system returns a complete test report to the user, including whether the content is AI-synthesized, the probability of synthesis, the blockchain evidence result and on-chain hash, the detection time, the evidence number or transaction ID, the credit score or credibility level, and the query interface link or QR code; S6. Traceability query during content dissemination: The system supports open query interfaces and verification mechanisms. Any dissemination node or recipient can verify the authenticity of the content through hash value comparison, blockchain data reading, API docking and embedding.

7. A method according to claim 6, characterized in that: In step S2, the image / video preprocessing module performs standardization on the uploaded multimedia content, specifically including adjusting the resolution, unifying the number of frames, image slicing, and color normalization operations to ensure that the input meets the model requirements.

8. A method according to claim 7, characterized in that: In step S2, the feature extraction module uses a pre-trained neural network model to extract image features, texture, frequency domain distribution, and artifact trace information, performs frame-level sampling and time series feature extraction on the video content, and combines it with optical flow analysis to enhance the recognition ability of AI-generated dynamic content.

9. A method according to claim 8, characterized in that: In step S2, the discrimination module outputs a probability value of AI-synthesized content based on the trained discrimination model and the deep feature input of the image or video frame; the system sets a threshold for the probability value, and if it is higher than the threshold, it is preliminarily considered that the content is AI-synthesized, and the threshold can be adjusted to adapt to different application scenarios.

10. A method according to claim 9, characterized in that: When reading blockchain data in step S6, by inputting the hash value, transaction ID or QR code, the system queries the corresponding summary data and AI detection results on the chain. The visual front-end supports displaying the detection record, trust level, and storage time content; When API is connected and embedded, the platform operator can embed the verification module into its content management system to automatically identify and label AI-generated content, or provide "authenticity verification" functions for ordinary users.

Images