Machine learning on encrypted data can address concerns related toprivacy and legality of sharing sensitive data with untrustworthy service providers . Fully Homomorphic Encryption (FHE) is a promising technique toenable machine learning and inferencing while providing strict guarantees against information leakage . We consider a Machine Learning as a Service (MLaaS) scenario where both input data and model parameters are secured using FHE . Our empirical study shows that choice of aforementioned designparameters result in significant trade-offs between accuracy, security level,and computational time . Encrypted inference experiments on the MNIST dataset show other design choices such as ciphertext packing strategy and parallelization using multithreading are also critical in determining the speed of the inference process . Other design choices also crucial in determining speed and latency of the . inference process are also crucial to determining the inferenication process, such as multithreadreading using multitasking using multiterreaders using multitsurfication using multi-discrimination using multididiation using multityotryryryrsrsrsnithreaders and multitreaders are also key in determining accuracy, latency and latency . The authors conclude that the FHE

Author(s) : Nayna Jain, Karthik Nandakumar, Nalini Ratha, Sharath Pankanti, Uttam Kumar

Links : PDF - Abstract

Code :
Coursera

Keywords : determining - data - learning - encrypted - process -

Leave a Reply

Your email address will not be published. Required fields are marked *