Biblio

The proliferation of mobile devices, equipped with numerous sensors and Internet connectivity, has laid the foundation for the emergence of a diverse set of crowdsourcing services. By leveraging the multitude, geographical dispersion, and technical abilities of smartphones, these services tackle challenging tasks by harnessing the power of the crowd. One such service, Crowd GPS, has gained traction in the industry and research community alike, materializing as a class of systems that track lost objects or individuals (e.g., children or elders). While these systems can have significant impact, they suffer from major privacy threats. In this paper, we highlight the inherent risks to users from the centralized designs adopted by such services and demonstrate how adversaries can trivially misuse one of the most popular crowd GPS services to track their users. As an alternative, we present Techu, a privacy-preserving crowd GPS service for tracking Bluetooth tags. Our architecture follows a hybrid decentralized approach, where an untrusted server acts as a bulletin board that collects reports of tags observed by the crowd, while observers store the location information locally and only disclose it upon proof of ownership of the tag. Techu does not require user authentication, allowing users to remain anonymous. As no user authentication is required and cloud messaging queues are leveraged for communication between users, users remain anonymous. Our security analysis highlights the privacy offered by Techu, and details how our design prevents adversaries from tracking or identifying users. Finally, our experimental evaluation demonstrates that Techu has negligible impact on power consumption, and achieves superior effectiveness to previously proposed systems while offering stronger privacy guarantees.

The proliferation of Internet-of-Things (IoT) devices within homes raises many security and privacy concerns. Recent headlines highlight the lack of effective security mechanisms in IoT devices. Security threats in IoT arise not only from vulnerabilities in individual devices but also from the composition of devices in unanticipated ways and the ability of devices to interact through both cyber and physical channels. Existing approaches provide methods for monitoring cyber interactions between devices but fail to consider possible physical interactions. To overcome this challenge, it is essential that security assessments of IoT networks take a holistic view of the network and treat it as a "system of systems", in which security is defined, not solely by the individual systems, but also by the interactions and trust dependencies between systems. In this paper, we propose a way of modeling cyber and physical interactions between IoT devices of a given network. By verifying the cyber and physical interactions against user-defined policies, our model can identify unexpected chains of events that may be harmful. It can also be applied to determine the impact of the addition (or removal) of a device into an existing network with respect to dangerous device interactions. We demonstrate the viability of our approach by instantiating our model using Alloy, a language and tool for relational models. In our evaluation, we considered three realistic IoT use cases and demonstrate that our model is capable of identifying potentially dangerous device interactions. We also measure the performance of our approach with respect to the CPU runtime and memory consumption of the Alloy model finder, and show that it is acceptable for smart-home IoT networks.