Leveraging graphical models to improve accuracy and reduce privacy risks of mobile sensing

Abstract

The proliferation of sensors on mobile phones and wearables has led to a plethora of context classifiers designed to sense the individual's context. We argue that a key missing piece in mobile inference is a layer that fuses the outputs of several classifiers to learn deeper insights into an individual's habitual patterns and associated correlations between contexts, thereby enabling new systems optimizations and opportunities. In this paper, we design CQue, a dynamic bayesian network that operates over classifiers for individual contexts, observes relations across these outputs across time, and identifies opportunities for improving energy-efficiency and accuracy by taking advantage of relations. In addition, such a layer provides insights into privacy leakage that might occur when seemingly innocuous user context revealed to different applications on a phone may be combined to reveal more information than originally intended. In terms of system architecture, our key contribution is a clean separation between the detection layer and the fusion layer, enabling classifiers to solely focus on detecting the context, and leverage temporal smoothing and fusion mechanisms to further boost performance by just connecting to our higher-level inference engine. To applications and users, CQue provides a query interface, allowing a) applications to obtain more accurate context results while remaining agnostic of what classifiers/sensors are used and when, and b) users to specify what contexts they wish to keep private, and only allow information that has low leakage with the private context to be revealed. We implemented CQue in Android, and our results show that CQue can i) improve activity classification accuracy up to 42%, ii) reduce energy consumption in classifying social, location and activity contexts with high accuracy(>90%) by reducing the number of required classifiers by at least 33%, and iii) effectively detect and suppress contexts that reveal private information.

On Architectural Qualities and Tactics for Mobile Sensing

“…Examples: Toftkjaer et al [50] present methods to improve QoS for indoor positioning by fusing input from inertial sensors, GPS sensors, and building models using particle filters. Parate et al [34] present CQue fusing output of individual classifiers and sensors to derive an individual's habitual patterns and associated correlations with context.…”

“…For testing, [35] presents debugging tools for energy-related bugs, and [25] describes a testing framework, which allows for automatically replicating heterogeneous sensor data. Finally, [34] addresses security by means for privacy protection. However, further detailing scenarios and tactics for these attributes in the context of mobile sensing is subject to future work.…”

Location Anonymization With Considering Errors and Existence Probability

“…Many algorithms [32], [33] for protecting location data and attributes of users other than previously mentioned research studies have been proposed, but they have not considered the location error.…”

A stochastic game for privacy preserving context sensing on mobile phone

“…Specifically, smartphone users usually transit between different contexts (e.g., a user goes to a particular hospital after eating at a coffee shop), whose sensitivities are different to the users. Moreover, the contexts are usually correlated, which has already been studied for different goals [8]- [10]. Thus, the adversaries can learn the connections between contexts by exploiting the temporal correlations, and then use such correlations to infer user's sensitive contexts based on their observations on nonsensitive contexts.…”

“…On the one hand, the untrusted application extracts the user's context using certain context recognition approaches (e.g. [8], [10]). On the other hand, the untrusted application leak the modified sensing data to an adversary.…”