The Accuracy-Privacy Tradeoff of Mobile Crowdsensing

Mohammad Abu Alsheikh, Yutao Jiao, Dusit Niyato, Ping Wang, Derek Leong, Zhu Han

Abstract

Mobile crowdsensing has emerged as an efficient sensing paradigm which combines the crowd intelligence and the sensing power of mobile devices, e.g., mobile phones and Internet of Things (IoT) gadgets. This article addresses the contradicting incentives of privacy preservation by crowdsensing users and accuracy maximization and collection of true data by service providers. We firstly define the individual contributions of crowdsensing users based on the accuracy in data analytics achieved by the service provider from buying their data. We then propose a truthful mechanism for achieving high service accuracy while protecting the privacy based on the user preferences. The users are incentivized to provide true data by being paid based on their individual contribution to the overall service accuracy. Moreover, we propose a coalition strategy which allows users to cooperate in providing their data under one identity, increasing their anonymity privacy protection, and sharing the resulting payoff. Finally, we outline important open research directions in mobile and people-centric crowdsensing.

Introduction

Mobile Crowdsensing (MCS) relies on individuals using their smartphones or IoT devices to collect data—such as GPS coordinates, acceleration, or environmental metrics—for a central service provider. However, a fundamental "Accuracy-Privacy Tradeoff" exists. Service providers require high-quality, "true" data to train accurate machine learning models, while users are often hesitant to share raw data due to privacy concerns.

Most existing incentive models focus purely on monetary rewards for completing a task (e.g., a flat fee per data point), ignoring the privacy cost to the user. This research addresses the challenge of designing a system where users can choose their own privacy levels and receive a fair payment that reflects the actual value their specific, anonymized data adds to the service.

System Framework and Methodology

The proposed framework introduces a truthful marketplace for data through three main components: data anonymization, marginal contribution-based payoffs, and user coalitions.

The system architecture features a mediator that facilitates auctions and data anonymization between user coalitions and service providers.
  1. Data Anonymization: Users protect their data by adding Gaussian noise. The level of noise corresponds to their "privacy level," specifically aiming for t-closeness.
  2. Marginal Contribution Payoff: Instead of a flat fee, the system uses a mechanism based on the Vickrey-Clarke-Groves (VCG) auction. A user’s payoff is calculated as the difference between the model's accuracy with their data included and the accuracy without it. This identifies "pivotal users" who significantly boost accuracy and rewards them accordingly. Conversely, if a user provides data so noisy that it hurts the model's accuracy, they may receive a negative payoff.
  3. Crowdsensing Coalitions: To further enhance privacy, users can form groups (coalitions) to achieve $k$-anonymity, making an individual indistinguishable from $k-1$ others. The total payoff for the group is distributed among members using the Shapley value, ensuring a fair split based on each person's individual contribution to the group's data quality.

Experimental Evaluation and Key Results

The framework was evaluated using a real-world activity recognition dataset consisting of accelerometer data from 36 users. A deep learning architecture was used to develop the prediction service, recognizing activities such as walking, jogging, and sitting.

Payoff allocation decreases as privacy levels increase, illustrating the cost of data anonymization and the risk of negative payoffs for over-anonymized data.

Key Insights

  • Quality Over Quantity: Experiments showed no direct correlation between the amount of data a user provides and the resulting service accuracy. Payoff is strictly tied to data quality and its utility to the specific machine learning model.
  • The "Critical Point": There is a threshold of anonymization beyond which a user's data becomes a liability. At this point, the service provider is incentivized to reject the data entirely, as shown by the negative payoff trends in Figure 6.
  • Truthfulness as a Strategy: Because the payoff is tied to the marginal improvement of the model, the authors demonstrate that "sending true data" (with the least possible noise the user is comfortable with) becomes a weakly-dominant strategy for participants.
  • Fairness in Groups: By using the Shapley value, the system prevents "free-riders" within a coalition. Even when users hide under a single group identity for privacy, the mediator can still reward them based on the underlying quality of their contributions.

Conclusion and Future Work

This framework successfully quantifies the cost of privacy in a data marketplace, achieving a baseline accuracy of 92.5% when users provide raw data. While the current model assumes a single service provider, future research will explore competition among multiple providers and the use of fog computing to move data processing to the "edge," further enhancing privacy and reducing central computational burdens.

Summary updated

Abstract

Mobile crowdsensing has emerged as an efficient sensing paradigm which combines the crowd intelligence and the sensing power of mobile devices, e.g.,~mobile phones and Internet of Things (IoT) gadgets. This article addresses the contradicting incentives of privacy preservation by crowdsensing users and accuracy maximization and collection of true data by service providers. We firstly define the individual contributions of crowdsensing users based on the accuracy in data analytics achieved by the service provider from buying their data. We then propose a truthful mechanism for achieving high service accuracy while protecting the privacy based on the user preferences. The users are incentivized to provide true data by being paid based on their individual contribution to the overall service accuracy. Moreover, we propose a coalition strategy which allows users to cooperate in providing their data under one identity, increasing their anonymity privacy protection, and sharing the resulting payoff. Finally, we outline important open research directions in mobile and people-centric crowdsensing.
BibTeX
@article{alsheikh2017accuracy,
  author = {Mohammad Abu Alsheikh and Yutao Jiao and Dusit Niyato and Ping Wang and Derek Leong and Zhu Han},
  title = {{The Accuracy-Privacy Tradeoff of Mobile Crowdsensing}},
  journal = {IEEE Communications Magazine},
  year = {2017},
  doi = {10.1109/mcom.2017.1600737},
}
Figure