Multilingual neural machine translation (MNMT) models degrade in performance as input context length increases, causing positional encoding schemes to misinterpret token distances.Existing absolute and relative positional encodings rely on fixed token indices and implicitly assume uniform semantic density, which breaks down for long-context inputs.We introduce DCARPE, a tokenization-aware adaptive positional encoding that conditions relative positional bias on inputlevel sequence length and fragmentation statistics, allowing the model to reinterpret positional distance when tokenization-induced inflation arises rather than semantic factors.Evaluations on JW300 and out-of-distribution FLORES-200 demonstrate consistent improvements in long-context robustness, achieving gains of up to +10.81 ChrF++ and +8.00 BLEU over baselines.

Emerging resource-constrained cellular Internet of Things (IoT) applications such as drone swarms, autonomous vehicles, and remote surgery via mixed reality demand millisecond-level time synchronization. Narrow-Band IoT (NB-IoT), the leading low-power wide-area technology, struggles to meet these requirements. The root cause lies in the non-deterministic delays inherent in the 5G protocol design. Uplink reliability and scheduling mechanisms introduce asymmetric latencies that disrupt conventional time synchronization algorithms such as the Network Time Protocol (NTP). Time-critical packets are further affected by deep-sleep wake-up latency, base station scheduling delays, uplink/downlink asymmetry, and unpredictable drift from inexpensive oscillators. Together, these factors can accumulate into timing errors on the order of hundreds of milliseconds. In this paper, we first quantify timing errors across five dimensions on a commercial NB-IoT network. We then present SynchroNB, an on-device framework that combines lightweight machine learning with a cross-layer control loop. SynchroNB forecasts 5G network volatility and crystal drift to adaptively wake the cellular modem, reserves uplink resources just in time, switches into resilience mode when the wireless link degrades, and prioritizes time synchronization packets in the MAC-layer queue. We deploy SynchroNB on commercial NB-IoT hardware and evaluate it over a live 5G network. Our experiments show that SynchroNB achieves single-millisecond-level synchronization accuracy under NB-IoT uplink/downlink asymmetry and, diverse wireless conditions, while requiring only \(36\%\) of the radio-on time and \(25\%\) of the bandwidth of the NTP baseline, transforming NB-IoT time synchronization from a reactive protocol into an intelligent, self-tuning control loop.

Mixed Reality (MR) enables rich, embodied collaboration; however, it is uncertain whether sensor- and system-logged behavioral signals capture how users experience that collaboration. This disconnect stems from a fundamental gap. Behavioral signals are observable and continuous, while collaboration is interpreted subjectively and shaped by internal states like presence, cognitive availability, and social awareness. Our core insight is that sensor signals serve as observable manifestations of subjective experiences in MR collaboration, and they can be captured through sensor data such as shared gaze, speech, spatial movement, and other system-logged performance metrics. We propose the Sensor-to-Subjective (S2S ) Mapping Framework, a conceptual model that links observable interaction patterns to users’ subjective perceptions of collaboration and internal cognitive states through sensor-based indicators and task performance metrics. To evaluate this model, we conducted an exploratory study with 48 participants across 12 MR groups engaged in a collaborative image-sorting task. Our findings show a correlation between sensed behavior and perceived collaboration, particularly through shared attention and proximity.

Accurate and secure timekeeping is a fundamental requirement for the dependable operation of cyber-physical systems and distributed applications deployed in virtualized environments. However, existing timing mechanisms are susceptible to compromise by privileged adversaries, such as the hypervisor, thereby undermining the integrity of local timing services. This paper introduces Seti,a system that employs an out-of-band inspection strategy to establish trust in time sources. Rather than modifying timing access mechanisms to preclude attacks, Setimonitors the integrity of existing timing paths, thus reducing system complexity and cost. It utilizes Remote Direct Memory Access (RDMA) in conjunction with System Management Mode (SMM)-based introspection to enable guest virtual machines (VMs) to access trusted time sources without interruption, while simultaneously detecting and mitigating malicious interference with timer interrupts. Setiis designed to impose minimal overhead on both host and guest systems, preserving high time accuracy and enabling prompt detection of adversarial activities. Experimental evaluation demonstrates that Setieffectively pro-tects critical timing functions achieving millisecond-level attack detection latency with less than 1% CPU overhead.

This paper investigates the security vulnerabilities of prompt-learning-based FL systems in a healthcare setting. Specifically, we use a backdoor attack that leverages learnable prompt vectors in vision-language medical foundation models to execute stealthy adversarial manipulations. We evaluate our attack across diverse healthcare datasets and FL configurations, showing that while FL is useful as a privacy-preserving mechanism, it is susceptible to targeted backdoor attacks that pose a threat to medical applications.

Timekeeping is a fundamental component of modern computing; however, the security of system time remains an overlooked attack surface, leaving critical systems vulnerable to manipulation. This paper examines time manipulation attacks that exploit kernel vulnerabilities to distort an application's perception of time. We categorize these attacks into constant, incremental, and randomized delay strategies and analyze their impact on system performance. Through experimental evaluation, we demonstrate how adversaries can manipulate system time via dynamic library injection and syscall modification, disrupting time-sensitive applications. While Trusted Execution Environments (TEEs) offer partial isolation, they fail to address time security concerns fully. Our results highlight the challenges of securing system time and underscore the need for robust mitigation strategies.

In-flight Wi-Fi provides high-speed Internet connectivity to travelers at 30,000 feet at premium fees. In this paper, we present the first systematic study of the architecture and security policies of in-flight Wi-Fi paywall systems using network tomography analysis. We discover that attackers can exploit the inherent architectural shortcomings of airborne networks to create covert channels and conceal data packets within certain ''always-allowed'' traffic for free Internet access. Moreover, broken device authentication policies in these systems allow unlimited complimentary Internet connectivity. Finally, insecure ARP policies allow attackers to steal paid users' bandwidth to access the free Internet even faster. We validate these issues in practice over two major in-flight Wi-Fi providers using common protocols, e.g., UDP, DNS, etc. We also find that the root causes of these issues stem from different design choices in the architectures of these systems and propose countermeasures to address these flaws and prevent similar attacks.

Distractions in mixed reality (MR) environments can significantly influence user experience, affecting key factors such as presence, reaction time, cognitive load, and Break in Presence (BIP). Presence measures immersion, reaction time captures user responsiveness, cognitive load reflects mental effort, and BIP represents moments when attention shifts from the virtual to the real world, breaking immersion. While prior work has established that distractions impact these factors individually, the relationship between these constructs remains underexplored, particularly in MR environments where users engage with both real and virtual stimuli. To address this gap, we have presented a theoretical model to understand how congruent and incongruent distractions affect all these constructs. We conducted a within-subject study (N = 54) where participants performed image-sorting tasks under different distraction conditions. Our findings show that incongruent distractions significantly increase cognitive load, slow reaction times, and elevate BIP frequency, with presence mediating these effects.

Autonomous robots, autonomous vehicles, and humans wearing mixed-reality headsets require accurate and reliable tracking services for safety-critical applications in dynamically changing real-world environments. However, the existing tracking approaches, such as Simultaneous Localization and Mapping (SLAM), do not adapt well to environmental changes and boundary conditions despite extensive manual tuning. On the other hand, while deep learning-based approaches can better adapt to environmental changes, they typically demand substantial data for training and often lack flexibility in adapting to new domains. To solve this problem, we propose leveraging the neurosymbolic program synthesis approach to construct adaptable SLAM pipelines that integrate the domain knowledge from traditional SLAM approaches while leveraging data to learn complex relationships. While the approach can synthesize end-to-end SLAM pipelines, we focus on synthesizing the feature extraction module. We first devise a domain-specific language (DSL) that can encapsulate domain knowledge on the essential attributes for feature extraction and the real-world performance of various feature extractors. Our neurosymbolic architecture then undertakes adaptive feature extraction, optimizing parameters via learning while employing symbolic reasoning to select the most suitable feature extractor. Our evaluations demonstrate that our approach, neurosymbolic Feature EXtraction (nFEX), yields higher-quality features. It also reduces the pose error observed for the state-of-the-art baseline feature extractors ORB and SIFT by up to 90% and up to 66%, respectively, thereby enhancing the system’s efficiency and adaptability to novel environments.

Data heterogeneity among Federated Learning (FL) users poses a significant challenge, resulting in reduced global model performance. The community has designed various techniques to tackle this issue, among which Knowledge Distillation (KD)-based techniques are common. While these techniques effectively improve performance under high heterogeneity, they inadvertently cause higher accuracy degradation under model poisoning attacks (known as \emph{attack amplification}). This paper presents a case study to reveal this critical vulnerability in KD-based FL systems. We show why KD causes this issue through empirical evidence and use it as motivation to design a hybrid distillation technique. We introduce a novel algorithm, Hybrid Knowledge Distillation for Robust and Accurate FL (HYDRA-FL), which reduces the impact of attacks in attack scenarios by offloading some of the KD loss to a shallow layer via an auxiliary classifier. We model HYDRA-FL as a generic framework and adapt it to two KD-based FL algorithms, FedNTD and MOON. Using these two as case studies, we demonstrate that our technique outperforms baselines in attack settings while maintaining comparable performance in benign settings.

The importance of timing services in edge systems makes them a lucrative target for privileged adversaries. Mali-cious agents with Operating System (OS) privileges can stealthily manipulate timing services and provide altered timestamps to user applications. In this paper, we first demonstrate the ad-verse impact of time attacks on the accuracy of sensor fusion algorithms at the edge. Then, we introduce Timeg Uard, our proposed architecture that protects against time attacks and provides trusted time to user applications. Timeguard's design leverages the secure interrupt and memory primitives of trusted execution environments (TEEs) to bypass untrusted privileged software and acquire time securely. Yet, these secure primitives come at a high computational cost. Timeg Uardalso introduces a probabilistic security framework - bounded by a time error - to limit the cost of our timing service. We prototype our design on ARM TrustZone - the dominant secure architecture in edge systems, and evaluate the trade-off in security, accuracy, and system overhead. TIMEGUARD's secure performance ranges from a microsecond to tens of milliseconds at 3.9% and 1.2% CPU overhead respectively, catering to a variety of application requirements.

Synchronizing clocks is a resource-intensive and a resource-rigid task; this makes it challenging to align time across resource-constrained and heterogeneous IoT devices. Just as low-power IoT devices harvest energy from the environment, we propose to scavenge timing information from environmental events to conserve device power and bandwidth. We convert ambient events in an environment - sensed by various sensing modalities - into synchronization signals. Our approach, HAEST, leverages prevalent sensors on commodity platforms such as accelerometer, microphone, and optical sensors to timestamp commonly observed events. We present a light-weight and robust approach to detect events across different types of sensors and devices. This creates ample opportunities to align time and simultaneously avoid the clocks to drift apart. Through evaluation on a hardware testbed in a smart home and a wireless body area network, we show that HAEST achieves clock accuracy as low as sub-milliseconds with no cost for IoT devices. Importantly, our use of off-the-shelf IoT platforms in our evaluations, establishes the universality and applicability of our approach to a variety of smart spaces with heterogeneous devices.

Measuring presence is critical to improving user involvement and performance in Mixed Reality (MR). Presence, a crucial aspect of MR, is traditionally gauged using subjective questionnaires, leading to a lack of time-varying responses and susceptibility to user bias. Inspired by the existing literature on the relationship between presence and human performance, the proposed methodology systematically measures a user's reaction time to a visual stimulus as they interact within a manipulated MR environment. We explore the user reaction time as a quantity that can be easily measured using the systemic tools available in modern MR devices. We conducted an exploratory study (N = 40) with two experiments designed to alter the users' sense of presence by manipulating place illusion and plausibility illusion. We found a significant correlation between presence scores and reaction times with a correlation coefficient -0.65, suggesting that users with a higher sense of presence responded more swiftly to stimuli. We develop a model that estimates a user's presence level using the reaction time values with high accuracy of up to 80%. While our study suggests that reaction time can be used as a measure of presence, further investigation is needed to improve the accuracy of the model.

A prerequisite to improving the presence of a user in mixed reality (MR) is the ability to measure and quantify presence. Traditionally, subjective questionnaires have been used to assess the level of presence. However, recent studies have shown that presence is correlated with objective and systemic human performance measures such as reaction time. These studies analyze the correlation between presence and reaction time when technical factors such as object realism and plausibility of the object's behavior change. However, additional psychological and physiological human factors can also impact presence. It is unclear if presence can be mapped to and correlated with reaction time when human factors such as conditioning are involved. To answer this question, we conducted an exploratory study ($N=60$) where the relationship between presence and reaction time was assessed under three different conditioning scenarios: control, positive, and negative. We demonstrated that human factors impact presence. We found that presence scores and reaction times are significantly correlated (correlation coefficient of -0.64), suggesting that the impact of human factors on reaction time correlates with its effect on presence. In demonstrating that, our study takes another important step toward using objective and systemic measures like reaction time as a presence measure.

Driver distraction caused by over-reliance on automotive technology is one of the leading causes of accidents in semi-autonomous vehicles. Existing driver's attention-gauging approaches are intrusive and as such emphasize constant driver engagement. In case of an urgent traffic event, they fail to measure the event's criticality and subsequently generate timely alerts. In this paper, we re-position the driver's attention-gauging approach as a way to improve the driver's situational awareness during critical situations. We exploit how a vehicle captures its surroundings information to convert an automotive decision into defining the criticality and timeliness of an alert. For this, we identify the relationship between the traffic event, the type of automotive sensing technologies, and its processing resources to capture that event to design the driver's attention gauge. We evaluate the timeliness of alerts for different traffic scenarios over a prototype built using NVIDIA Jetson Xavier AGX and Carla. Our results show that we can improve the timeliness of an alert by up to 75x as compared to existing state-of-the-art approaches, while also providing feedback on its criticality.

Universal time is critical for coordinating function-alities among co-located as well as geographically distributed IoT devices. Current time alignment approaches for IoT devices rely on radio-based communication that puts an extra burden on the already resource-constrained devices. Other timing approaches depend either on customized hardware frontends or on fixed networking capabilities. Intermittent network connectivity fur-ther deteriorates timing performance for these devices. These constraints motivate us to create a new design that actively embeds time information into the surroundings of IoT devices that can harvest timing signals with off-the-shelf sensing capabil-ities. Our design decouples clock performance from the network uncertainties, introduce resource efficiency and extensibility, and requires no modification in existing devices. A unique property of our design is that it leverages the ubiquitous Electric Network Frequency (ENF) fluctuations as global time reference for the sensing devices, and takes on a variety of challenges at the in-tersection of sensing and signal processing to provide a universal sense of time without custom hardware frontends and network dependability. We evaluate the extensibility of our design in remote setups and show its robustness in real world settings.

There is a lack of datasets for visual-inertial odometry applications in Extended Reality (XR). To the best of our knowledge, there is no dataset available that is captured from an XR headset with a human as a carrier. To bridge this gap, we present a novel pose estimation dataset --- called HoloSet --- collected using Microsoft Hololens 2, which is a state-of-the-art head mounted device for XR. Potential applications for HoloSet include visual-inertial odometry, simultaneous localization and mapping (SLAM), and additional applications in XR that leverage visual-inertial data.

Split Learning (SL) and Federated Learning (FL) are two prominent distributed collaborative learning techniques that maintain data privacy by allowing clients to never share their private data with other clients and servers, and find extensive IoT applications in smart healthcare, smart cities and smart industry. Prior work has extensively explored the security vulnerabilities of FL in the form of poisoning attacks. To mitigate the effect of these attacks, several defenses have also been proposed. Recently, a hybrid of both learning techniques has emerged (commonly known as SplitFed) that capitalizes on their advantages (fast training) and eliminates their intrinsic disadvantages (centralized model updates).

Critical machine learning applications (medical image guidance, task prediction, anomaly detection) require large amounts of data that could not be sufficiently supplied from a single entity, so multiple edge devices collaboratively train their collected data. But this raises privacy and overhead concerns. Federated learning (FL) can be a promising solution to enable these applications while preserving data privacy and mitigating communication overhead. However, an FL model originating from edge deployments with heterogeneous resources may be biased towards a set of devices. We observe that existing bias mitigation techniques in FL focus mainly on the bias that originates from label heterogeneity (due to the skewed distribution of data). We argue that sample feature heterogeneity due to different feature representations at devices is a major contributor to bias in FL. In this paper, we present an analysis of the bias that arises from sampling feature heterogeneity, and analyze the potential of existing performance enhancing techniques (normalization) to overcome bias. Our results demonstrate that normalization techniques do not eliminate bias and motivate the need for dedicated bias mitigation techniques in FL.

Network Function Virtualization (NFV) migrates the carrier-grade LTE Evolved Packet Core (EPC) that runs on commodity boxes to the public cloud. In the new virtualized environment, LTE EPC must offer high availability to its mobile users upon failures. Achieving high service availability is challenging because failover procedure must keep the latency-sensitive control-plane procedures intact during failures. Through our empirical study, we show that existing recovery mechanisms on the cloud and standardized LTE solutions are coarse-grained, thus unable to quickly recover from failures. They incur LTE service outage, lost network connectivity, and slow recovery. To address these issues, we describe a new design for fault-tolerant LTE EPC. It provides quick failure detection and timely recovery from failed operations. To reduce failure detection time, it leverages frequent retransmission of LTE control-plane signaling within EPC as an indication of failure. To recover from failure, it adopts a checkpointing based rollback recovery approach in the LTE context and addresses the shortcomings known in the classic checkpointing approach. Our design is LTE standard-compliant and works as a plug-and-play without modifying existing LTE implementations. Our results show that this approach can recover from the failure in 2.6 seconds and only incurs tens of milliseconds of overhead.

This paper studies the feasibility of key reinstallation attacks in the 4G LTE network.It is well known that LTE uses session keys for confidentiality and integrity protection of its control-plane signaling packets.However, if the keys are not updated and counters are reset, key reinstallation attacks may arise.In this paper, we show that several design choices in the current LTE security setup are vulnerable to key reinstallation attacks.Specifically, on the control plane, the LTE security association setup procedures, which establish security between the device and the network, are disconnected.The keys are installed through one procedure, whereas their associated parameters (such as uplink and downlink counters) are reset through another different procedure.The adversary can thus exploit the disjoint security setup procedures, and launch the key stream reuse attacks.He consequently breaks message encryption, when he tricks the victim to use the same pair of keys and counter value to encrypt multiple messages.This control-plane attack hijacks the location update procedure, thus rendering the device to be unreachable from the Internet.Moreover, it may also deregister the victim from the LTE network.We have confirmed our findings with two major US operators, and found that such attacks can be launched with software-defined radio devices that cost about $299.We further propose remedies to defend against such threats.

The explosion of mobile applications and phenomenal adoption of mobile connectivity by end users make all-IP based 4G LTE as an ideal choice for providing Internet access on the go. LTE core network which handles device control-plane and data-plane traffic becomes susceptible to network resource constraints. To ease these constraints, Network Function Virtualization (NFV) provides high scalability and flexibility by enabling dynamic allocation of LTE core network resources. NFV achieves this by decomposing LTE Network Functions (NF) into multiple instances. However, LTE core network architecture which is designed considering fewer NF boxes does not fit well where decomposed NF instances add delays in network event execution. Certain control-plane events being time critical hurt data-plane traffic requirements defined by LTE standard. This paper proposes Fat-proxy which acts as a stand-alone execution engine of these critical network events. Through space uncoupling, we execute several signalling messages in parallel while skipping unnecessary messages to reduce event execution time and signalling overhead while ensuring highly available service access. We build our system prototype of open source LTE core network over virtualized platform. Our results show that we can reduce event execution time and signalling overhead upto 50% and 40%, respectively.

This paper exhaustively explores the threat landscape of coordinated spatiotemporal attacks in mixed reality systems. Novel devicelevel and cross-device time translation and spatial shift attacks are launched, and their impact on deep learning based sensor fusion is evaluated. A major focus of this work is to establish stealthiness in the presence of sophisticated security mechanisms with an added constraint that mixed reality systems allow minimal time durations for covert operation. The efficacy of proposed attacks is evaluated through a preliminary study on inertial and visual data streams.

We propose a new clock synchronization architecture for systems under time transfer attacks. Facilitated by a feedforward control with feedback trim --based clock adjustment, coupled with packet filtering and frequency shaping techniques, our proposed architecture bounds the clock errors in the presence of a powerful network attacker capable of attacking packets between a master and a client. A key advantage is consistent measurements, timely coordination, and synchronized actuation in distributed systems. In contrast, current time synchronization architectures behave poorly under attacks due to assumptions that the network is benign and delays are symmetric. The usage of feedback controllers aggravates poor performance. We provide an architecture that is indifferent to delays and eases the integration to traditional protocols. We implement a delay attack--resistant precision time protocol and validate the results on a hardware-supported testbed.

An accurate sense of elapsed time is essential for the safe and correct operation of hardware, software, and networked systems. Unfortunately, an adversary can manipulate the system's time and violate causality, consistency, and scheduling properties of underlying applications. Although cryptographic techniques are used to secure data, they cannot ensure time security as securing a time source is much more challenging, given that the result of inquiring time must be delivered in a timely fashion. In this paper, we first describe general attack vectors that can compromise a system's sense of time. To counter these attacks, we propose a secure time architecture, TIMESEAL that leverages a Trusted Execution Environment (TEE) to secure time-based primitives. While CPU security features of TEEs secure code and data in protected memory, we show that time sources available in TEE are still prone to OS attacks. TIMESEAL puts forward a high-resolution time source that protects against the OS delay and scheduling attacks. Our TIMESEAL prototype is based on Intel SGX and provides sub-millisecond (msec) resolution as compared to 1-second resolution of SGX trusted time. It also securely bounds the relative time accuracy to msec under OS attacks. In essence, TIMESEAL provides the capability of trusted timestamping and trusted scheduling to critical applications in the presence of a strong adversary. It delivers all temporal use cases pertinent to secure sensing, computing, and actuating in networked systems.

Having a shared and accurate sense of time is critical to distributed Cyber-Physical Systems (CPS) and the Internet of Things (IoT). Thanks to decades of research in clock technologies and synchronization protocols, it is now possible to measure and synchronize time across distributed systems with unprecedented accuracy. However, applications have not benefited to the same extent due to limitations of the system services that help manage time, and hardware-OS and OS-application interfaces through which timing information flows to the application. Due to the importance of time awareness in a broad range of emerging applications, running on commodity platforms and operating systems, it is imperative to rethink how time is handled across the system stack. We advocate the adoption of a holistic notion of Quality of Time (QoT) that captures metrics such as resolution, accuracy, and stability. Building on this notion we propose an architecture in which the local perception of time is a controllable operating system primitive with observable uncertainty, and where time synchronization balances applications' timing demands with system resources such as energy and bandwidth. Our architecture features an expressive application programming interface that is centered around the abstraction of a timeline - a virtual temporal coordinate frame that is defined by an application to provide its components with a shared sense of time, with a desired accuracy and resolution. The timeline abstraction enables developers to easily write applications whose activities are choreographed across time and space. Leveraging open source hardware and software components, we have implemented an initial Linux realization of the proposed timeline-driven QoT stack on a standard embedded computing platform. Results from its evaluation are also presented.

In this paper we propose a fast and energy efficient service provisioning approach. In our work, we focus towards the management of already discovered services. We assert that if the sensor nodes share the important service information among each other, then the service re-discovery can be reduced at greater extent. Hence a significant amount of Service Discover (SD) time and the network energy cost can be saved. We propose the threshold-based technique by devising the formula that determines the importance of the service, and a mechanism of sharing that service through the proposed scheme. We also discuss the performance of different SD protocols with and without implementation of proposed Fast and Energy Efficient Service Provisioning (FESP).

Service discovery in Ubiquitous Sensor Networks has been targeted mostly for services available within certain proximity, but the service availability only in close vicinity is no longer feasible in the pervasive and ubiquitous era. In order to address the ubiquity in service discovery, we have proposed a framework that makes use of the Electronic Number Mapping (ENUM) protocol. Our network architecture consists of sensor nodes associated with few relatively powerful nodes called master nodes. Only master nodes within IPv6 enabled Low power Personal Area Networks (6LoWPANs) are assigned unique E.164 numbers so that the services offered by the network could be accessed globally. Services destined for sensor nodes first reach the master node to which they are associated to, using the E.164 number of the master node. The gateway of the network performs the task of converting attribute-value pair based human readable queries to E.164 numbers. Also the gateway facilitates its network by running ENUM protocol for service sharing like multimedia, mail, web and many other services. Moreover, a significant improvement in service discovery cost in terms of latency and traffic overhead is observed.