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.

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.

Millimeter-wave (mmWave) radar is an effective sensor for human pose estimation, particularly in challenging environments where vision-based systems may struggle. However, the sparse, noisy nature of mmWave radar point clouds, along with multipath effects, presents difficulties for precise pose reconstruction. This paper introduces the Geometry-aware Feature Decoupling Network (GF-DecNet), an approach that employs a dynamic graph convolutional network (DGCN) framework to overcome these challenges. GF-DecNet combines Set Abstraction (SA) modules with Geometry-aware Feature Decoupling (GFD) blocks for effective multi-scale feature learning. This design captures both local geometric details and global context, which are crucial for accurate pose estimation. The GFD blocks are designed to disentangle local geometric features from higher-level semantic features, enhancing the network’s ability to process sparse and noisy data. A multi-scale feature aggregation module merges these learned features, providing a comprehensive representation of human pose that integrates low-level geometric details with high-level semantic information. Additionally, a learnable geometric constraint mechanism, grounded in biomechanical models, enhances robustness by ensuring consistency with human body proportions. Extensive experiments on four established public datasets (MiliPoint, MM-Fi, mRI, and mmBody) and our curated WavePose dataset demonstrate that GF-DecNet consistently achieves competitive performance, achieving significant improvements in Mean Per Joint Position Error (MPJPE). Ablation studies further validate the effectiveness of each key component of GF-DecNet, These findings establish GF-DecNet as a strong benchmark in the field of mmWave radar-based pose estimation.

Wi-Fi sensing technology can infer information about the environment, including user privacy. With the evolution of the smart home, more and more Wi-Fi devices are placed in a home. The ubiquitous Wi-Fi devices lead to concerns over privacy breaches caused by Wi-Fi sensing. In this paper, we discuss the insider threat model, where the smart device may be used to spy on the user through Wi-Fi sensing. To protect user privacy in the insider model, we propose WiShield, a bolt-on countermeasure against abusing smart devices for malicious Wi-Fi sensing in a per-device manner. WiShield works by attaching it to the smart devices. It utilizes low-cost radio frequency (RF) components to obfuscate the smart devices’ sensing by switching among multiple propagation paths with different gains and lengths. The user can recover a certain device’s sensing by disabling the corresponding WiShield through a policy configuration or command. We validate the effectiveness of WiShield against two state-of-the-art sensing strategies. The result shows that WiShield is effective in hiding most information contained in the Wi-Fi signals.

The privacy of fingerprints is a growing concern due to the risk of data breaches and subsequent attacks. A key issue is that the information of the same fingerprint may exist on multiple devices, and device-level protection mechanisms are fundamentally limited in their coverage. Consequently, information leakage from any device potentially affects all enrolled fingerprint recognition devices. In this paper, we introduce a novel fingerprint enrollment method called MaskPrint, which allows users to enroll in various fingerprint recognition systems using distinct information. This approach can largely mitigate the risk of data breaches, providing a user-transparent, device-agnostic fingerprint protection measure. Our method involves collecting the original fingerprint information, selecting a minimum feature set, synthesizing protective fingerprints, and fabricating physical ones for enrollment. Users can complete these procedures on their own. We validate the effectiveness and usability of MaskPrint on commercial fingerprint recognition systems.

This paper presents Centaur, an input system that enables tangible interaction on displays, e.g., untouchable computer monitors. Centaur’s tangibles are built from low-cost optical mouse sensors, or can alternatively be emulated by commercial optical mice already available. They are trackable when put on the display, rendering a real-time and high-precision tangible interface. Even for ordinary personal computers, enabling Centaur requires no new hardware and installation burden. Centaur’s cost-effectiveness and wide availability open up new opportunities for tangible user interface (TUI) users and practitioners. Centaur’s key innovation lies in its tracking method. It embeds high-frequency light signals into different portions of the display content as location beacons. When the tangibles are put on the screen, they are able to sense the light signals with their optical mouse sensors, and thus determine the locations accordingly. We develop four applications to showcase the potential usage of Centaur.

QUIC's connection migration provides an opportunity for network middleboxes to control paths of network traffic in a fine time granularity without worrying about breaking down the transport layer connections. This paper takes the first step to investigate its feasibility and effectiveness. We apply connection migration to a multi-carrier Wi-Fi relay to counter the latency caused by cellular network suspension. We identify a deadlock in such a situation, which eliminates the gain of passive migration. We analyze the root cause and propose a simple but effective approach to resolve the deadlock. Our findings may also motivate similar applications in other middlebox scenarios.

This paper proposes MousePath, a novel lightweight communication system between PC web pages and smartphones. MousePath works by putting the optical mouse on top of the smartphone's screen, then its transmission starts and is instantly finished without association and pairing friction. It encodes data into the movement of smartphone's display content and leverages the optical mouse of the computer to sense the movement for decoding the data. We prototype and evaluate the system with commercial computers and smartphones. A key benefit of MousePath is that it can be seamlessly integrated into web pages. Two representative web applications, i.e., sensor sharing and message sharing, have been developed to demonstrate MousePath's potential in enhancing PC web page functionalities.