Predicting Privacy Attitudes Using Phone Metadata

Abstract

With the increasing usage of smartphones, there is a corresponding increase in the phone metadata generated by individuals using these devices. Managing the privacy of personal information on these devices can be a complex task. Recent research has suggested the use of social and behavioral data for automatically recommending privacy settings. This paper is the first effort to connect users' phone use metadata with their privacy attitudes. Based on a 10-week long field study involving phone metadata collection via an app, and a survey on privacy attitudes, we report that an analysis of cell phone metadata may reveal vital clues to a person's privacy attitudes. Specifically, a predictive model based on phone usage metadata significantly outperforms a comparable personality features-based model in predicting individual privacy attitudes. The results motivate a newer direction of automatically inferring a user's privacy attitudes by looking at their phone usage characteristics.

Automatic assessment of mental health using phone metadata

“…An individual's interaction behavior might provide prescriptive clues to her mental and physical processes. Recent literature on phone‐based latent attribute identification has used a number of such behavioral features (De Choudhury et al, 2013; Ghosh & Singh, ). Based on this literature we consider the following features: In Out Ratio, Missed Calls Percentage, Call Response Rate, Call Response Latency, and Longest Call Percentage.…”

“…An individual's interaction behavior might provide prescriptive clues to her mental and physical processes. Recent literature on phone-based latent attribute identification has used a number of such behavioral features (De Choudhury et al, 2013;Ghosh & Singh, 2016…”

“…For instance, a combination of an animal's behavioral traits (e.g., exploration, loyalty) have been demonstrated to have significant predictive power on an animal's social stature, survivability, metabolism rates, reproductive success, and other life outcomes (Sih et al, 2004;Wilson et al, 1994). Similarly, recent literature on mobile computing has connected diversity of contacts and novelty over time with individual outcomes (Ghosh & Singh, 2016;Singh et al, 2013). Hence, we have identified the following features to characterize diversity and novelty in the participant's contacts.…”

Inteligencia Artificial Y Justicia Penal: Más Allá De Los Resultados Lesivos Causados Por Robots

“…Bajo esta denominación se pueden incluir numerosas fuentes de nuevos tipos datos derivados de la actividad de los usuarios con las TIC, y que integran en sus procesos analíticos variables como por ejemplo el medio de creación de los datos, finalidad de los datos, hora y fecha de creación, creador o autor de los datos, ubicación en una red informática donde se crearon los datos, normas utilizadas, tamaño de archivo, calidad de los datos, fuente de los datos, proceso utilizado para crear los datos, entre otros metadatos, que han mostrado una igual o superior capacidad para la identificación de sujetos concretos 101 . Esta miopía técnica del legislador podría quedar parcialmente justificada ante las enormes dificultades existentes para la cuantificación y calificación de un número de atributos o metadatos altamente fluctuante y que es resultado de las constantes actualizaciones de las plataformas digitales, especialmente de redes sociales 102 , y cuya inobservancia, sea cual sea el fin al que respondan, puede comprometer la correcta aplicabilidad de los principios básicos de protección de datos, esto es, la legalidad, equidad y transparencia, tan relevantes en un ámbito tan sensible como es el de la prevención e investigación del delito.…”

Towards a Broadening of Privacy Decision-Making Models: The Use of Cognitive Architectures

“…Ghosh and Singh [17] provide one of the use cases for applying machine learning to build a predictive model of privacy concerns. Feeding phone usage metadata, refined through classification algorithms, they manage to achieve enough accuracy (for their classification design) to predict users' privacy attitude category.…”