PPMark: An Architecture to Generate Privacy Labels Using TF-IDF Techniques and the Rabin Karp Algorithm

Abstract

Layman and non-layman users often have difficulties to understand privacy policy texts. The amount of time spent on reading and comprehending a policy poses a challenge to the user, who rarely pays attention to what he or she is agreeing to. Given this scenario, this paper aims to facilitate privacy policy terms presentation regarding data collection and sharing by introducing a new format called Privacy Label. Using natural language processing techniques, a model able to extract information about data collection in privacy policies and present them in an automated and easy-to-understand way to the user was built. To validate this model we used a precision assessment method where the accuracy of the extracted information was measured. The precision of our model was 0.685 (69%) when recovering information regarding data handling, making it possible for the final user to understand which data is being collected without reading the whole policy. The PPMark architecture can facilitate the notice-and-choice by presenting privacy policy information in an alternative way for online users.