Research On Security Supervision Mechanisms For Consortium Blockchain

With the rapid development of blockchain technology,the consortium blockchain has been widely empowered in various fields of economy and society due to its privacy protection capability,regulatability,and high performance.It has become the leading technical route for developing blockchain in China.However,to further realize the independent and controllable consortium blockchain,the corresponding security supervision technology needs to be studied in depth to strengthen the risk prevention and control of the consortium blockchain and ensure the healthy development of the blockchain industry.When performing the supervisory function,the supervisory authority in the consortium blockchain network is usually undertaken by multiple nodes from different interest groups to prevent the control of solid organizations and avoid problems such as single point of failure.There are still the following two problems in the above-mentioned supervisory agencies’security supervision of the data on the chain.First,due to multiple supervisors,when the encrypted transaction data on the chain needs to be audited,there is a problem of low efficiency in the collaborative audit of private data.The current research on consortium chain data privacy and auditing does not involve the analysis of collaborative auditing of private data in the multi-supervisor scenario,and the proposed single-regulator privacy data audit scheme is difficult to deploy for the landing consortium blockchain application.Second,since realistic false information is difficult to be effectively identified and filtered by supervisors,there is still a risk of uploading public information containing fake content.The existing research on monitoring and governance of data content on the blockchain cannot prevent fake information from being uploaded to the chain,and the subjectivity and cost of the adopted trust model are not fully considered.Given the above problems in the security supervision of the consortium blockchain,this thesis designs and implements an efficient collaborative audit mechanism for encrypted on-chain data based on the group key method.It takes fake video information as an example,structures,and implements an effective regulatory mechanism for video data uploading based on the deep learning method.Specifically,it mainly includes the following three contents:1.Aiming at the low efficiency of collaborative audit of private data,an efficient co-audit mechanism for privacy-preserving data on consortium blockchains based on group key agreement is proposed.This mechanism allows transaction parties to use a one-time session key to encrypt transaction data,thereby achieving the effect of on-chain data privacy protection.By deeply coupling the asymmetric group key agreement technology with the blockchain,members of the supervisor group can generate a shared group encryption key and their own group decryption key.The session key is encrypted only once by the group encryption key and stored on the consortium blockchain along with the encrypted transaction.Each supervisor can obtain the plaintext content of the private transaction by combining the data on the chain and his group decryption key,thereby realizing co-audit of privacy-preserving data.The mechanism also takes into account dynamic changes in the group of supervisors.Finally,the experimental results show that,compared with the traditional scheme,the mechanism proposed significantly improves the efficiency of co-audit of privacy-preserving data by multiple supervisors on the consortium blockchain with moderate overhead.2.Aiming at the problem that deepfake videos are uploaded and spread on the chain and are difficult to supervise,the regulation mechanism for harmful video data uploading to the blockchain based on deep learning is proposed.The videos on the chain are divided into two categories:original videos and secondary edited videos.The complete data will be stored on the IPFS network to reduce the storage overhead of the blockchain.The Deepfake detection model based on deep learning is used for the original video to check whether it is Deepfake.The detection model is based on the Meso Net-4 model^[63],and the local maximum ECR algorithm is used to replace the original random frame selection strategy in face extraction.The improved approach reduces data redundancy and increases the diversity of video frame selection.The secondary edited video needs to be checked for legal authorization.The use of the chaincode and the tamper-proof records on the blockchain will be used to test the validity of the authorization.The mechanism has designed a complete process from video classification,storage,review,and post-event supervision,enabling supervisors to effectively conduct penetrating supervision of video information on the consortium blockchain,while ensuring that only legal video data can be successfully stored on the chain.The experimental results show that this mechanism can effectively regulate all kinds of video data with moderate overhead and solve the problem of harmful video information spreading on the consortium blockchain.3.Designed and implemented the consortium blockchain security supervision prototype system.The design details of the prototype system are described from the modules of blockchain and IPFS maintenance,plaintext data release and plaintext data supervision.The main functions of each module are described,and the operation results of the designed method in the system are displayed in a graphical interface.