Multiparty Computation for Privacy Preserving Multisource Data Mining in Industrial IoT

Within the context of the Industrial Internet of Things (IIoT), interorganizational collaboration and data interchange enable enhanced data analytics and extraction. The joint data mining yields improved data access, analytical precision, and uncovers concealed patterns. However, during the process of joint data mining, there are data privacy concerns such as data leakage and misuse, necessitating the protection of data privacy. In this article, we propose a secure multiparty computation (MPC) approach, aiming to achieve the separation of data ownership and usage which ensures that multiple parties can perform computations without revealing sensitive inputs. Additionally, we utilize Beaver triples to reduce polynomial degrees and lower communication costs. Furthermore, we demonstrate consistency checks and computation verification to attest to the result consistency and data confidentiality under the semihonest adversary model. When evaluated with 20 parties on an Intel i7-10700 CPU, a single 128-bit field multiplication transmits 2.6 KB and completes in 21 ms over a 1 Mbps link. Even at 500 parties the budget remains 62.6 kB (490 ms). Compared with six representative MPC and homomorphic schemes, the proposed design achieves the lowest overall communication cost and competitive computation time. Integrated into a privacy-preserving DBSCAN implementation, the framework clusters 1000 2-D points with plaintext-level accuracy in millisecond-level, demonstrating practical applicability to multisource data mining.