03 Apr Study Says Machine Learning Can be Taught Privacy
In a new paper, researchers from the University of Texas recently claimed to have developed a system that enables the use of cloud machine-learning-as-a-service while preserving privacy.
MLaaS currently require all training data to be revealed to the service operator, which could expose it to risk from nefarious actors. In their paper called “Chiron: Privacy-preserving Machine Learning as a Service” researcher Tyler Hunt and his team says that they used Intel’s Software Guard Extensions (SGX) and uses it for the Ryoan sandbox to protect the user’s training data, as well as trained model queries and outputs, while in the cloud.
However, the researchers revealed that Chiron cannot be used with GPUs, which will make implementation far from ideal until GPU manufacturers decide to explore security measures more seriously.