The paper’s description of a framework for mobile devices to block private information from going to apps includes the problem definition, components, and related experimental results.
Section 1 introduces the guardian-estimator-neutralizer (GEN) framework, a feature learning framework that learns from data to establish tradeoff between user privacy and app utility. The second section gives a formal definition of the problem. Section 3 details the framework’s three components: the neutralizer component is an optimizer with a specific objective function for the tradeoff; the guardian is a feature learning component; and the estimator is for checking accuracy by quantification.
The fourth section presents details of the experiments. Two real-world datasets are used to measure data transformation efficiency and information leakage. The paper ends with related works and a conclusion.
The paper assumes that readers have a background in architectures, machine learning, and optimization algorithms.