Data confidentiality in cloud-based pervasive system

Abstract

Data confidentiality and privacy is a serious concern in pervasive systems where cloud computing is used to process huge amount of data such as matrix multiplications typically used in HPC. Due to limited processing capabilities, smart devices need to rely on cloud servers for heavy-duty computations such as matrix multiplication. Conventional security mechanisms such as public key encryption is not an option to safeguard data from cloud servers to see them. Ensuring client data confidentiality in cloud computing can be achieved using data obfuscating techniques instead of encryption. In a matrix multiplication application, clients can protect their data from dishonest or curious cloud servers which perform multiplication operations on matrices without `knowing or seeing' actual values of input matrices. In our approach, we introduce random noise to the data, and generate several matrices randomly from each matrix in order to cloak data from cloud servers. The main idea is to mask the data as well as confuse the cloud server so it is unable to derive or guess the actual values of matrices as well as computer results.