Privacy-preserving Linear Models over Homomorphically Encrypted Data
The data involved in machine learning are often confidential as, eg. medical or federal data. The security and authenticity of such data sets needs to be guaranteed. In this thesis we implement logistic and linear regression using the homomorphic Fan-Vercauteren encryption scheme to tackle this problem. Homomorphic encryption allows us to make predictions on the encrypted data without decrypting it. We then examine the feasibility and runtime performance of our implementation.
Privacy-preserving Linear Models over Homomorphically Encrypted Data
| Supervisor(s): | Dr. Huang Xiao |
| Status: | finished |
| Topic: | Others |
| Author: | Melissa Paul |
| Submission: | 2017-11-15 |
| Type of Thesis: | Bachelorthesis |
| Proof of Concept | No |
Thesis topic in co-operation with the Fraunhofer Institute for Applied and Integrated Security AISEC, Garching
|
|
Astract:The data involved in machine learning are often confidential as, eg. medical or federal data. The security and authenticity of such data sets needs to be guaranteed. In this thesis we implement logistic and linear regression using the homomorphic Fan-Vercauteren encryption scheme to tackle this problem. Homomorphic encryption allows us to make predictions on the encrypted data without decrypting it. We then examine the feasibility and runtime performance of our implementation. |
|