The increasing prevalence of IoT technology in smart homes has significantly enhanced convenience but also introduced new security and safety challenges. Traditional security solutions, reliant on sequences of IoT-generated event data (e.g., notifications of device status changes and sensor readings), are vulnerable to cyberattacks, such as message forgery and interception and delaying attacks, and fail to monitor non-smart devices. Moreover, fragmented smart home ecosystems require vendor cooperation or system modifications for comprehensive monitoring, limiting the practicality of the existing approaches. To address these issues, we propose IoTBystander, a non-intrusive dual-channel smart home security monitoring framework that utilizes two ubiquitous platform-agnostic signals, i.e., audio and network, to monitor user and device activities. We introduce a novel dual-channel aggregation mechanism that integrates insights from both channels and cross-verifies the integrity of monitoring results. This approach expands the monitoring scope to include non-smart devices and provides richer context for anomaly detection, failure diagnosis, and configuration debugging. Empirical evaluations on a real-world testbed with nine smart and eleven non-smart devices demonstrate the high accuracy of IoTBystander in event recognition: 92.86% for recognizing events of smart devices, 95.09% for non-smart devices, and 94.27% for all devices. A case study on five anomaly scenarios further shows significant improvements in anomaly detection performance by combining the strengths of both channels.

Visual place recognition is crucial in robot localization and autonomous driving. Most existing methods focus on inter-domain appearance changes, but strug

The Eulerian–Eulerian approach, coupled with the Reynolds-averaged Navier–Stokes (RANS) technique, is a practical combination for simulating industrial slurry flows. In such simulations, the turbulent dispersion force plays a critical role in driving particles from regions of high concentration to low concentration, thereby influencing particle distribution and overall hydrodynamic behaviors. The performance of the proposed numerous turbulent dispersion force models has not been evaluated systematically in simulating sand–water slurry flows in horizontal pipes. This study investigates turbulent slurry flows in horizontal pipes under various flow conditions using the Eulerian–Eulerian approach combined with RANS. Simulations without turbulent dispersion force and with three widely used turbulent dispersion force models—Lopez de Bertodano, Simonin, and Burns—are analyzed based on predictions of streamwise velocity, secondary flow, solid concentration distribution, and liquid turbulent kinetic energy. A comparison of the simulation results with experimental data from the literature on streamwise velocity and solid concentration distribution highlights the critical importance of incorporating the turbulent dispersion force for accurate predictions. The Lopez de Bertodano model significantly underestimates the turbulent dispersion effect and, therefore, is not recommended for such flows. The Simonin and Burns models exhibit better performance; however, further refinement is required to achieve higher predictive accuracy. A computational efficiency analysis is conducted. Considering both accuracy and efficiency, the Burns model is identified as the optimal choice for the flow cases examined in this study. These findings may serve as a useful reference for slurry simulations in horizontal pipelines and assist researchers in making more informed model selections.

Smart home IoT systems are increasingly vulnerable to event-targeted attacks that manipulate device-reported event streams through masking, spoofing, or disordering, thereby undermining conventional event-based monitoring and automation. We present IoTAudVeri, an audio-assisted verification framework that cross-validates platform-observed device event streams with independently inferred audio evidence captured by embedded microphones in commodity devices (e.g., doorbells, cameras, and speakers). We evaluate IoTAudVeri on a public dataset (ESC-5), a home-collected dataset (HomeSet-8), and in a 40-day continuous real-world smart home deployment. IoTAudVeri achieves up to 97.5% audio event recognition accuracy with only a few minutes of training data per event, and detects three categories of event-targeted attacks with high recall (up to 98.5%) and very low false-positive rates (0–1.2%), highlighting the benefit of semantic audio-assisted verification over representative cyber-only and non-semantic acoustic baselines under event-targeted attacks.

Recent advances in embodied AI have increasingly utilized foundation models to address robot manipulation challenges. Among these, Vision-Language-Action (VLA) models significantly enhance manipulation tasks performance due to their integrated vision, language, and action capabilities. However, prevailing VLA architectures, including those that employ dual systems to separately process vision-language and action modalities, still rely on latent vector representations which suffer from knowledge degradation and thus constrain the generalizability and reusability of off-the-shelf Vision-Language Models (VLMs). To address this, we explore VLM-generated graphical intent annotations as explicit intermediate representations between high-level VLM planning and low-level VLA execution. Through systematic experiments, we demonstrate that these annotations significantly improve success rates, sample efficiency, and cross-scene generalization. Notably, our key findings reveal that the downstream VLA policy tends to learn from the graphical annotations rather than from the actual scene background without any auxiliary supervision. This work establishes graphical annotations as a interpretable pathway for effective knowledge transfer between high-level planners and low-level embodied policies.