Research on key technologies of tripartite privacy computing system for sea-rail intermodal transport

In the era of big data, data has become a key production factor for sea-rail intermodal transport, and the efficient circulation of data elements can effectively empower the innovative development of sea-rail intermodal transport. In order to enhance its development level, the railway sector urgently needs to mine potential sources of goods suitable for sea-rail intermodal transport based on its own capacity, and then develop marketable transport products and marketing strategies in the data of railways, ports, and customs, so as to enhance the willingness of shippers to adopt them. Currently, sea-rail intermodal transport data is shared in various ways, mainly including the standardisation mechanism based on the ‘one-order system’ and ‘one-box system’, EDI message exchange and other technical paths, but most of them are the interoperability of the basic production information, which is not enough to support the port schedules, manifests, railway capacity flow, and customs statistics. However, it is not enough to support the detailed analysis of key data such as port schedule, manifest, railway capacity flow, and customs statistics, which involves a large amount of cargo owner's privacy and commercially sensitive data, and the traditional sharing mechanism has significant deficiencies in terms of privacy protection and trust construction. Privacy computing technology provides a new path for secure data sharing. In this paper, we analyse the demand for tripartite interconnection among railways, ports and customs, and design a tripartite privacy computing interconnection framework based on privacy computing, an interconnection implementation process, and a privacy protection implementation process using technologies such as federated learning, multi-party security computing and blockchain. Application validation shows that the architecture can achieve joint training of railways, ports and customs under the premise of privacy computing protection, and the potential cargo model introduces a complete privacy computing framework with almost no loss in classification performance.