INFOCOM 2024 论文列表

Anomaly-based network intrusion detection systems (A-NIDS) use unsupervised models to detect unforeseen attacks. However, existing A-NIDS solutions suffer from low throughput, lack of interpretability, and high maintenance costs. Recent in-network intelligence (INI) exploits programmable switches to offer line-rate deployment of NIDS. Nevertheless, current in-network NIDS are either model-specific or only apply to supervised models. In this paper, we propose Genos, a general in-network framework for unsupervised A-NIDS by rule extraction, which consists of a Model Compiler, a Model Interpreter, and a Model Debugger. Specifically, observing benign data are multi-modal and usually located in multiple subspaces in the feature space, we utilize a divide-and-conquer approach for model-agnostic rule extraction. In the Model Compiler, we first propose a tree-based clustering algorithm to partition the feature space into subspaces, then design a decision boundary estimation mechanism to approximate the source model in each subspace. The Model Interpreter interprets predictions by important attributes to aid network operators in understanding the predictions. The Model Debugger conducts incremental updating to rectify errors by only fine-tuning rules on affected subspaces, thus reducing maintenance costs. We implement a prototype using physical hardware, and experiments demonstrate its superior performance of 100 Gbps throughput, great interpretability, and trivial updating overhead.

Unmanned aerial vehicles (UAVs)-enabled aerial communication provides a flexible, reliable, and cost-effective solution for a range of wireless applications. However, due to the high line-of-sight (LoS) probability, aerial communications between UAVs are vulnerable to eavesdropping attacks, particularly when multiple eavesdroppers collude. In this work, we aim to introduce distributed collaborative beamforming (DCB) into UAV swarms and handle the eavesdropper collusion by controlling the corresponding signal distributions. Specifically, we consider a two-way DCB-enabled aerial communication between two UAV swarms and construct these swarms as two UAV virtual antenna arrays. Then, we minimize the two-way known secrecy capacity and the maximum sidelobe level to avoid information leakage from the known and unknown eavesdroppers, respectively. Simultaneously, we also minimize the energy consumption of UAVs for constructing virtual antenna arrays. Due to the conflicting relationships between secure performance and energy efficiency, we consider these objectives as a multi-objective optimization problem. Following this, we propose an enhanced multi-objective swarm intelligence algorithm via the characterized properties of the problem. Simulation results show that our proposed algorithm can obtain a set of informative solutions and outperform other state-of-the-art baseline algorithms. Experimental tests demonstrate that our method can be deployed in limited computing power platforms of UAVs and is beneficial for saving computational resources.

Large-scale datacenter networks are increasingly using in-network aggregation (INA) and remote direct memory access (RDMA) techniques to accelerate deep neural network (DNN) training. However, existing research trends suggest that these two techniques are on an inevitable collision course. To fill this gap, we present FreeINA, a host-driven in-network aggregation aimed at providing RDMA reliable connection (RC) for multi-tenant learning settings. FreeINA relies on dual transmission paths to support RC compatibility, with one path for INA and another one for aggregation on end-host parameter server. With dynamic control of these two paths, FreeINA can leave the traditional in-server aggregation unaffected while ensuring INA’s reliability without modifying RDMA network interfaces (RNICs). To support multi-tenant learning, FreeINA employs all-reduce-level memory allocation, which can capture the well-known "on and off" DNN training pattern and thus improve switch memory efficiency. We have implemented a FreeINA prototype using P4-programmable switch and commercial RNICs, and evaluated it extensively using 100Gbps testbed. The results show that compared to the state-of-the-art solution—ATP, FreeINA improves single-job training speedup ratio by 1.20×, while improving the aggregation throughput by 2.65× in multi-job scenario.

Local clustering aims to find a high-quality cluster near a given vertex. Recently, higher-order units are introduced to local clustering, and the underlying information has been verified to be essential. However, original edges are underestimated in these techniques, leading to the degeneration of network information. Moreover, most of the higher-order models are designed for static networks, whereas real-world networks are generally large and evolve rapidly. Repeatedly conducting a static algorithm at each snapshot is usually computationally impractical, and recent approaches instead track a cluster by updating the cluster sequentially. However, errors would accumulate over lengthy evolutions, and the complete cluster needs to be recalculated periodically to maintain the accuracy, which naturally affects the efficiency. To bridge the two gaps, we design a multi-order hypergraph, and present a hybrid model for dynamic clustering. In particular, we propose an incremental method to track a personalized PageRank vector in the evolving hypergraph, which converges to the exact solution at each snapshot when significantly reducing the complexity. We further develop a dynamic sweep to identify a cut in each vector, whereby a cluster can be incrementally updated with no accumulated errors. We provide rigorous theoretical basis and conduct comprehensive experiments, which demonstrate the effectiveness.

Virtualized radio access networks (vRAN) enable network operators to run RAN functions on commodity servers instead of proprietary hardware. It has garnered significant interest due to its ability to reduce costs, provide deployment flexibility, and offer other benefits, particularly for operators of 5G private networks. However, the non-deterministic computing platforms pose difficulties to effective quality of service (QoS) provision, especially in the case of hybrid deployment of time-critical and throughput-demanding applications. Existing approaches including network slicing and other resource management schemes fail to provide fine-grained and effective QoS support at the User Equipments level. In this paper, we propose UQ-vRAN, a UE-level QoS provision framework. UQ-vRAN presents the first comprehensive analysis of the complicated impacts among key network parameters, e.g., network function splitting, resource block allocation, and modulation/coding scheme selection and builds an accurate and comprehensive network model. UQ-vRAN also provides a fast network configurator which gives feasible configurations in seconds, making it possible to be practical in actual 5G vRAN. We implement UQ-vRAN on OpenAirInterface and use simulation and testbed-base experiments to evaluate it. Results show that compared with existing works, UQ-vRAN reduces the delay violation rate by 12%–41% under various network settings, while minimizing the total energy consumption.

Recently, deep autoencoders have gained traction as a powerful method for implementing goal-oriented semantic communications systems. The idea is to train a mapping from the source domain directly to channel symbols, and vice versa. However, prior studies often focused on rate-distortion tradeoff and transmission delay, at the cost of increasing end-to-end complexity and thus latency. Moreover, the datasets used are often not reflective of real-world environments, and the results were not validated against real-world baseline systems, leading to an unfair comparison. In this paper, we study the problem of remote camera pose estimation and propose AdaSem, an adaptive semantic communications approach that optimizes the tradeoff between inference accuracy and end-to-end latency. We develop an adaptive semantic codec model, which encodes the source data into a dynamic number of symbols, based on the latent space distribution and the channel state feedback. We utilize a lightweight model for both transmitter and receiver to ensure comparable complexity to the baseline implemented in a real- world system. Extensive experiments on real-environment data show the effectiveness of our approach. When compared to a real implementation of a client-server camera relocalization service, AdaSem outperforms the baseline by reducing the end-to-end delay and estimation error by over 75% and 63%, respectively.

Recently, deploying static chargers to construct timely and robust Wireless Rechargeable Sensor Networks (WRSNs) has become an important research issue for solving the limited energy problem of wireless sensor networks. However, the established fixed power distribution lacks flexibility in response to dynamic charging requests from sensors and may render some sensors to be continuously impacted by destructive wave interference. This results in a gap between energy supply and practical demand, making the charging process less efficient. In this paper, we focus on the real-time sensor charging requests and formulate a dynamic power disTributIon controlling for Directional chargErs (TIDE) problem to maximize the overall charging utility. To solve the problem, we first build a charging model for directional chargers while considering wave interference and extract the candidate charging orientations from the continuous search space. Then we propose the neighbor set division method to narrow the scope of calculation. Finally, we design a dynamic power distribution controlling algorithm to update the neighbor sets timely and select optimal orientations for chargers. Our experimental results demonstrate the effectiveness and efficiency of the proposed scheme, it outperforms the comparison algorithms by 142.62% on average.

Sound eavesdropping using light has been an area of considerable interest and concern, as it can be achieved over long distances. However, previous work has often lacked stealth (e.g., active emission of laser beams) or been limited in the range of realistic applications (e.g., using direct light from a device’s indicator LED or a hanging light bulb). In this paper, we present EchoLight, a non-intrusive, passive and long-range sound eavesdropping method that utilizes the extensive reflection of ambient light from vibrating objects to reconstruct sound. We analyze the relationship between reflection light signals and sound signals, particularly in situations where the frequency response of reflective objects and the efficiency of diffuse reflection are suboptimal. Based on this analysis, we have introduced an algorithm based on cGAN to address the issues of nonlinear distortion and spectral absence in the frequency domain of sound. We extensively evaluate EchoLight’s performance in a variety of real-world scenarios. It demonstrates the ability to accurately reconstruct audio from a variety of source distances, attack distances, sound levels, light intensity, light sources, and reflective materials. Our results reveal that the reconstructed audio exhibits a high degree of similarity to the original audio over 40 meters of attack distance.

Automated spectrum analytics inform critical decisions in dynamic spectrum access networks such as (i) how to allocate network resources to clients, (ii) when to enforce penalties due to malicious or disruptive activity, and (iii) how to chart policies for future regulations. The insights gleaned from a spectrum trace, however, are as objective as the trace itself, and artifacts introduced by sensor imperfections or improper configuration will inevitably affect analysis outcomes. Yet, spectrum analytics have been largely developed in isolation from the underlying data collection and are oblivious to sensor-induced artifacts.To address this challenge, we develop VIA, a framework that attributes sensor properties and configuration to spectrum data fidelity, and models the relationship between spectrum analytics performance and data quality. VIA does not require expert input or intervention and can be used to profile the fidelity of unknown sensors. VIA takes as an input a spectrum trace and the sensor configuration, and benchmarks data quality along three dimensions: (i) Veracity, or how truthfully a scan captures spectrum activity, (ii) Intermittency, characterizing the temporal persistence of spectrum scans and (iii) Ambiguity quantifying the likelihood of false detection. We employ VIA to measure the data fidelity of five common sensor platforms. We then predict the outcome of several spectrum analysis tasks including occupancy and transmitter detection, and modulation recognition using both controlled and real-world measurements. We demonstrate high prediction performance with an average mean squared error of 0.0013 across all tasks using both regression and neural network models.

Distributed deep learning has been widely employed to train deep neural network over large-scale dataset. However, the commonly used parameter server architecture suffers from long synchronization time in data-parallel training. Although the existing solutions are proposed to reduce synchronization overhead by breaking the synchronization barriers or limiting the staleness bound, they inevitably experience low convergence efficiency and long synchronization waiting. To address these problems, we propose Gsyn to reduce both synchronization overhead and staleness. Specifically, Gsyn divides workers into multiple groups. The workers in the same group coordinate with each other using the bulk synchronous parallel scheme to achieve high convergence efficiency, and each group communicates with parameter server asynchronously to reduce the synchronization waiting time, consequently increasing the convergence efficiency. Furthermore, we theoretically analyze the optimal number of groups to achieve a good tradeoff between staleness and synchronization waiting. The evaluation test in the realistic cluster with multiple training tasks demonstrates that Gsyn is beneficial and accelerates distributed training by up to 27% over the state-of-the-art solutions.

Modern telco networks have experienced a significant paradigm shift in the past decade, thanks to the proliferation of network softwarization. Despite the benefits of softwarized networks, the constituent software data planes cannot always guarantee predictable performance due to resource contentions in the underlying shared infrastructure. Performance predictions are thus paramount for network operators to fulfill Service-Level Agreements (SLAs), especially in high-speed regimes (e.g., Gigabit or Terabit Ethernet). Existing solutions heavily rely on in-band feature collection, which imposes non-trivial engineering and data-path overhead. This paper proposes a non-invasive performance prediction approach, which complements state-of-the-art solutions by measuring and analyzing low-level features ubiquitously available in the network infrastructure. Accessing these features does not hamper the packet data path. Our approach does not rely on prior knowledge of the input traffic, VNFs’ internals, and system details. We show that (i) low-level hardware features exposed by the NFV infrastructure can be collected and interpreted for performance issues, (ii) predictive models can be derived with classical ML algorithms, (iii) and can be used to predict performance impairments in real NFV systems accurately. Our code and datasets are publicly available1.

With edge intelligence, AI models are increasingly push to the edge to serve ubiquitous users. However, due to the drift of model, data, and task, AI model deployed at the edge suffers from degraded accuracy in the inference serving phase. Model retraining handles such drifts by periodically retraining the model with newly arrived data. When colocating model retraining and model inference serving for the same model on resource-limited edge servers, a fundamental challenge arises in balancing the resource allocation for model retraining and inference, aiming to maximize long-term inference accuracy. This problem is particularly difficult due to the underlying mathematical formulation being time-coupled, non-convex, and NP-hard. To address these challenges, we introduce a lightweight and explainable online approximation algorithm, named ORRIC, designed to optimize resource allocation for adaptively balancing the accuracy of model training and inference. The competitive ratio of ORRIC outperforms that of the traditional Inference-Only paradigm, especially when data drift persists for a sufficiently lengthy time. This highlights the advantages and applicable scenarios of colocating model retraining and inference. Notably, ORRIC can be translated into several heuristic algorithms for different resource environments. Experiments conducted in real scenarios validate the effectiveness of ORRIC.

Mobile networks have increased spectral efficiency through advanced multiplexing strategies that are coordinated by base stations (BS) in licensed spectrum. However, external interference on clients leads to significant performance degradation during dynamic (unlicensed) spectrum access (DSA). We introduce the notion of network tomography for DSA, whereby clients are transformed into spectrum sensors, whose joint access statistics are measured and used to account for interfering sources. Albeit promising, performing such tomography naively incurs an impractical overhead that scales exponentially with the multiplexing order of the strategies deployed – which will only continue to grow with 5G/6G technologies.To this end, we propose a novel, scalable network tomography framework called NeTo-X that estimates joint client access statistics with just linear overhead, and forms a blue-print of the interference, thus enabling efficient DSA for future networks. NeTo-X’s design incorporates intelligent algorithms that leverage multi-channel diversity and the spatial locality of interference impact on clients to accurately estimate the desired interference statistics from just pair-wise measurements of its clients. The merits of its framework are showcased in the context of resource management and jammer localization applications, where its performance significantly outperforms baseline approaches and closely approximates optimal performance at a scalable overhead.

RAN virtualization is revolutionizing the telco industry, enabling 5G Distributed Units to run using general-purpose platforms equipped with Hardware Accelerators (HAs). Recently, GPUs have been proposed as HAs, hinging on their unique capability to execute 5G PHY operations efficiently while also processing Machine Learning (ML) workloads. While this ambivalence makes GPUs attractive for cost-effective deployments, we experimentally demonstrate that multiplexing 5G and ML workloads in GPUs is in fact challenging, and that using conventional GPU-sharing methods can severely disrupt 5G operations. We then introduce YinYangRAN, an innovative O-RAN-compliant solution that supervises GPU-based HAs so as to ensure reliability in the 5G processing pipeline while maximizing the throughput of concurrent ML services. YinYangRAN performs GPU resource allocation decisions via a computationally-efficient approximate dynamic programming technique, which is informed by a neural network trained on real-world measurements. Using workloads collected in real RANs, we demonstrate that YinYangRAN can achieve over 50% higher 5G processing reliability than conventional GPU sharing models with minimal impact on co-located ML workloads. To our knowledge, this is the first work identifying and addressing the complex problem of HA management in emerging GPU-accelerated vRANs, and represents a promising step towards multiplexing PHY and ML workloads in mobile networks.

Large-scale storage systems boost I/O performance via flash caching, but the underlying storage medium of flash caching incurs significant financial costs and also exhibits low endurance. Previous studies adopt compression-after-deduplication to mitigate writing redundant contents into the flash cache, so as to address the cost and endurance issues. However, deduplication and compression have conflicting preferable cases, and compression-after-deduplication essentially compromises the space-saving benefits of either deduplication or compression. To simultaneously preserve the benefits of both approaches, we explore compression-before-deduplication, which applies compression to eliminate byte-level redundancies across data blocks, followed by deduplication to write only a single copy of duplicate compressed blocks into the flash cache. We present CDCache, a space-efficient flash caching system that realizes compression-before-deduplication. It proposes to dynamically adjust the compression range of data blocks, so as to preserve the effectiveness of deduplication on the compressed blocks. Also, it builds on various design techniques to approximately estimate duplicate data blocks and efficiently manage compressed blocks. Trace-driven experiments show that CDCache improves the read hit ratio and the write reduction ratio of a previous compression-after-deduplication approach by up to 1.3× and 1.6×, respectively, while it only has small memory overhead for index management.

Point cloud registration (PCR) can significantly extend the visual field and enhance the point density on distant objects, thereby improving driving safety. However, it is very challenging for vehicles to perform online registration between long-range point clouds. In this paper, we propose an online long-range PCR scheme in VANETs, called LoRaPCR, where vehicles achieve long-range registration through multi-hop short-range highly-accurate registrations. Given the NP-hardness of the problem, a heuristic algorithm is developed to determine best registration paths while leveraging the reuse of registration results to reduce computation costs. Moreover, we utilize an optimized dynamic programming algorithm to determine the transmission routes while minimizing the communication overhead. Results of extensive simulations demonstrate that LoRaPCR can achieve high PCR accuracy with low relative translation and rotation errors of 0.55 meters and 1.43°, respectively, at a distance of over 100 meters, and reduce the computation overhead by more than 50% compared to the state-of-the-art method.

Recently, federated learning (FL) has gained momentum because of its capability in preserving data privacy. To conduct model training by FL, multiple clients exchange model updates with a parameter server via Internet. To accelerate the communication speed, it has been explored to deploy a programmable switch (PS) in lieu of the parameter server to coordinate clients. The challenge to deploy the PS in FL lies in its scarce memory space, prohibiting running memory consuming aggregation algorithms on the PS. To overcome this challenge, we propose Federated Learning in-network Aggregation with Compression (FediAC) algorithm, consisting of two phases: client voting and model aggregating. In the former phase, clients report their significant model update indices to the PS to estimate global significant model updates. In the latter phase, clients upload global significant model updates to the PS for aggregation. FediAC consumes much less memory space and communication traffic than existing works because the first phase can guarantee consensus compression across clients. The PS easily aligns model update indices to swiftly complete aggregation in the second phase. Finally, we conduct extensive experiments by using public datasets to demonstrate that FediAC remarkably surpasses the state-of-the-art baselines in terms of model accuracy and communication traffic.

Federated Learning (FL) enables training Artificial Intelligence (AI) models over end devices without compromising their privacy. As computing tasks are increasingly performed by a combination of cloud, edge, and end devices, FL can benefit from this End-Edge-Cloud Collaboration (EECC) paradigm to achieve collaborative device-scale expansion with real-time access. Although Hierarchical Federated Learning (HFL) supports multitier model aggregation suitable for EECC, prior works assume the same model structure on all computing nodes, constraining the model scale by the weakest end devices. To address this issue, we propose Agglomerative Federated Learning (FedAgg), which is a novel EECC-empowered FL framework that allows the trained models from end, edge, to cloud to grow larger in size and stronger in generalization ability. FedAgg recursively organizes computing nodes among all tiers based on Bridge Sample Based Online Distillation Protocol (BSBODP), which enables every pair of parent-child computing nodes to mutually transfer and distill knowledge extracted from generated bridge samples. This design enhances the performance by exploiting the potential of larger models, with privacy constraints of FL and flexibility requirements of EECC both satisfied. Experiments under various settings demonstrate that FedAgg outperforms state-of-the-art methods by an average of 4.53% accuracy gains and remarkable improvements in convergence rate. Our code is available at https://github.com/wuzhiyuan2000/FedAgg.

Mix networks, or mixnets, are one of the fundamental building blocks of anonymity systems. To defend against epistemic attacks, existing free-route mixnet designs require all clients to maintain a consistent, up-to-date view of the entire key directory. This, however, inevitably raises the performance concern under system scale-out: in a larger mixnet, a client will consume more bandwidth for updating keys in the background.This paper presents Periscoping, a key distribution protocol for mixnets at scale. Periscoping relaxes the download-all requirement for clients. Instead, it allows a client to selectively download a constant number of entries of the key directory, while guaranteeing the privacy of selections. Periscoping achieves this goal via a novel Private Information Retrieval scheme, constructed based on constrained Pseudorandom Functions. Moreover, the protocol is integrated seamlessly into the mixnet operations, readily applicable to existing mixnet systems as an extension at a minimal cost. Our experiments show that, with millions of mixes, it can reduce the traffic load of a mixnet by orders of magnitude, at a minor computational and bandwidth overhead.