Privacy in Sensor-Driven Human Data Collection: A Guide for Practitioners Working Paper

Arkadiusz Stopczynski, Riccardo Pietri, Alex Pentland, David Lazer2014

TL;DR

This paper surveys the landscape of privacy practices in sensor-driven human data collection for research, highlighting significant shortcomings in informed consent, data security, and control. It proposes a "Living Informed Consent" framework and actionable items to empower participants and build more trustworthy, sustainable data collection practices.

What Problem Does This Paper Address?

The paper tackles the growing challenge of ensuring privacy in the era of "big data" generated by sensor-driven human data collection. As individuals increasingly share personal information through online platforms and self-tracking devices, and researchers collect unprecedentedly detailed behavioral data, the sensitivity of this information has skyrocketed. However, the authors observe that in many academic research studies, privacy is not adequately addressed. This manifests as unclear study purposes, poorly defined informed consent, and insufficient disclosure of security and sharing protocols, leading to potential misuse of sensitive data and erosion of public trust. The paper aims to provide a comprehensive overview of current practices and propose solutions for practitioners.

Methods

The paper's primary method is a survey and analysis of existing sensor-driven human data collection studies. The authors categorize these studies into generic frameworks and specialized applications. They then analyze these works through the lens of privacy, examining their approaches to:

  • Informed Consent: How participants understand and agree to data collection.
  • Data Security: Techniques used to protect data from unauthorized access.
  • Privacy Implementations: Methods like noise injection and anonymization.
  • Information Ownership and Disclosure: How data is shared and controlled.

Based on this review, the authors propose a novel conceptual framework called "Living Informed Consent," which emphasizes continuous participant engagement, understanding, and control over their data throughout the data lifecycle. They also outline "Privacy Actionable Items" as concrete recommendations for improving current practices.

Datasets

The paper does not introduce new datasets for training or evaluation. Instead, it analyzes and references numerous existing studies that utilize various forms of sensor-driven human data. These datasets are derived from sources such as:

  • Online social networks (e.g., Facebook, Twitter)
  • Smartphone sensors (e.g., GPS, call logs, application usage)
  • Wearable sensors
  • Pervasive environments (e.g., smart homes)
  • Traditional surveys and interviews

Key Insights and Ideas

  • The "all-or-nothing" approach to informed consent is insufficient. Current models often force participants to grant broad permissions without granular control, leading to a false sense of informed consent.
  • "Living Informed Consent" is crucial. Participants need to understand and be able to revisit their data authorizations and control their data throughout its lifecycle, not just at the initial enrollment. This includes the ability to change permissions, request deletion, and audit data access.
  • De-identification techniques are often insufficient. Modern algorithms can re-identify individuals even from anonymized datasets, highlighting the need for more robust privacy measures.
  • Privacy is dynamic and context-dependent. What is considered private changes over time and across cultures, necessitating flexible and adaptable privacy frameworks.
  • Security and privacy are intertwined but distinct. While strong security is necessary for privacy, it doesn't automatically guarantee it. The focus needs to be on how data is used and controlled, not just how it's stored.
  • Reproducibility is a challenge for dynamic privacy. If data can be altered or deleted by users, replicating past research findings becomes difficult, requiring mechanisms like data freezing.
  • Legal and contractual agreements are essential complements to technical solutions. While technical measures are important, the threat of legal and business consequences can be a powerful deterrent against data misuse.

Key Results

The paper doesn't present quantitative results from a specific experiment. Instead, its key "results" are:

  • A comprehensive survey of privacy practices in sensor-driven human data collection, identifying common shortcomings.
  • The proposal of the "Living Informed Consent" framework, offering a more dynamic and participant-centric approach to privacy.
  • A set of "Privacy Actionable Items" providing concrete recommendations for researchers to improve their practices in areas like informed consent, data security, and privacy implementation.
  • An emphasis on the need for reusable tools and shared standards to avoid reinventing privacy solutions for every new study.

Limitations and Future Work

The authors acknowledge several limitations and suggest future directions:

  • Current privacy solutions are often not widely used or easily deployable. Many require trusted computing bases or client extensions that reduce usability.
  • The effectiveness of some privacy techniques is still uncertain. Assured deletion and data expiration, while promising, require further technical scrutiny.
  • The challenge of balancing privacy with reproducibility remains. Mechanisms for data freezing are proposed but need further development.
  • The legal and technical frameworks for data control are still evolving. The paper calls for better alignment between business, legal, and technical dimensions.
  • Future work should focus on developing more robust, reusable, and standardized privacy solutions. This includes building platforms that enable users to visualize and control their information flows and promoting a paradigm shift where users truly own their sensitive information. The authors also suggest that future studies may increasingly leverage existing user data, managed by users themselves, rather than collecting new data.

Summary updated

Abstract

In recent years, the amount of information collected about human beings has increased dramatically. This development has been partially driven by individuals posting and storing data about themselves and friends using online social networks (such as Facebook or Twitter) or collecting their data for selftracking purposes (quantified-self movement). Data regarding human behavior is also collected through the environment, embedded RFIDs, cameras, traffic monitoring, business transactions, etc.. Across the sciences, researchers conduct studies collecting data with an unprecedented resolution and scale. Using computational power combined with mathematical models, such rich datasets can be mined to infer underlying patterns, thereby providing insights into human nature. Much of the data collected is sensitive. It is private in the sense that most individuals would feel uncomfortable sharing their collected personal data publicly. For this reason, the need for solutions to ensure the privacy of the individuals generating data has grown alongside the data collection efforts. Out of all the massive data collection efforts, this paper focuses on efforts directly instrumenting human behavior, and notes that—in many cases—the privacy of participants is not sufficiently addressed. For example, study purposes are often not explicit, informed consent is ill-defined, and security and sharing protocols are only partially disclosed. This paper provides a survey of the work related to addressing privacy issues in research studies that collect detailed sensor data on human behavior. Reflections on the key problems and recommendations for future work are included. We hope the overview of the privacy-related practices in massive data collection studies can be used as a frame of reference for practitioners in the field. Although focused on data collection in an academic context, we believe that many of the challenges and solutions we identify are also relevant and useful for other domains where massive data collection takes place, including businesses and governments.
BibTeX
@misc{stopczynski2014privacy,
  author = {Arkadiusz Stopczynski and Riccardo Pietri and Alex Pentland and David Lazer},
  title = {{Privacy in Sensor-Driven Human Data Collection: A Guide for Practitioners Working Paper}},
  year = {2014},
}
Figure