Research on Cloud Computing Security & Privacy Protection of Massive Unstructured Multi-Modal Data

Abstract

With the rapid development of big data and AI technologies, the sharing and opening-up, development and utilization of big data from government, medical becomes into one of our priorities. However, a problem that has puzzled us for a long time is that, the usage rate of massive unstructured multi-modal data as an absolute main body is still very low, which seriously restricts the circulation of data factors and faces the increasingly serious challenge of security-compliance. To speed up the circulation process and value release of data factors, this paper proposed a security-compliant and lightweight cloud computing model for massive unstructured multi-modal data. The implementation method includes multi-modal feature recognition and extraction from unstructured big data with NLP, CNN, image processing, audio processing technologies, fast and safe cipher-text calculation, and cipher-text retrieval with HE (homomorphic encryption) against the feature index database, and multi-modal data fusion, etc. With the above model & algorithms including the optimization solution of the HE algorithm, we verified the “compute-able but invisible” of massive cross-media unstructured data, and proved the security, validity, and time efficiency.