Recent hardware and software advances allow efficient local machine learning inference on edge devices such as smart cameras. This feature addresses users’ privacy concerns, as they no longer need to upload sensitive raw data, such as images and videos, to cloud servers for data analysis. This paper studies potential privacy breaches in this emerging computing paradigm. Specifically, we show that it is possible to manipulate production neural network models to turn them into spyware. The inference results of these models are almost the same as they are expected to be, but can convey information about the raw input content, allowing the curious cloud service provider managing the edge devices to covertly spy on the private information without being noticed. In the end, we discuss possible defense approaches.

The increasing use of camera streams on mobile systems has raised significant privacy concerns due to unauthorized visual data access by applications. Existing solutions either burden users with excessive interaction or lack semantic understanding of contextual privacy norms. This paper introduces PrivacyAgent, a novel visual privacy protection framework leveraging multimodal large language models (LLMs) to enable context-aware and fine-grained privacy control on mobile systems. PrivacyAgent intercepts camera streams via a virtualized I/O layer and restricts untrusted apps to privacy-compliant content with minimal user overhead.

Fingerprint scanners are widely used in security applications. In this work, we demonstrate that capacitive fingerprint scanners can also serve as universal communication devices. In this demonstration, we show that a reliable data transmission channel can be established between a wearable device and the fingerprint scanner, using the human body as the medium. We also envision several promising applications that arise from this new opportunity.

As assisted and autonomous driving systems become more prevalent, the need for accurate interpretation of road traffic signs is critical for driving safety and functionality. Current camera-based recognition methods face challenges due to the variability of traffic signs and environmental conditions, leading to potential inaccuracies. To address this, we propose LiDARMarker, a type of machine-readable traffic sign using infrared materials, making it invisible to human drivers but detectable by LiDAR-equipped vehicles. This paper introduces the design, fabrication, and efficient decoding methods of LiDARMarker. LiDARMarker is tailored to the emerging capabilities and needs of modern vehicles, enhancing their ability and accuracy in traffic sign recognition while avoiding interference with human drivers. Through the proposal of LiDARMarker, we aim to inspire the rethinking of the design of traffic sign systems in the context of modern vehicles.

This paper presents EMU, a framework that enables the emulation, snipping, and multiplexing of LoRa chirps on commercial IoT devices equipped with low-power sub-GHz transceivers, including those supporting LoRa itself. Chirp snipping consists in artificially removing a sequence of chips and in putting the radio in low-power mode, which allows to reduce energy consumption while still commu-nicating reliably. Chirp multiplexing exploits the gaps introduced by chirp snipping to transmit portions of another chirp on a sep-arate channel, which allows to concurrently transmit two LoRa packets and to increase the throughput. We build EMU as a modu-lar framework and implement support for off-the-shelf LoRa and non-LoRa transceivers. We then evaluate its performance by com-paring the reliability, efficiency, and receiver sensitivity achieved by EMU with that of traditional LoRa for different physical layer settings. We finally showcase EMU's ability to send packets over two channels simultaneously, thereby improving the uplink throughput of LoRaWan, and demonstrate that even non-LoRa transceivers employing EMU can communicate to a LoRaWan gateway, enabling new use cases and expanding the applicability of LoRa technology.

Recently, several datasets shedding light on connectivity aspects in real-world LoRa networks have been provided to the community. However, they typically only involve a limited number of nodes, deal with unidirectional communication only, or focus on very specific physical layer settings. More importantly, existing datasets typically lack fine-grained environmental information such as the temperature in the surroundings of each node, which is known to have a strong impact on communication performance. In this work, we provide the community with a comprehensive dataset that fills all these gaps. We have collected detailed connectivity information in an outdoor LoRa network composed of 21 nodes for more than four months. Our dataset does not only focus on network-level performance (e.g., the average number of correctly-exchanged packets), but sheds light on link-level information such as the received signal strength, signal-to-noise ratio, and the number of available neighbours over time. We further collect environmental information from an online weather site, as well as the on-board temperature of each node in the network, which varies considerably across the deployed locations. We collect all this information while perpetually changing physical layer settings such as the spreading factor and the RF channel. A preliminary analysis of our dataset, which is available in Zenodo1, reveals that temperature has a significant correlation with the link quality and connectivity in the outdoor LoRa network, confirming the findings of earlier studies.