Jindan Zhu

An adaptive privacy-preserving scheme for location tracking of a mobile user

Many popular mobile applications require the continuous monitoring and sharing of a mobile users location. However, exploiting a users location leads to disclosing sensitive information about the users daily activity. Several location privacy-preserving schemes have been proposed, but it remains challenging for a user to achieve visibility of the associated threats as well as to control the impact of those threats. This paper presents an adaptive location privacy-preserving system (ALPS) that allows for a user to control the level of privacy disclosure with different quality of location-based service (LBS). We have identified key attack models on location tracking using powerful map-matching algorithms, and then defined a scheme that allows a user to control the privacy of tracking information. We have implemented ALPS on Android OS and evaluated the implementation extensively via trace-based simulation, showing the effectiveness of user-controllable privacy preservation.

Using Deep Learning for Energy Expenditure Estimation with wearable sensors

Energy Expenditure (EE) Estimation is an important step in tracking personal activity and preventing chronic diseases such as obesity, diabetes and cardiovascular diseases. Accurate and online EE estimation using small wearable sensors is a difficult task, primarily because most existing schemes work offline or using heuristics. In this work, we focus on accurate EE estimation for tracking ambulatory activities (walking, standing, climbing upstairs or downstairs) of individuals wearing mobile sensors. We use Convolution Neural Networks (CNNs) to automatically detect important features from data collected from triaxial accelerometer and heart rate sensors. Using CNNs, we find a significant improvement in EE estimation compared to other state-of-the-art models. We compare our results against state-of-the-art Activity-Specific Linear Regression as well as Artificial Neural Networks (ANN) based models. Using a universal CNN model, we obtain an overall low Root Mean Square Error (RMSE) of 1.12 which is 30% and 35% lower than existing models. The results were calibrated against a COSMED K4b2 indirect calorimeter readings.

Using Smartphone Sensors for Improving Energy Expenditure Estimation

Energy expenditure (EE) estimation is an important factor in tracking personal activity and preventing chronic diseases, such as obesity and diabetes. Accurate and real-time EE estimation utilizing small wearable sensors is a difficult task, primarily because the most existing schemes work offline or use heuristics. In this paper, we focus on accurate EE estimation for tracking ambulatory activities (walking, standing, climbing upstairs, or downstairs) of a typical smartphone user. We used built-in smartphone sensors (accelerometer and barometer sensor), sampled at low frequency, to accurately estimate EE. Using a barometer sensor, in addition to an accelerometer sensor, greatly increases the accuracy of EE estimation. Using bagged regression trees, a machine learning technique, we developed a generic regression model for EE estimation that yields upto 96% correlation with actual EE. We compare our results against the state-of-the-art calorimetry equations and consumer electronics devices (Fitbit and Nike+ FuelBand). The newly developed EE estimation algorithm demonstrated superior accuracy compared with currently available methods. The results were calibrated against COSMED K4b2 calorimeter readings.

WearIA: Wearable device implicit authentication based on activity information

Privacy and authenticity of data pushed by or into wearable devices are of important concerns. Wearable devices equipped with various sensors can capture users activity in fine-grained level. In this work, we investigate the possibility of using users activity information to develop an implicit authentication approach for wearable devices. We design and implement a framework that does continuous and implicit authentication based on ambulatory activities performed by the user. The system is validated using data collected from 30 participants with wearable devices worn across various regions of the body. The evaluation results show that the proposed approach can achieve as high as 97% accuracy rate with less than 1% false positive rate to authenticate a user using a single wearable device. And the accuracy rate can go up to 99.6% when we use the fusion of multiple wearable devices.

Navigating in Signal Space: A Crowd-Sourced Sensing Map Construction for Navigation

Indoor navigation is typically achieved by an operational localization system among a range of location-based services it provides. However, the construction of a localization map, which is prerequisite for binding sensed observation in sensing space to individual locations in geographic space, remains a challenging task to date. In this work, we propose an indoor navigation system that alleviates the need for constructing a localization map and instead provides navigation in signal space. The main idea behind our approach is to construct a sensing map consisting of signal observations (WiFi Clusters) and connecting dead-reckoning segments obtained through mobile sensing capabilities (traces of accelerometer and digital compass reading). To this end, we design a prototype to demonstrate effective construction of such sensing map with energy-efficient sensors and crowd-sourcing, and its ability to support accurate navigation.

Verification of User-Reported Context Claims with Context Correlation Model

Context-aware services nowadays offer incentive to user-reported context information , which inevitably solicits malicious users to cheat by submitting fabricated context claims. Conventional countermeasures based on Trusted Computing Base typically focus on particular context of interest, while disregarding the availability of various types of context information and the intrinsic correlation among them. In this work we propose a context claim verification scheme that interrogates correlated contexts of multiple dimensions to corroborate or contradict the reported context. Specifically, it first learns and models the context correlation with a Bayesian Multinet. Given a claim consisting of reported context and witnessing evidence, the scheme performs Bayesian inference with the evidence to verify the reported context. The verification process is light-weight, and can be applied to arbitrary types of context with a single model learnt. Evaluations on Reality Mining dataset and synthetic dataset validates choice of Multinet for data modeling, and demonstrate the feasibility of our scheme in context verification.

Time and energy efficient localization

Time-critical Location Based Service (LBS) applications in mobile ad hoc networks require fast localization. The conventional localization techniques are, unfortunately, unsuitable for such applications, for they neglect the time needed for localization. As a result, time-critical information may become obsolete, and the mobile users such as vehicles may have moved to new locations before the localization procedure is completed. To address this issue, we formulate a notion of On-Demand Fast Localization (ODFL) and devise a framework to implement this concept over existing routing protocols in MANETs. We present analytical and simulation results to demonstrate that ODFL can significantly reduce the time solely needed for localization before starting time-critical applications. Moreover, we show that ODFL can also improve location privacy and reduce energy consumptions.