Data Obfuscation for Privacy and Confidentiality in Cloud Computing

Abstract

This paper proposes a data obfuscation approach in outsourcing matrix multiplication to cloud computing. It is primarily based on splitting the rows and columns of matrices to alter their actual dimension coupled with adding random noise and shuffling in order to ensure confidentiality and privacy. In our approach, obfuscated matrices are sent to servers without any public key encryption. While it computes on matrices, the server is unable to extract or derive actual values either from obfuscated matrices or from computed multiplication results. Whereas, clients can extract actual computed values using a very insignificant computing effort from results produced by the server.