INFOCOM 2026 Papers

The commercial success of live stream- ing—especially for major sports events—is putting increasing pressure on Internet Service Providers (ISPs) and raises questions about pricing, service quality, and network neutrality. A key open problem is understanding how congestion, pricing, and side payments affect user choices and market outcomes. This paper introduces a modeling framework to study in- teractions among users, ISPs, and content providers (CPs) in congestion-sensitive environments. The framework combines quality-of-experience modeling, discrete choice theory, and evo- lutionary population dynamics, with particular attention to side payments from CPs to ISPs and their impact on user allocation and revenues. Using this framework, we analyze how side payments may affect ISP market shares and give rise to potential neutrality and fairness concerns. Results show that side payments can benefit both CPs and ISPs, but that their allocation among ISPs plays a critical role in limiting induced market distortions.

Service mesh has become an essential infrastructure for managing cloud-native microservices, widely adopted by major providers to streamline service orchestration and reduce operational overhead. A core component of service mesh architecture is the sidecar proxy, managing policy-based routing, Layer-7 load balancing, and related functions. Traditional per-pod distributed sidecar deployments route all inter-pod communication through local proxies, introducing substantial resource consumption and inefficiencies. Recent approaches advocate for centralized proxy deployments to offload compute-intensive modules to gateway nodes; however, this design introduces non-trivial traffic detours and exacerbates network congestion.To overcome these limitations, we propose PiggyCar, a novel service mesh communication system that piggybacks Layer-7 traffic control onto programmable network devices. In PiggyCar, resource-intensive tasks are offloaded to intermediary switches along inter-pod paths, enabling in-network execution of advanced network functions. PiggyCar incorporates a latency-aware offline planner for optimal network policy placement and a stability-oriented online scheduler that dynamically adapts to traffic fluctuations in real time. We prototype PiggyCar on a testbed comprising six servers and four programmable switches. Experimental results show that PiggyCar cuts latency by up to 97.2% compared to Canal, boosts throughput by up to 1.82× under high RPS conditions, and consistently delivers the lowest network jitter.

Time-Sensitive Networking (TSN) underpins time-critical industrial systems with ultra-low-latency, deterministic communication. However, long-term observations in real factories reveal that temporal uncertainty from legacy devices (jitter and clock drift) and topology churn from reconfigurable workflows break the assumptions of existing TSN schedulers, sharply degrading their performance. Therefore, we present RhythmScheduler, a fast, scalable TSN scheduler that explicitly models jitter and drift via three components: a temporal–stochastic graph encoder capturing timing variation and topology; a dual-head Transformer decoupling offset selection from guard-band allocation; and a Bayesian meta-adaptation loop on a hybrid digital twin, achieving sub-second reconfiguration with 3–4 ms inference. Extensive experiments on both simulation and hardware-in-the-loop testbeds show that RhythmScheduler improves schedulability by 12.2%/25.9%, reduces guard-band waste by 55.6%/62.6%, and generates Gate Control Lists (GCLs) 11.2×/14.8× faster on 200-link networks compared to state-of-the-art RL/heuristics-based methods; and achieves a further 52.9% schedulability gain, 52.5% waste reduction, and 10.9× speedup on 400-link networks, where prior RL baselines fail due to resource exhaustion. By combining efficiency with robustness, RhythmScheduler bridges deterministic TSN theory and noisy shop-floor reality, removing traffic scheduling as a bottleneck to agile industrial automation.

Modern storage systems replicate new data for fast access and later encode aged data into erasure coding for storage efficiency. However, the encoding process incurs substantial cross-rack traffic and affects the future repair cost. Existing approaches compromise either post-encoding repair efficiency or on-encoding performance due to overlooking the characteristics of the erasure-coded layout. We propose Repair-Oriented Encoding (ROE), a scheme for erasure codes with localities that jointly minimizes the on-encoding cross-rack traffic and post-encoding repair cost. ROE targets the repair-optimal coding layout to reduce future repair cost and proactively aligns the replication layout to match the coding layout. It further exploits localities to offload the computation of local parity blocks to their destination racks, often eliminating the on-encoding traffic entirely. Experiments show that ROE reduces the encoding time by up to 97.8% and post-encoding repair time by up to 94.7%, while maintaining comparable replication read/write performance.

Wi-Fi localization and tracking are increasingly becoming the key enabler for intelligent services and lead to a wide range of pioneering techniques. However, most existing systems rely on prior knowledge of the device position (DP) and the user’s initial position (IP) to model, which is often unavailable in dynamic real-world deployments. To overcome this limitation, we propose AutoLoc, a system that jointly estimates DP and IP without intensive manual calibration or dedicated hardware. Our core insight lies in the spatial heterogeneity of Wi-Fi signal features. Specifically, we develop a reconstruction pipeline, which first predicts user trajectories based on observed features, then infers positions by aligning reconstructed features from trajectories with observations. To further enhance accuracy and efficiency, we introduce an alternating iteration and multi-trace optimization. We build a prototype with the COTS device and conduct experiments across diverse real-world scenarios. Extensive experiments demonstrate that our system achieves localization errors of 0.43m for the device and 0.45m for the initial position with only one labeled link, while also achieving tracking accuracy comparable to state-of-the-art systems. AutoLoc aims to offer a low-cost, practical solution for Wi-Fi sensing and paves the way toward plug-and-play deployment.

Network traffic anomaly detection represents a critical cybersecurity task, yet widespread encryption makes this task increasingly challenging. In response, image-based methods that model traffic as visual patterns have emerged as the dominant approach. However, this work pioneers the identification of a pervasive "full-frequency"characteristic and an associated limitation termed "spectral mismatch"within this paradigm. Specifically, while encrypted traffic exhibits prominent high-frequency components, mainstream reconstruction methods demonstrate an inherent bias toward learning low-frequency information. This fundamental mismatch results in incomplete representations that consequently degrade anomaly detection performance. To address this challenge, we propose FreeUp, a novel frequency-decoupled framework designed explicitly for encrypted traffic analysis. FreeUp decomposes traffic data into distinct low- and high-frequency bands, processing them through separate, dedicated branches along with a customized training strategy that ensures stable and independent frequency-specific learning. Furthermore, recognizing that simple reconstruction error proves inadequate for evaluating dual-branch architectures, we introduce an uncertainty-inspired fusion scoring mechanism. This mechanism quantifies the reconstruction uncertainty of the frequency-specific branches and dynamically integrates their outputs, yielding a more comprehensive and reliable anomaly score. Extensive experiments across multiple benchmarks demonstrate that FreeUp consistently outperforms state-of-the-art baselines. The code is available at https://github.com/ikun0124/FreeUp.