Vaibhav Kulkarni

MobiDict: a mobility prediction system leveraging realtime location data streams

Mobility prediction is becoming one of the key elements of location-based services. In the near future, it will also facilitate tasks such as resource management, logistics administration and urban planning. To predict human mobility, many techniques have been proposed. However, existing techniques are usually driven by large volumes of data to train user mobility models computed over a long duration and stored in a centralized server. This results in inherently long waiting times before the prediction model kicks in. Over this large training data, small time bounded user movements are shadowed, due to their marginality, thus impacting the granularity of predictions. Transferring highly sensitive location data to third party entities also exposes the user to several privacy risks. To address these issues, we propose MobiDict, a realtime mobility prediction system that is constantly adapting to the user mobility behaviour, by taking into account the movement periodicity and the evolution of frequently visited places. Compared to the existing training approaches, our system utilises less data to generate the evolving mobility models, which in turn lowers the computational complexity and enables implementation on handheld devices, thus preserving privacy. We test our system using mobility traces collected around Lake Geneva region from 168 users and demonstrate the performance of our approach by evaluating MobiDict with six different prediction techniques. We find a satisfactory prediction accuracy as compared to the baseline results obtained with 70% of the user dataset for majority of the users.

A mobility prediction system leveraging realtime location data streams: poster

Location-based services today, exceedingly depend on user mobility prediction, in order to push context aware services ahead of time. Existing location forecasting techniques are driven by large volumes of data to train the prediction models in a centralised server. This amounts to considerably long waiting times before the model kicks in. Disclosing highly sensitive location information to third party entities also exposes the user to several privacy risks. To address these issues, we put forth a mobility prediction system, able to provide swift realtime predictions, evading the strenuous training procedure. We enable this by constantly adapting the model to substantive user mobility behaviours that facilitate accurate predictions even on marginal time bounded movements. In comparison to existing frameworks, we utilise less volumes of data to produce satisfactory prediction accuracies. This in turn lowers the computational complexity making implementation on mobile devices feasible and a step towards privacy preservation. Here, only the predicted location can be sent to such services to maintain the utility/privacy tradeoff. Our preliminary evaluations based on real world mobility traces corroborate our hypothesis.

Privacy-Preserving Location-Based Services by using Intel SGX

We are witnessing a rapid proliferation of location-based services, due to the useful context-aware services they provide their users. However, sharing sensitive location traces with untrusted service-providers has many privacy implications. Although, user-data monetization is the core economic model of such services, offering private services to concerned users will be a beneficial functionality in the coming years. Existing solutions include location perturbation, k-anonymity and cryptographic primitives that trade service accuracy or latency for enhanced user privacy. We introduce a novel approach for privacy preserving location-based services by using the Intel Software Guard eXtensions (SGX). We implement a simple location-based service using SGX and gauge its performance in terms of efficiency and effectiveness, in comparison with its bare-metal implementation. Our evaluation results show that SGX contributes a marginal overhead but also provides near-to-the-perfect results in contrast to spatial cloaking with k-anonymity whose performance deteriorates as the degree of desired privacy increases. We show that hardware-based trusted execution-environments are a promising alternative for offering proactive and de-facto location-privacy in the context of location-based services.

Generating synthetic mobility traffic using RNNs

Mobility trajectory datasets are fundamental for system evaluation and experimental reproducibility. Privacy concerns today however, have restricted sharing of such datasets. This has led to the development of synthetic traffic generators, which simulate moving entities to create pseudo-realistic trajectory datasets. Existing work on traffic generation, superficially matches a-priori modeled mobility characteristics, which lacks realism and does not capture the substantive properties of human mobility. Critical applications however, require data that contains these complex, candid and hidden mobility patterns. To this end, we investigate the effectiveness of Recurrent Neural Networks (RNN) to learn these hidden patterns contained in an original dataset to produce a realistic synthetic dataset. We observe that, the ability of RNNs to learn and model problems over sequential data having long-term temporal dependencies is ideal for capturing the inherent properties of location traces. Additionally, the lack of intuitive high-level spatiotemporal structure and instability, guarantees trajectories that are different from the ones seen in the training dataset. Our preliminary evaluation results show that, our model effectively captures the sleep cycles and stay-points commonly observed in the considered training dataset, along with preserving the statistical characteristics and probability distributions of the movement transitions. Although, many questions remain to be answered, we show that generating synthetic traffic by learning the innate structure of human mobility through RNNs is a promising approach.

Controlled Interference Generation for Wireless Coexistence Research

In recent years, we have witnessed a proliferation of wireless technologies and devices operating in the unlicensed bands. The resulting escalation of wireless demand has put enormous pressure on available spectrum. This raises a unique set of communication challenges, notably co-existence, Cross Technology Interference (CTI), and fairness amidst high uncertainty and scarcity of interference-free channels. Consequently, there is a strong need for understanding and debugging the performance of existing wireless protocols and systems under various patterns of interference. Therefore, we need to augment testbeds with tools that can enable repeatable generation of realistic interference patterns. This would primarily facilitate wireless coexistence research experimentation. The heterogeneity of the existing wireless devices and protocols operating in the unlicensed bands makes interference hard to model. Meanwhile, researchers working on wireless coexistence generally use interference generated from various radio appliances. The lack of a systematic way of controlling these appliances makes it inconvenient to run experiments, particularly in remote testbeds. In this paper, we present a Controlled Interference Generator (CIG) framework for wireless networks. In the design of CIG, we consider a unified approach that incorporates a careful selection of interferer technologies (implemented in software), to expose networks to realistic interference patterns. We validate the resemblance of interference generated by CIG and interference from represented RF devices, by showing the accuracy in temporal and spectral domains.