Privacy-preserving big data analytics for cyber-physical systems

Abstract

Cyber-physical systems (CPS) generate big data collected from combining physical and digital entities, but the challenge of CPS privacy-preservation demands further research to protect CPS sensitive information from unauthorized access. Data mining, perturbation, transformation and encryption are techniques extensively used to preserve private information from disclosure whilst still providing insight, but these are limited in their effectiveness in still allowing high-level analysis. This paper studies the role of big data component analysis for protecting sensitive information from illegal access. The independent component analysis (ICA) technique is applied to transform raw CPS information into a new shape whilst preserving its data utility. The mechanism is evaluated using the power CPS dataset, and the results reveal that the technique is more effective than four other privacy-preservation techniques, obtaining a higher level of privacy protection. In addition, the data utility is tested using three machine learning algorithms to estimate their capability of identifying normal and attack patterns before and after transformation.