Google extends the coverage and functionality of TensorFlow, its artificial intelligence kit for developers, and has added an automatic learning module focusing on privacy.
Currently, the Automatic Learning Development Kit is widely used by developers to create applications based on this artificial intelligence technique in a much more efficient and faster way. Now the platform announced the integration of a module that will help improve data protection for user data.
TensorFlowPrivacy, is the new module of the development kit that uses artificial intelligence to help protect user data efficiently.
Google remains committed to user security and privacy, Mountain View product manager Carey Radebaugh said: “If we don’t get something like differentiated privacy in TensorFlow, we simply know that it won’t be easy for computers inside and outside Google to use it.
The “differential privacy” mechanism, as the name suggests, is based on complex mathematical models used to train neural networks. The models trained with machine learning or automatic learning make it possible to encrypt users’ personal data in such a way that they are difficult for any attacker to access.
Google uses different privacy policies to ensure that Gmail’s intelligent response never reveals personally identifiable information.
The focus of this methodology is on “mathematical security,” as a Google scientist points out. “It’s a technique that removes identifiable outliers from records without changing the aggregate meaning of that data,” Erlingsson told The Verge.
While this module has been further improved and optimized with the release of this first version, the company expects many AI developers to deploy TensorFlow and take full advantage of the Development Kit.