Available at: http://digitalcommons.calpoly.edu/theses/1403
Date of Award
MS in Computer Science
Work in the field of Homomorphic Encryption has exploded in the past 5 years, after Craig Gentry proposed the first encryption scheme capable of performing Homomorphic Encryption. Under the scheme one can encrypt data, perform computations on the encrypted result (without needing the original data), and then decrypt the data to get the result as if the computations had been run on the unencrypted data.
Such a scheme has wide reaching implications for cloud computing. Computations on sensitive data, just like regular data, could now be performed in the cloud with the added security that even the cloud service provider couldn't "see" the secure data. With such a benefit one might ask why the encryption scheme is not used currently? It is because, while Craig Gentry's scheme was theoretically sound, it was not quick. As such, recent work has been in finding ways to speed up the scheme. Several improvements in speed have been made and several implementations of those improved schemes have been developed: one being HElib.
As of now HElib is self described as an "assembly language for HE". Our work focused on creating HEIDE, a Homomorphic Encryption IDE, where researchers could write tests at a high-level. This high-level code is then "compiled" into the operations provided by HElib. HElib, like most encryption schemes, can be configured using different setup parameters. These parameters change the run-time and security of the scheme. As such we have also provided an easy way for researchers to simultaneously run their tests using different setup parameters. To support that, timing and memory metrics are provided for each test so that researchers can determine which parameters worked best.