Search for: All records

Computationally efficient, camera-based, real-time human position tracking on low-end, edge devices would enable numerous applications, including privacy-preserving video redaction and analysis. Unfortunately, running most deep neural network based models in real time requires expensive hardware, making widespread deployment difficult, particularly on edge devices. Shifting inference to the cloud increases the attack surface, generally requiring that users trust cloud servers, and increases demands on wireless networks in deployment venues. Our goal is to determine the extreme to which edge video redaction efficiency can be taken, with a particular interest in enabling, for the first time, low-cost, real-time deployments with inexpensive commodity hardware. We present an efficient solution to the human detection (and redaction) problem based on singular value decomposition (SVD) background removal and describe a novel time- and energy-efficient sensor-fusion algorithm that leverages human position information in real-world coordinates to enable real-time visual human detection and tracking at the edge. These ideas are evaluated using a prototype built from (resource-constrained) commodity hardware representative of commonly used low-cost IoT edge devices. The speed and accuracy of the system are evaluated via a deployment study, and it is compared with the most advanced relevant alternatives. The multi-modal system operates at a frame rate ranging from 20 FPS to 60 FPS, achieves awIoU_0.3score (see Section 5.4) ranging from 0.71 to 0.79, and successfully performs complete redaction of privacy-sensitive pixels with a success rate of 91%–99% in human head regions and 77%–91% in upper body regions, depending on the number of individuals present in the field of view. These results demonstrate that it is possible to achieve adequate efficiency to enable real-time redaction on inexpensive, commodity edge hardware. 

PeerTube is an open-source video sharing platform built as a decentralized alternative to YouTube. With software like Mastodon and Friendica, PeerTube is part of a series of federated social media platforms built partly in response to growing concerns about centralized control and ownership of the incumbent ones. In this paper, we present the first characterization of PeerTube, including its underlying infrastructure and the content being shared on its network. Our findings reveal concerning trends toward centralization that echo patterns observed in other contexts, exacerbated by the limited degree of content replication. PeerTube instances are mostly located in North America and Western Europe, with about 70% hosted in Germany, the USA, and France, and over 50% hosted on the top 7 ***ASes. We also find that over 92% of videos are stored without any redundancy in spite of PeerTube's native support for video redundancy. 

IoT devices like smart cameras and speakers provide convenience but can collect sensitive information within private spaces. While research has investigated user perception of comfort with information flows originating from these types of devices, little focus has been given to the role of the sensing hardware in influencing these sentiments. Given the proliferation of trusted execution environments (TEEs) across commodity- and server-class devices, we surveyed 1049 American adults using the Contextual Integrity framework to understand how the inclusion of cloud-based TEEs in IoT ecosystems may influence comfort with data collection and use. We find that cloud-based TEEs significantly increase user comfort across information flows. These increases are more pronounced for devices manufactured by smaller companies and show that cloud-based TEEs can bridge the previously-documented gulfs in user trust between small and large companies. Sentiments around consent, bystander data, and indefinite retention are unaffected by the presence of TEEs, indicating the centrality of these norms.