We present a visual encoding method for semantic mapping in indoor environments, designed to minimize redundancy in image data and support federated learning across a fleet of service robots. Our pipeline combines 2D LiDAR-based segmentation with RGB image filtering based on geometric orientation, distance, visibility, and uniqueness. The result is a compact set of representative visual samples suitable for downstream semantic tasks such as object recognition or language grounding. We evaluate our method in a Gazebo simulation using a TurtleBot platform and compare it against a naive odometry-based sampling strategy. Our approach achieves up to 57.5% reduction in collected images while preserving scene coverage. Additionally, we demonstrate how multiple robots can collaboratively improve the visual map in a federated setup, reducing collection time and enabling model generalization across diverse environments. The proposed method offers an efficient and scalable solution for semantic mapping under bandwidth and computation constraints.
While artificial intelligence (AI) is improving the performance of O-RAN, it will also expose the network to adversarial machine learning (AML) attacks. For this reason, in this paper, we are the first to investigate AML in the context of deep reinforcement learning (DRL)-based O-RAN xApps. What separates AML in O-RAN from traditional settings is the need to design and analyze adversarial attacks based on RAN-specific Key Performance Measures (KPMs) such as transmitted bit rate, downlink buffer occupancy, transmitted packets, etc. As such, we propose the AdvO-RAN framework, which includes (i) a new adversarial perturbation generator using preference-based reinforcement learning (PbRL) to learn the perturbation that most violate the user service level agreements (SLA) and (ii) a robust training module for enhancing DRL agent resilience to the attacks in (i). We experimentally evaluate AdvO-RAN on the Colosseum network emulator. Experimental results show that AdvO-RAN can enhance xApp performance by reducing SLA violations from 44% to 27% on average and reducing by 46% the latency under the most challenging attack scenario for Ultra-Reliable Low-Latency Communications (URLLC) traffic. AdvO-RAN can improve up to 75% of throughput in the victim Enhanced Mobile Broadband (eMBB) slice users during a constant bit-rate traffic scenario.
Personal characteristics fundamentally shape virtual reality (VR) experiences, yet their integration into predictive models remains underexplored. This paper studies how to incorporate personal attributes (age, gender, prior VR experience) into Bayesian networks for cybersickness prediction via: (i) direct inclusion as root nodes, (ii) a two-stage model that learns a susceptibility score from personal attributes, and (iii) a stratified model. Using 26,040 samples from VR maze-navigation experiments, direct inclusion attains 82.53% accuracy (+14.02 percentage points over a 68.51% no-personal baseline). The two-stage approach reaches 77.32% while supporting cold-start prediction for unseen users, and stratified models achieve 73.62%. Using participant-level cross-validation to avoid subject leakage, we find that personalization consistently improves cybersickness prediction. These results argue that personal attributes should be treated as first-class signals in cybersickness models, with clear design trade-offs between maximal accuracy and deployability for unseen users, informing personalized VR systems and adaptive content delivery.
Cognitive attacks in mixed reality (MR), e.g., latency perturbations that induce frame-time jitter, can divert visual attention and degrade task performance. We study 2D gaze prediction under such disturbances and propose a time-aware sequence model that handles irregular sampling by supplying elapsed times triangle t between observations and conditions on sparse event/object context available at prediction time via learned token embeddings. Using time-based windows, we evaluate within-user and cross-user temporal generalization on MR recordings spanning multiple attack intensities. Results indicate accurate, time-robust gaze regression under latency perturbations, supporting adaptive MR interfaces in adversarial settings.
Internet browsing exposes private user information to websites and intermediate network providers, such as geographically linkable IP addresses and browsing history. Privacy proxies aim to address this internet privacy problem at scale. They are affordable for a large user base and do not compromise performance or usability-with a trend towards enabling them by default. Due to network encryption and closed-source code, their privacy promises cannot be verified using passive observations. We are the first to analyze the client-side implementations of three prominent privacy proxies available to over a billion users: Apple's iCloud Private Relay, Google's IP Protection, and Microsoft's Edge Secure Network VPN. We develop a deep understanding of these systems by reverse-engineering closed-source components and creating a client to bring Apple's, Google's, and Microsoft's solutions to Firefox on Linux. While privacy proxies offer privacy-enhancing features, we identify multiple issues deeply anchored in their architecture that weaken user privacy and security promises across all solutions. Our responsible disclosure and design recommendations aim to further strengthen this novel technology.
Industrial edge devices participating in federated visual anomaly detection systems play a vital role in maintaining operational efficiency. However, the complexity of high-dimensional visual data, such as images, poses significant computational challenges in resource-constrained environments. Efficient processing of such data is essential in settings characterized by heterogeneous processing power, memory capacity, and bandwidth availability. Deploying a single, uniform deep neural network architecture across this diverse landscape may overwhelm lightweight devices or underutilize more capable nodes, leading to imbalanced resource allocation and suboptimal system performance. Additionally, strict privacy constraints prevent the aggregation of centralized data, underscoring the need for decentralized intelligence systems. To address these challenges, we propose a novel hybrid framework combining deep neural networks (DNNs) tailored for federated learning (FL), using the Federated Proximal (FedProx) algorithm. The architecture combines ShuffleNetV2, a lightweight convolutional backbone designed for efficient feature extraction that distills raw visual input into compact, high-salience representations, with the Asymmetrical Consensus Mamba (ACMamba) module, an advanced state-space model that facilitates region-level consensus for robust anomaly detection under strict resource constraints at the edge. This design minimizes computational footprint, reduces communication overhead, and enhances scalability within FL environments. Experiments performed on a Raspberry Pi 3B+ edge device demonstrated improved communication efficiency reflected by reduced inference time, CPU load, and RAM usage, while maintaining robust performance across all 15 categories of the MVTec AD dataset. These results demonstrate the viability of the framework for self-sufficient and scalable anomaly detection on federated industrial edge platforms.
The dynamic nature of wireless environments presents significant challenges for machine learning (ML) models in real-world radio frequency applications, where impairments such as noise, fading, and frequency shifts disrupt performance. To address these challenges and build trust in ML models, we present Augmented Input Resilience Analysis (AURA), a test framework designed for IQ-based RF models to rigorously assess ML model performance by simulating RF impairments and identifying critical vulnerabilities. AURA systematically applies test-time augmentations to provide a detailed examination of model strengths and weaknesses. Key contributions include (1) Score-CAM adapted to 1D IQ in time and a frequency-selective variant to localize spectral features, and (2) embedding similarity evaluation to quantify distribution shifts caused by impairments. By integrating these methods, AURA enhances interpretability, promoting trust in ML decision-making. We demonstrate AURA's utility in exposing critical vulnerabilities in well-cited models, such as over-reliance on power-based features, including instances where random noise is misclassified as a legitimate signal with 99.7% accuracy. AURA also evaluates remediation strategies, such as noise classes, which reduce misclassifications to less than 1% in the noise augmentation case. This framework aims to advance the design of trustworthy and resilient AI-driven systems for future RF ML technologies.
Decentralized federated learning (DFL) can greatly reduce communication costs due to its decentralized communication structure compared to traditional centralized federated learning (FL). Existing works on FL with partial client participation often considered idealized scenarios (such as all clients participate in a round with the same probability), or required using clients' past gradient/model information which can be too costly to implement, or focused on centralized FL. In this paper, we study lightweight decentralized federated learning that does not use any client's past gradient/model information. We first present a novel sample-path-based cyclic convergence analysis for lightweight DFL with arbitrary client participation for the non-convex objectives case. The cyclic convergence analysis bounds clients' local model drifts due to partial participation over multiple rounds within a cycle and the cyclic consensus error via a per-cycle descent approach, while capturing the effect of client participation through a single unified term. By analyzing this term, we propose Cyclic Decentralized Federated Learning (CDFL), which enables general cyclic client participation by requiring only that each client performs the same total number of local updates per cycle. Our results show that CDFL achieves a convergence rate that matches existing benchmarks. We further propose a cyclic control framework that is both training-round and energy efficient to adaptively select participating clients and determine their number of local updates. Numerical experiments using real-world datasets verify our theoretical results and demonstrate the effectiveness of CDFL and the adaptive cyclic control framework.
Modern cars monitor their surroundings and record video to deter intruders, but current surveillance systems operate independently without communication. Connected vehicles can share detected features of suspicious individuals, improving tracking and alerting approaching drivers before they park in vulnerable spots. This paper explores this use case, where connected vehicles exchange features over the network upon detecting suspicious individuals. Real-world data analysis shows that connected vehicle surveillance improves detection accuracy by 38%.
We present a practical in-network compression system for multimedia delivery under dynamic wireless conditions at the network edge. Our system leverages Dynamic Multi-Level AutoEncoder (DMAE), allowing edge nodes to apply additional compression in response to local congestion. This enables runtime, feedback-free adaptive compression, ensuring timely delivery of multimedia traffic. We demonstrate the system's feasibility on a programmable three-node testbed.
Volumetric video (VV) offers photorealistic 3D capture for immersive educational VR, often created by instructors through live-streamed lessons or prerecorded demonstrations. While enhancing engagement and presence, such instructor-produced content can unintentionally expose sensitive objects, personal information, or biometric identifiers, and may intensify feelings of surveillance. This poster examines these privacy risks in using VV for educational VR and presents a research agenda focused on integrating diminished reality (DR) techniques and real-time 3D scene understanding into VV pipelines to dynamically sanitize environments while balancing realism and privacy.
This work aims to enable efficient digital rights management (DRM) for volumetric video by introducing attribute-based selective coordinate encryption for point clouds. By encrypting only a subset of coordinates, our approach reduces computational overhead and latency while maintaining necessary security. Selective encryption ensures that point cloud frames-and, by extension, entire volumetric videos-are sufficiently obfuscated so that, while the content remains viewable, it appears highly distorted and visually unpleasant, preventing meaningful unauthorized viewing. We propose a flexible framework that allows varying the amount and type of coordinate encryption (e.g. x,y,z, or combinations), and we assess visual degradation using established point cloud quality metrics. Our results show that encrypting only.. coordinates cuts encryption and decryption times by 37% and 46%, respectively, compared to full-frame encryption, while.. and.. encryption achieves 20% and 36% reductions, both still significantly degrading visual quality. Leveraging Attribute-Based Encryption (ABE) further enables content to be securely cached and efficiently distributed in its protected form, eliminating the need for re-encryption, thereby reducing computational load and latency. While our current evaluation is limited to individual point cloud frames, future work will extend to entire volumetric video streams, including analysis of caching gains during streaming with ABE.
Bluetooth Low Energy (BLE) direction finding, a feature introduced in BLE version 5.1, enables precise localization through Angle of Arrival (AoA) estimation. However, this advancement introduces new risk to BLE direction finding based localization system. Specifically, the AoA estimation based on phase sampling of constant-tone-extension (CTE) is susceptible to the signal injection attack. This paper presents SADIF, a feasible spoofing attack mechanism to mislead the locators into mistaking the positioning result as a continuous path. By eavesdropping on BLE packets and injecting attack signals containing pre-designed disturbing phase shift, SADIF subtly alters the AoA estimation without detection, thus interfere the localization results. Moreover, SADIF address the challenges posed by hardware imperfections by proposing an injection timing optimization to improve attack robustness. Extensive experiments demonstrates the effectiveness of SADIF in successfully attacking multiple BLE targets in real-time scenarios. In conclusion, our findings reveal critical security risks in BLE direction finding feature and provide insights into strengthening its defenses.
Connected vehicles use Vehicle-to-Everything (V2X) communication to improve situational awareness. However, transmitting raw or irrelevant data may congest the communication channel, and limit the scalability and effectiveness of safety applications. This paper introduces the Context-Aware V2X concept, which assesses message relevance and adds contextual meaning before transmission, ensuring that only essential information reaches relevant receivers. We demonstrate the e!ectiveness of the Context-Aware V2X concept through simulations of the tailgating scenario, where results show that Context-Aware V2X can improve driver awareness and reduce risks associated with tailgating.
Ensuring the safety of neural network (NN) models in mixed reality (MR) systems is challenging due to adversarial manipulation of system parameters. We present PolySafe, which extends DeepPoly and Prover to validate safety of LSTM-based MR models. PolySafe unrolls temporal dependencies, introduces multi-plane abstractions for tighter bounds, and establishes probabilistic safety guarantees. It further includes an adaptive search that identifies minimal sets of critical parameters required to be constrained for defense. Evaluation on an MR engagement prediction model shows that PolySafe provides rigorous and actionable safety assurances for deployment.
The challenge of classification at the network edge is that due to limited computational resources, the edge must transmit the data to a server for processing. However, the communication constraints at the edge necessitate that these devices compress data before transmission. The question this paper aims to answer is how to efficiently compress and transmit this information in order to achieve timely and accurate edge classification. To that end, we develop scheduling algorithms that optimize age of information (AoI) and classification accuracy. Our analysis reveals that in scenarios with multiple available compression levels, an algorithm that selects at most two compression levels can achieve good theoretical performance guarantees. Numerical results indicate that double-level compression algorithms yield near-optimal performance, suggesting that for many classification tasks, numerous compression levels are unnecessary-only two are sufficient, significantly reducing the storage demands on devices and simplifying the overall system design.
We consider the partitioning of a deep neural network (DNN) inference job and offloading part of it from a resource-constrained device to a resource-rich server. The inference job is required to finish within a delay constraint, but it is allowed to perform early exit at some intermediate layer of the DNN, at the cost of lower accuracy. Since in practice both the processing delay and the communication delay of offloading usually are unknown ahead of time, this is naturally modelled as an online delay-constrained accuracy maximization problem. We propose Accuracy-Optimal Delay Constrained Online Partitioning (AODPart), a lightweight online algorithm that uses an adaptive thresholding strategy to solve the offloading problem. We derive the competitive ratio for AODPart and show that it is optimal in the sense that no other online algorithm can achieve a lower deterministic competitive ratio. Furthermore, we show that AODPart is robust and provides worst-case performance guarantee even with parameter estimation error. Through experimenting with common vision and language learning models, we demonstrate that AODPart substantially outperforms state-of-the-art alternatives and returns near optimal accuracy in practice.
The fog node placement problem has become an important topic in the field of the internet of things. In this paper, we explore a specific scenario where all fog nodes (FNs) are connected to a base station through a minimum spanning tree and we assume they all have the same radius of coverage. We also assume each FN has the capacity to process all the data coming from the end users (EUs) within the radius of coverage. We evaluate the effectiveness of difference placement algorithms using information elasticity concepts.
With the rapid adoption of extended reality (XR) technologies in both consumer and enterprise domains, continuous and unobtrusive user authentication has become increasingly important. Existing authentication methods are often intrusive, static, or insufficiently secure for immersive environments. In this work, we propose a novel passive authentication framework that leverages users' realtime pupil light reflex (PLR) in response to visual stimuli rendered in XR. By treating screen brightness as a natural, time-varying challenge and modeling the user's pupil response as the biometric signal, our system learns to extract identity-specific features that are invariant to environmental content. We implement our prototype on two commercial XR headsets and evaluate it through a user study involving eight participants across diverse XR applications. Our system achieves an equal error rate (EER) of 0.093 with a 2-minute prediction window. These results demonstrate the feasibility of pupillary dynamics as a behavioral biometric for secure, continuous authentication in immersive environments. This study lays the foundation for future work on scalable, multimodal, and adaptive biometric authentication in XR.
Accurate localization in dense urban areas remains a significant challenge due to the limitations of Global Navigation Satellite Systems (GNSS) in environments with obstacles and reflections, such as urban canyons. While the most recent 3GPP standards offer sophisticated network-centric positioning techniques, their widespread deployment will take time and is hindered by high infrastructure costs and complexity. In this work, we present mm-NOLOC, a UE-centric localization system, designed as a practical fallback when GNSS fails to deliver high accuracy, that leverages the growing deployment of 5G mmWave infrastructure in dense urban areas. Unlike traditional approaches, mm-NOLOC operates independently of 3GPP location support and utilizes only standardized control-plane information collected solely on the UE side - Synchronization Signal Block (SSB) Indices that are mapped to 5G mmWave beam directions - to obtain robust position estimations. To address the uncertainty introduced by urban multipath, mm-NOLOC models the SSB-to-angle relationship as a discrete and multimodal distribution, based on empirical measurements in operational 5G mmWave networks, and uses a particle filter to refine position estimates by integrating probabilistic observations with UE-side motion dynamics. We validate mm-NOLOC through experiments over commercial 5G mmWave deployments, as well as trace-based simulations. Our results show that mm-NOLOC achieves a median localization error below 3 m and a 95th percentile error below 10 m, offering a practical fallback localization solution in urban canyon scenarios for 5G networks without network location support.
