INFOCOM 2026 Papers

Survivable routing is crucial in backbone networks to ensure connectivity, even during failures, including natural and man-made disasters. Service-level agreements (SLAs) often define availability targets in terms of "number of nines" (e.g., 99.99%, 99.999%), and network operators aim to meet or exceed these targets by provisioning redundant paths between source-destination pairs. Nevertheless, ensuring extremely high availabilities comes with extra cost in terms of total network bandwidth usage. This study formalizes the problem of availability-aware routing based on risk zone failure probabilities. On the downside, we find the problem is $\mathcal{N}\mathcal{P}$-hard, with minimizing unavailabilities being even inapproximable in polynomial time (supposing $\mathcal{P} \ne \mathcal{N}\mathcal{P}$). Motivated by the hardness results, we initiate the study of heuristics providing availability-aware routing considering regional failures. Our methods are suitable for providing a range of routing plans, navigating the trade-off between availability and bandwidth usage. At the core of the heuristics, we rely on an efficiently solvable problem variant, where failure probabilities of risk zones are changed to capacities. We validate our findings through extensive simulations based on real-world input data. With reasonably low bandwidth usage, our heuristic is close to a lower bound on unavailability, the median and average gap being 1.23% and 2.21%, respectively.

While the adoption of programmable data planes is increasing, writing correct and efficient P4 programs remains a challenging and error-prone process. This hinders the rapid development of custom in-network functionality and limits the accessibility of data-plane programming to a small set of experts. In this paper, we present T4G, a system that automatically synthesizes P4 programs from network behavior expressed as input-output network traces. Unlike prior approaches that rely on high-level abstraction or domain-specific languages, T4G operates with example packet captures, eliminating the need for manual abstraction or rule specifications from network operators. T4G introduces a novel input-output grouping technique that identifies logical packet transformation behavior from packet traces, and performs a divide-and-conquer synthesis that scales to complex forwarding logic. Our experimental results show the ability of T4G to generate different types of programs and demonstrate the potential of T4G in generating complex P4 programs.

Underwater acoustic (UWA) communications are essential for high-speed marine data transmission but remain severely constrained by limited bandwidth, significant propagation loss, and sparse multipath structures. Conventional underwater acoustic multiple-input multiple-output (MIMO) systems primarily utilize spatial diversity but suffer from limited array resolution, causing angular ambiguity and insufficient spatial degrees of freedom (DoFs). This paper addresses these limitations through acoustic Reconfigurable Intelligent Surfaces (aRIS) to actively generate orthogonally distinguishable virtual paths, significantly enhancing spatial DoFs and channel capacity. An ocean-specific DoF-channel coupling model is established, explicitly deriving conditions for spatial rank enhancement. Subsequently, the optimal geometric locus, termed the Light-Point, is analytically identified, where deploying a single aRIS maximizes DoFs by introducing two and three additional resolvable paths in deep-sea and shallow-sea environments, respectively. Furthermore, an active simultaneous transmitting and reflecting (ASTAR) aRIS architecture with independent beam control and adaptive beam-tracking mechanism integrating unmanned underwater vehicles (UUVs) and acoustic intensity gradient sensing is proposed. Extensive simulations validate the proposed joint aRIS deployment and beamforming framework, demonstrating substantial UWA channel capacity improvements-up to 265% and 170% in shallow-sea and deep-sea scenarios, respectively.

True Wireless Stereo earbuds (TWS) have evolved into multifunctional devices with biometric sensing and motion tracking, making secure authentication essential. However, existing methods often require user interaction or access to specific hardware, limiting their applicability in commercial TWS earbuds like AirPods. In this paper, we propose EarAuth, an in-ear Ballistocardiogram-based authentication system that uses the built-in IMU to capture cardiac vibrations for passive, continuous authentication. Unlike movement-dependent IMU methods, EarAuth requires no user effort and is resistant to spoofing, leveraging internal physiological signals. To improve signal reliability, we develop signal processing modules and introduce a personalized heartbeat simulator for increasing physiological diversity. Additionally, using a Siamese network-based metric learning framework, we map in-ear Ballistocardiogram signals to a discriminative embedding space, ensuring robust authentication and adaptation to long-term cardiac feature drift. Our solution is scalable, energy-efficient, and integrates seamlessly with COTS TWS earbuds, preserving core functionalities.

Serverless computing is a promising method to serve Machine Learning (ML) inference via on-demand functions. Due to the time- and memory-consuming ML library and model (i.e., ML artifact) loading, serverless inference endures notable startup overhead and memory waste issues. In this paper, we advocate for hierarchical ML artifact pre-loading and sharing to balance loading and memory efficiency. Building on this, we propose Rocket, a serverless ML inference system that accelerates function startup while reducing memory waste. Rocket dynamically pre-loads partial, shared ML artifacts, each implying a hierarchy of trade-offs between the loading latency and memory usage. Specifically, with a dual-timescale invocation prediction, Rocket first estimates the pre-loading timing for each function, and then schedules them via a sharing-aware agglomerative clustering to improve ML artifact sharing efficiency. In particular, Rocket learns to make the online hierarchical pre-loading decision for function containers based on a lightweight contextual bandit algorithm. Finally, we implement Rocket and evaluate it with realistic workloads. Experimental results display that Rocket outperforms existing solutions by up to 38.7% on startup latency and up to 43.8% on memory saving.

Collective communication becomes increasingly crucial as large language models rapidly evolve, but the RDMA it uses inevitably faces network performance anomalies (NPAs). Vedrfolnir is an accurate and efficient diagnosis system for RDMA NPAs in collective communication, which (1) constructs waiting graphs through algorithm decomposition, (2) adaptively detects anomalies while efficiently collecting diagnostic data, and (3) precisely analyzes performance bottlenecks and root causes. Evaluation shows that Vedrfolnir can achieve accurate diagnosis results with low overhead.

Decentralized Storage Networks (DSNs) are emerging as a foundational infrastructure for Web 3.0, offering global peer-to-peer storage. However, a critical vulnerability persists: user privacy during file retrieval remains largely unaddressed, risking the exposure of sensitive information. To overcome this, we introduce PIR-DSN, the first DSN protocol to integrate Private Information Retrieval (PIR) for both single and multi-server settings. Our key innovations include a novel secure mapping method that transforms sparse file identifiers into compact integer indexes, enabling both public verifiability of file operations and efficient private retrieval. Furthermore, PIR-DSN guarantees Byzantine-robust private retrieval through file replication across multiple miners. We implement and rigorously evaluate PIR-DSN against three prominent industrial DSN systems. Experimental results demonstrate that PIR-DSN achieves comparable overhead for file upload and deletion. While PIR inherently introduces an additional computational cost leading to higher retrieval latency, PIR-DSN maintains comparable throughput. These findings underscore PIR-DSN’s practical viability for privacy-sensitive applications within DSN environments.

The rise of IoT networks has driven demand for robust, privacy-preserving machine learning in distributed, resource-constrained environments. Federated Learning (FL) addresses this by enabling on-device training without sharing raw data, but remains vulnerable to data poisoning from adversarial clients. Ensuring auditable data quality and verifiable model contributions is essential for securing FL and challenging due to limited client data visibility. Existing defenses often rely on trusted central servers for client selection or aggregation, offer limited visibility into client data quality and often fail to identify clients contributing low-quality or poisoned data. To overcome these limitations, we propose Statistical Auditing and Filtering for Enabling Robust FL (SAFER-FL), a data-aware framework for adversarially robust FL in IoT networks. SAFER-FL builds a global, privacy-preserved statistical profile by aggregating Gaussian-noised client-side data moments (mean, covariance) in latent space. Clients use this to filter poisoned samples via Mahalanobis distance before training, ensuring only clean data influences the model. This proactive approach eliminates the need for validation data or pretrained models. We provide theoretical analysis on convergence, filtering efficacy, and poisoning resilience. Experiments on five IoT datasets show SAFER-FL significantly improves robustness and accuracy under various attacks, making it practical for real-world IoT FL deployments.

Jamming attacks pose a critical threat to wireless networks, yet existing detection methods remain largely uni-modal, centralized and resource-intensive, limiting their performance, scalability, and deployment feasibility, respectively. To address these limitations, we present FedJam, a multimodal Federated Learning (FL) framework for on-device jamming detection and classification. FedJam locally fuses spectrograms and cross-layer network Key Performance Indicators (KPIs) using a lightweight dual-encoder architecture with an integrated fusion module and multimodal projection head, that enables privacy-preserving training and inference without transmitting raw data. We prototype and deploy FedJam on a wireless experimental testbed and evaluate it using the first, over-the-air multimodal dataset comprising synchronized samples across benign and three distinct jamming attack types. FedJam outperforms state-of-the-art unimodal baselines by up to 15% in accuracy, while requiring 60% fewer communication rounds to converge, and maintains low resource utilization. Its advantage is especially pronounced in realistic scenarios, where it remains extremely robust under heterogeneous data distributions across devices.

The emergence of Open Radio Access Networks (O-RAN) promises intelligent sensing applications over 5G and beyond mobile networks provided 5G IQ samples for reference signals can be accessed from the RAN stack (DU) at the applications running over RAN Intelligent Controller (RICs). However, reference signals like DMRS generate multi-Gbps data streams per RU and across multiple RUs can easily overwhelm the network connection between the DU and RIC in typical deployments. Existing compression approaches—transform coding assuming signal sparsity, compressive sensing with high computational overhead, and machine-learning models requiring bespoke hardware or extensive training—either degrade sensing fidelity or violate sub-second near-RT RIC latency deadlines. We present TAPS, a Three-Dimensional Amplitude–Phase–Spatial algorithm that leverages (i) cross-symbol amplitude correlation, (ii) linear phase structure across subcarriers, and (iii) spatial redundancy across antennas, all within a CPU-only, O-RAN telemetry pipeline. With DMRS data collected from an outdoor 5G O-RAN testbed, TAPS delivers 10-30× data reduction and reduces reconstruction error by 50–67% relative to state-of-the-art baselines, while maintaining end-to-end compression and decompression latency below 100 ms, thereby overcoming the technical bottlenecks for high-fidelity sensing over O-RAN based 5G and beyond mobile networks.

Non-line-of-sight (NLoS) radar has the potential to significantly improve road safety. Signals in the sub-6 GHz bands are well known to provide good signal strength in such scenarios. This paper considers a bistatic radar that uses a 5G base station (BS) as transmitter (TX), while a user equipment (UE) integrated into an ego vehicle serves as receiver (RX); the goal is to identify target vehicles that may not have line-of-sight (LoS) to the ego vehicle. However, a key challenge is the oscillator/clock synchronization of TX and RX since traditional synchronization algorithms designed for communications purposes are insufficient for radar sensing. We address this issue by developing an algorithm that leverages map priors to identify significant reflected multipath components (MPCs) whose delay and Doppler is leveraged for determining and tracking the timing offset (TO) and carrier frequency offset (CFO) of the RX clock. We also develop a new clutter suppression method, as traditional background subtraction does not work in scenarios with moving RX. We demonstrate the practical feasibility of our proposed techniques via extensive real-world experiments. Our results show absolute synchronization with mean absolute error (MAE) of TO < 4 ns and CFO < 2 Hz, and successful identification of NLoS targets.

With the proliferation of heterogeneous software-hardware infrastructures in camera deployment, video analytics pipelines (VAPs) are increasingly burdened by spatiotemporal workload imbalance, where uneven task distributions lead to latency constraint violations and degraded quality of experience (QoE). Eliminating this imbalance is challenging due to the inherent complexity of adjusting large-scale parameters and the dynamic nature of VAP runtime environments. To this end, we propose Hier-EI, a novel scheduling framework that combines the two-phase hierarchical design with embodied intelligence, which adaptively tunes system parameters to mitigate imbalance and maintain long-term service-level objective (SLO) performance for modern VAP systems. To tackle the complexity of combinatorial decision-making, we introduce a hierarchical collaboration mechanism with macro-micro coordination that transforms the exponential search space into a linear coarse-to-fine workflow. To adapt to runtime dynamics, we present an embodied feedback mechanism that employs closed-loop feedback to converge toward real-time optimal solutions as a Markov decision process. Extensive evaluations on a real-world prototype system built on KubeEdge demonstrate that Hier-EI achieves a 3.6× improvement in latency compliance, and a 67.4% reduction in P95 latency compared with state-of-the-art scheduling methods.

Serverless computing is increasingly adopted for AI-driven workloads due to its automatic scaling and pay-as-you-go model. However, its function-based architecture creates significant security risks, including excessive privilege allocation and poor permission management. In this paper, we present ALPS, an automated framework for enforcing least privilege in serverless environments. Our system employs serverless-tailored static analysis to extract precise permission requirements from function code and a fine-tuned Large Language Model (LLM) to generate language- and vendor-specific security policies. It also performs real-time monitoring to block unauthorized access and adapt to policy or code changes, supporting heterogeneous cloud providers and programming languages. In an evaluation of 8,322 real-world functions across AWS, Google Cloud, and Azure, ALPS achieved 94.8% coverage for least-privilege extraction, improved security logic generation quality by 220% (BLEU), 124% (ChrF++) and 100% (ROUGE-2), and added minimum performance overhead. These results demonstrate that ALPS provides an effective, practical, and vendor-agnostic solution for securing serverless workloads.

In the era of massive network traffic and sophisticated cyber threats, accurate per-flow cardinality measurement has become the cornerstone of modern traffic analytics and network security. However, traditional sketch-based solutions face a fundamental dilemma: they either excel at cardinality estimation or super-spreader detection, but not both, while being limited to either biased or unbiased estimations. This limitation forces network operators to deploy multiple specialized systems, increasing complexity and resource overhead. One-Sketch represents a paradigm shift in network measurement by unifying dual functionality within a single architecture. Specifically, our approach introduces a novel hybrid design combining a heavy part for tracking elephant flows with specialized unbiased cardinality estimators for remaining traffic. This architecture delivers simultaneous high-accuracy cardinality estimation and super-spreader detection while providing flexible bias control to meet diverse application demands. Furthermore, our approach introduces a cardinality buffering technique, which dramatically enhances throughput by enabling real-time estimation capabilities that traditional single-flow estimators cannot achieve. Extensive evaluation on real-world network traces demonstrates One-Sketch’s superior performance: achieving the highest accuracy for super-spreader detection while maintaining comparable cardinality estimation precision, with significantly higher throughput than existing solutions.

Recent advancements in array-camera videography enable real-time capturing of ultra-high-definition (Ultra-HD) videos, providing rich visual information in a large field of view. However, promptly processing such data using state-of-the-art transformer-based vision foundation models faces significant computational overhead in on-device computing or transmission overhead in cloud computing. In this paper, we present Hyperion, the first cloud-device collaborative framework that enables low-latency inference on Ultra-HD vision data using off-the-shelf vision transformers over dynamic networks. Hyperion addresses the computational and transmission bottleneck of Ultra-HD vision Transformer inference by exploiting the intrinsic property in vision Transformer models. Specifically, Hyperion integrates a collaboration-aware importance scorer that identifies critical regions at the patch level, a dynamic scheduler that adaptively adjusts patch transmission quality to balance latency and accuracy under dynamic network conditions, and a weighted ensembler that fuses edge and cloud results to improve accuracy. Experimental results on real-world prototypes and datasets demonstrate that Hyperion enhances frame processing rate by up to 1.61 × and improves the accuracy by up to 20.2% when compared with state-of-the-art baselines under various network environments.

Low earth orbit (LEO) satellites play a pivotal role in global connectivity—delivering high-speed Internet, cellular coverage, and massive IoT support. With ever-growing onboard computing and storage resources, LEO satellites herald a new cloud paradigm: space cloud computing. While container orchestration platforms (e.g., Kubernetes) excel in terrestrial data centers, they are ill-suited to LEO satellite networks, featuring geographic dispersion and frequent handovers. Those features bring high latency and intermittent management, leading to control plane failure in container orchestration. To address this, we propose KubeSpace, a low-latency and stable control plane specifically designed for container orchestration on LEO satellites. KubeSpace combines two key innovations: a distributed ground-control-node architecture that binds each satellite to its nearest controller for uninterrupted management, and an orbit-aware placement with dynamic assignment strategy that further minimizes communication latency and handover frequency. Extensive experiments based on real satellite traces demonstrate that compared to existing solutions, KubeSpace reduces the average management latency of satellite nodes by 59% without any management interruption time.

With the rapid advancement of communication technologies, wireless data transmissions have become pervasive. However, the open nature of the wireless medium makes these transmissions vulnerable to attacks, such as malicious detection and eavesdropping. Importantly, the prerequisite for these threats is the attackers’ ability to detect the existence of legitimate transmissions. Covert communication (CovCom), a physical-layer security (PLS) technique designed to conceal legitimate transmissions from detection by attackers, serves as a fundamental security measure unlike other PLS schemes. Nevertheless, most existing CovCom schemes rely on covering signals that are independent of the covert signals themselves. This independence allows attackers to enhance their signal processing capabilities to separate the covert signal from the covering signal, ultimately extracting the original signal. Furthermore, the necessity for dedicated covering signal generation incurs additional power and hardware costs. To address these deficiencies, we propose a novel scheme called Generative Covert Communication (GCC), aimed at achieving covert data transmissions. GCC leverages interactions among covering signals; by adjusting their power and/or phases, a desired secret signal intended for the covert receiver (Rx) can be produced. Since the covert signals are derived from covering signals, they become coupled, and hence prevent attackers from decoupling them and detecting the existence of, as well as the content carried by, the secret transmissions. We design two GCC schemes: Power Control GCC (PC-GCC) and Phase Control GCC (${\mathbb{F}}$C-GCC). We utilize (1) the Universal Software Radio Peripheral (USRP) platform for experimental validation of GCC’s feasibility and (2) comprehensive MATLAB simulations to demonstrate its effectiveness in achieving covert and secure data transmissions.

Sharded blockchains enhance scalability by partitioning the system into parallel shards. However, even with parallel processing, the substantial coordination overhead and suboptimal placement of interdependent contracts lead to significant performance degradation when executing cross-shard transactions (CSTxs). Therefore, efficiently managing costly CSTxs through a dependency-aware placement strategy presents a significant challenge. In this paper, we propose JiangZiya, a dependency-aware smart contract migration mechanism that improves CSTx execution efficiency. We model contract dependencies as a gas-weighted invocation graph to reflect realistic execution costs. Furthermore, we employ a genetic algorithm to optimize contract placement and a dual-mode protocol to execute remaining CSTxs, complemented by a lightweight storage module. Finally, we conduct extensive experiments to compare our work with state-of-the-art baselines on real-world workloads. The evaluation results show that JiangZiya improves throughput by up to 4.2× and reduces latency by up to 74%.

We present Loki, an approach to accelerate existing software-based CUs using commodity network hardware infrastructure, such as network switches and SmartNICs. Loki uses an in-band stateful migration protocol to handle the transfer of user-plane traffic between software and hardware. The protocol incorporates techniques to manage state inconsistency caused by packet loss and reordering. To improve performance, Loki dynamically moves user traffic processing between hardware and software to maximize hardware usage while reducing the workload on the software. We prototype Loki using an Intel Tofino2 switch coupled with an OCTEON TX2 CN96XX ARM SoC and integrate Loki with an open-source 5G stack. Our evaluations on a real 5G testbed show that Loki achieves orders-of-magnitude improvements in throughput and latency, demonstrating its effectiveness and practical deployability.