An Improved Network Coding-based Secret Sharing Scheme on Programmable Switches

With the development of the internet, multipath transmission technology that has been widely applied enables efficient data transmission via multiple transmission paths. In addition, secret sharing schemes have also begun to find many applications with the rise of big data. By splitting data into multiple sub-blocks with sub-secrets for transmission, this not only enhances security but also reduces the possibility of data loss. Only when a subset of all sub-secrets is collected can the original data be reconstructed. This technique is suitable for the transmission of private data and can also be applied to data backup. In this paper, we have chosen to implement secret sharing schemes on programmable switches, as internet device functionalities gradually increase. First, we encode packets based on random linear network coding, and we modify the generation method of the coefficient matrix to improve the uncertainty of decoding in traditional random linear network coding. Second, we modify the encoding mode of the electronic codebook in the block cipher used in previous random linear network coding. We encode each matrix with a different coefficient matrix to avoid data correlation in the encoded data when the data is highly repetitive. Finally, we generate the sub-secrets to be transmitted by XORing the encoded data with an index matrix. Compared to previous studies, we add redundancy to the data while maintaining the same level of security and guarantee decoding success. The proposed scheme reduces computation of random number generation and multiplication lookup table lookups by 28% and 66%, respectively.