Securely perturb big data by using inner product

Data privacy is one important consideration in the era of big data. This paper considers the problem of protecting data privacy when sharing external data to the other parties. The major difficulty is how to perturb the unique identifier in the external data source. We first give a complete system model for this problem. Specifically, we propose to use inner product to replace the classical method of using hash function. Then we give a security proof for our construction based on the Goldreich-Levin theorem. Finally, we implement our scheme and demonstrate that our solution is approximately 6 times more efficient than the classical hash function method.