The method and system use onboard data stream mining for extracting data patterns from data continuously generated by different components of a vehicle. The system stores the data patterns in an onboard micro database and discards the data. The system uses a resource-constrained, small, lightweight onboard data stream management processor, with onboard data stream mining, an onboard micro database, and a privacy-preserving communication module, which periodically and upon request communicates stored data patterns to a remote control center. The control center uses the data patterns to characterize the typical and unusual vehicle health, driving and fleet behavior.