INFOCOM 2022 论文列表

The advent of high-accuracy and resource-intensive deep neural networks (DNNs) has fulled the development of live video analytics, where camera videos need to be streamed over the network to edge or cloud servers with sufficient computational resources. Although it is promising to strike a balance between available bandwidth and server-side DNN inference accuracy by adjusting video encoding configurations, the influences of fine-grained network and video content dynamics on configuration performance should be addressed. In this paper, we propose CASVA, a Configuration-Adaptive Streaming framework designed for live Video Analytics. The design of CASVA is motivated by our extensive measurements on how video configuration affects its bandwidth requirement and inference accuracy. To handle the complicated dynamics in live video analytics streaming, CASVA trains a deep reinforcement learning model which does not make any assumptions about the environment but learns to make configuration choices through its experiences. A variety of real-world network traces are used to drive the evaluation of CASVA. The results on a multitude of video types and video analytics tasks show the advantages of CASVA over state-of-the-art solutions.

With the rapid development of mobile devices and the fast increase of sensitive data, secure and convenient mobile authentication technologies are desired. Except for traditional passwords, many mobile devices have biometric-based authentication methods (e.g., fingerprint, voiceprint, and face recognition), but they are vulnerable to spoofing attacks. To solve this problem, we study new biometric features which are based on the dental occlusion and find that the bone-conducted sound of dental occlusion collected in binaural canals contains unique features of individual bones and teeth. Motivated by this, we propose a novel authentication system, TeethPass, which uses earbuds to collect occlusal sounds in binaural canals to achieve authentication. We design an event detection method based on spectrum variance and double thresholds to detect bone-conducted sounds. Then, we analyze the time-frequency domain of the sounds to filter out motion noises and extract unique features of users from three aspects: bone structure, occlusal location, and occlusal sound. Finally, we design an incremental learning-based Siamese network to construct the classifier. Through extensive experiments including 22 participants, the performance of TeethPass in different environments is verified. TeethPass achieves an accuracy of 96.8% and resists nearly 99% of spoofing attacks.

In network measurement, data plane switches measure traffic and report events (e.g., heavy hitters) to the control plane via control channels. The control plane makes decisions to process events. However, current network measurement suffers from two problems. First, when traffic bursts occur, massive events are reported in a short time so that the control channels may be overloaded due to limited bandwidth capacity. Second, only a few events are reported in normal cases, making control channels underloaded and wasting network resources. In this paper, we propose Escala to provide the elastic scaling of control channels at runtime. The key idea is to dynamically migrate event streams among control channels to regulate the loads of these channels. Escala offers two components, including an Escala monitor that detects scaling situations based on realtime network statistics, and an optimization framework that makes scaling decisions to eliminate overload and underload situations. We have implemented a prototype of Escala on Tofino-based switches. Extensive experiments show that Escala achieves timely elastic scaling while preserving high application-level accuracy.

Eye tracking techniques have been widely adopted by a wide range of devices (e.g., AR/VR headsets, smartphones) to enhance user experiences. However, eye gaze data is private in nature, which can reveal users’ psychological and physiological features. Privacy protection techniques can be incorporated to preserve privacy of eye tracking information. Yet, most existing solutions based on Differential Privacy (DP) mechanisms cannot well protect privacy for individual users without sacrificing user experience. In this paper, we are among the first to propose a novel gaze model-based privacy control framework called Otus for eye tracking applications, which incorporates local DP (LDP) mechanisms to preserve user privacy and improves user experience in the meanwhile. First, we conduct a measurement study on real traces to illustrate that direct noise injection on raw gaze trajectories can significantly lower the utility of gaze data. To preserve utility and privacy simultaneously, Otus injects noises in two steps: (1) Extracting model features from raw data to depict gaze trajectories on individual users; (2) Adding LDP noises into model features so as to protect privacy. On one hand, established models can be used to recover user gaze data in order to improve service quality of eye tracking applications. On the other hand, we only need to add LDP noises to distort a small number of model parameters rather than every point on a trajectory to preserve privacy, which has less impact on the utility of gaze data given the same privacy budget. By applying the tile view graph model in step (1), we illustrate the entire workflow of Otus and prove its privacy protection level. For evaluation, we conduct extensive experiments using real gaze traces and the results show that Otus can effectively protect privacy for individual users without significantly compromising gaze data utility.

Network Function Virtualization (NFV) migrates network functions from proprietary hardware to commercial servers on the edge or cloud, making network services more cost-efficient, manage-convenient, and flexible. To facilitate these advantages, it is critical to find an optimal deployment of the chained virtual network functions, i.e. service function chains (SFCs), in hybrid edge-and-cloud environment, considering both resource and latency. It is an NP-hard problem. In this paper, we first limit the problem at the edge and design a constant approximation algorithm named chained next fit (CNF), where a sub-algorithm called double spanning tree (DST) is designed to deal with virtual network embedding. Then we take both cloud and edge resources into consideration and create a promotional algorithm called decreasing sorted, chained next fit (DCNF), which also has a provable constant approximation ratio. The simulation results demonstrate that the ratio between DCNF and the optimal solution is much smaller than the theoretical bound, approaching an average of 1.25. Moreover, DCNF always has a better performance than the benchmarks, which implies that it is a good candidate for joint resource and latency optimization in hybrid edge-and-cloud networks.

The COVID-19 pandemic has affected our lives and how we use network infrastructures in an unprecedented way. While early studies have started shedding light on the link between COVID-19 containment measures and mobile network traffic, we presently lack a clear understanding of the implications of the virus outbreak, and of our reaction to it, on the usage of mobile apps. We contribute to closing this gap, by investigating how the spatiotemporal usage of mobile services has evolved through different response measures enacted in France during a continued seven-month period in 2020 and 2021. Our work complements previous studies in several ways: (i) it delves into individual service dynamics, whereas previous studies have not gone beyond broad service categories; (ii) it encompasses different types of containment strategies, allowing to observe their diverse effects on mobile traffic; (iii) it covers both spatial and temporal behaviors, providing a comprehensive view on the phenomenon. These elements of novelty let us lay new insights on how the demands for hundreds of different mobile services are reacting to the new environment set forth by the pandemics.

Edge computing is a promising solution for reducing service latency by provisioning time-sensitive services directly from the network edge. However, upon workload peaks at the resource-limited edge, an edge service has to queue service requests, incurring high waiting time. Such quality of service (QoS) degradation ruins the reputation and reduces the long-term revenue of the service provider.To address this issue, we propose an admission control mechanism for time-sensitive edge services. Specifically, we allow the service provider to offer admission advice to arriving requests regarding whether to join for service or balk to seek alternatives. Our goal is twofold: maximizing revenue of the service provider and ensuring QoS if the provided admission advice is followed. To this end, we propose a threshold structure that estimates the highest length of the request queue. Leveraging such a threshold structure, we propose O2A, a mechanism to balance the trade-off between increasing revenue from accepting more requests and guaranteeing QoS by advising requests to balk. Rigorous analysis shows that O2A achieves the goal and that the provided admission advice is optimal for end-users to follow. We further validate O2A through trace-driven simulations with both synthetic and real-world service request traces.

The ability to dynamically scale in/out network functions (NFs) on multiple cores/servers to meet traffic demands is a key benefit of network function virtualization (NFV). The stateful NF operations make NFV scaling a challenging task: if care is not taken, NFV scaling can lead to incorrect operations and poor performance. We advocate two general abstractions, NFlow and Match-Value Table (MVT), for NFV packet processing pipelines. We present formal definitions of the abstractions and discuss how they can facilitate NFV scaling by minimizing or eliminating shared states. Using NFs implemented with the proposed abstractions, we conduct extensive experiments and demonstrate their efficacy in terms of correctness and performance of NFV scaling.

The Priority-based Flow Control (PFC) protocol is adopted to guarantee zero packet loss in many high-performance data centers. PFC, however, can induce deadlocks and in severe cases cause the entire network to be blocked. Existing solutions have focused on deadlock avoidance; unfortunately, they are not foolproof. Therefore, deadlock detection is a necessity. We propose ITSY, a novel system that correctly detects and resolves deadlocks entirely in the data plane. It works with any network topologies and routing algorithms. Unique to ITSY is the use of deadlock initial triggers, which contributes to efficient deadlock detection, mitigation, and recurrence prevention. ITSY provides three deadlock resolution mechanisms with different trade-off options. We implement ITSY for programmable switches in the P4 language. Experiments show that ITSY detects and resolves deadlocks rapidly with minimal overheads.

The fast development of mobile edge computing (MEC) and service virtualization brings new opportunities to the deployment of interactive applications, e.g., VR education, stream gaming, autopilot assistance, at the network edge for better performance. Ensuring quality of experience (QoE) for such services often requires the satisfaction of multiple quality of service (QoS) factors, e.g., short delay, high throughput rate, low packet loss. Nevertheless, existing mobile edge networks often fail to meet these requirements due to the mobility of end users and the volatility of network conditions. In this paper, we propose a novel scheme that both reduces delay and adjusts data throughput rate for QoE enhancement. We design an online service placement and throughput rate adjustment (SPTA) algorithm which coordinately migrates virtual services while tuning their data throughput rates based on real-time bandwidth fluctuation. By implementing a small-scale prototype supporting stream gaming at the edge, we show the necessity and feasibility of our work. Based on data from the experiments, we conduct real-world trace driven simulations to further demonstrate the advantages of our scheme over existing baselines.

The network topology design is critical for achieving low latency and high capacity in future integrated satellite and terrestrial networks (ISTN). However, existing studies mainly focus on the design of inter-satellite topology of ISTN, and very little is known about the design of satellite-ground topology, as well as its impact on the attainable network performance.In this paper, we conduct a quantitative study on the impact of various satellite-ground designs on the network performance of ISTN. We identify that the high-density and high-dynamicity characteristics of emerging mega-constellations have jointly imposed big challenges, such as significant routing instability, low network reachability, high latency and jitter on the ISTN paths. To alleviate the above challenges, we formulate the Low-latency Satellite-Ground Interconnecting (LSGI) problem, targeting at the integration of space and ground segment in the ISTN, while minimizing the maximum transmission latency and keeping routing stable. We further design algorithms to solve the LSGI problem through wisely coordinating the establishment of ground-to-satellite links among distributed ground stations. Comprehensive experiment results demonstrate that our solution can outperform existing related schemes by about 19% reduction of the latency and 70% reduction of the jitter on average, while sustaining the highest network reachability.

Voice assistant has been widely used for human-computer interaction and automatic meeting minutes. However, for multiple sound sources, the performance of speech recognition in voice assistant decreases dramatically. Therefore, it is crucial to separate multiple voices efficiently for an effective voice assistant application in multi-user scenarios. In this paper, we present a novel voice separation system using a 2D microphone array in multiple sound source scenarios. Specifically, we propose a spatial filtering-based method to iteratively estimate the Angle of Arrival (AoA) of each sound source and separate the voice signals with adaptive beamforming. We use BeamForming-based cross-Correlation (BF-Correlation) to accurately assess the performance of beamforming and automatically optimize the voice separation in the iterative framework. Different from cross-correlation, BF-Correlation further performs cross-correlation among the after-beamforming voice signals processed with each linear microphone array. In this way, the mutual interference from voice signals out of the specified direction can be effectively suppressed or mitigated via the spatial filtering technique. We implement a prototype system and evaluate its performance in real environments. Experimental results show that the average AoA error is 1.4 degree and the average ratio of automatic speech recognition accuracy is 90.2% in the presence of three sound sources.

Accurate estimation of Channel State Information (CSI) is essential to design MU-MIMO beamforming. However, errors in CSI estimation are inevitable in practice. State-of-the-art works model CSI as random variables and assume certain specific distributions or worst-case boundaries, both of which suffer performance issues when providing performance guarantees to the users. In contrast, this paper proposes a Data-Driven Beamforming (D2BF) that directly handles the available CSI data samples (without assuming any particular distributions). Specifically, we employ chance-constrained programming (CCP) to provide probabilistic data rate guarantees to the users and introduce ∞-Wasserstein ambiguity set to bridge the unknown CSI distribution with the available (limited) data samples. Through problem decomposition and a novel bilevel formulation for each subproblem, we show that each subproblem can be solved by binary search and convex approximation. We also validate that D2BF offers better performance than the state-of-the-art approach while meeting probabilistic data rate guarantees to the users.

Various congestion control (CC) algorithms have been designed to target specific scenarios. To automate this process, researchers have begun to use machine learning to automatically control the congestion window. These, however, often rely on heavyweight learning models (e.g., neural networks). This can make them unsuitable for resource-constrained mobile devices. On the other hand, lightweight models (e.g., decision trees) are often incapable of reflecting the complexity of diverse mobile wireless environments. To address this, we present Muses, a learning-based approach for generating lightweight congestion control algorithms. Muses relies on imitation learning to train a universal (heavy) LSTM model, which is then used to extract (lightweight) decision tree models that are each targeted at an individual environment. Muses then dynamically selects the most appropriate decision tree on a per-flow basis. We show that Muses can generate high throughput policies across a diverse set of environments, and it is sufficiently light to operate on mobile devices.

In this paper, we discuss the concept of Midpoint Optimization (MO) for Segment Routing (SR). It is based on the idea of integrating SR policies into the Interior Gateway Protocol (IGP) to allow various demands to be steered into them. We discuss the benefits of this approach when compared to end-to-end SR and potential challenges that might arise in deployment. We further develop an LP-based optimization algorithm to assess the Traffic Engineering capabilities of MO for SR. Based on traffic and topology data from a Tier-1 Internet Service Provider as well as other, publicly available data, we show that this algorithm is able to achieve virtually optimal results with regards to the maximum link utilization, that are on par with state-of-the-art end-to-end SR approaches. However, our MO approach requires substantially less policies to do so. For some instances the achieved reduction ranges up to more than 99%.

A large amount of disinformation on social media has penetrated into various domains and brought significant adverse effects. Understanding their roots and propagation becomes desired in both academia and industry. Prior literature has developed many algorithms to identify this disinformation, particularly rumor detection. Some leverage the power of deep learning and have achieved promising results. However, they all focused on building predictive models and improving forecast accuracy, while two important factors - popularity and conformity biases - that play critical roles in rumor spreading behaviors are usually neglected.To overcome such an issue and alleviate the bias from these two factors, we propose a rumor detection framework to learn debiased user preference and effective event representation in a causal view. We first build a graph to capture causal relationships among users, events, and their interactions. Then we apply the causal intervention to eliminate popularity and conformity biases and obtain debiased user preference representation. Finally, we leverage the power of graph neural networks to aggregate learned user representation and event features for the final event type classification. Empirical experiments conducted on two real-world datasets demonstrate the effectiveness of our proposed approach compared to several cutting-edge baselines.

Cloud Radio Access Network (C-RAN) is a novel centralized architecture for cellular networks. C-RAN can significantly improve spectrum efficiency by performing cooperative signal processing for multiple cells at a centralized baseband unit (BBU) pool. However, a new resource scheduler is needed before we can take advantage of C-RAN's multi-cell processing capability. Under C-RAN architecture, the scheduler must jointly determine RB allocation, MCS assignment, and beamforming matrices for all users under all covering cells. In addition, it is necessary to obtain a scheduling solution within each TTI (at most 1 ms) to be useful for the frame structure defined by 5G NR. In this paper, we present M3—a sub-ms scheduler for multi-cell MIMO networks under C-RAN architecture. M3addresses the stringent timing requirement through a novel multi-pipeline design that exploits parallelism. Under this design, one pipeline performs a sequence of operations for cell-edge users to explore joint transmission, and in parallel, the other pipeline is for cell-center users to explore MU-MIMO transmission. Experimental results show that M3is capable of offering a scheduling solution within 1 ms for 7 remote radio heads (RRHs), 100 users, 100 RBs, and 2×12 MIMO. Meanwhile, M3provides ~40%. throughput gain on average by employing joint transmission.

Consider a stream learning system with a source and a set of computation nodes that solves a machine learning task modeled as stochastic convex optimization problem over an unknown distribution D. The source generates i.i.d. data points from D and routes the data points to the computation nodes for processing. The data points are processed in a streaming fashion, i.e., each data point can be accessed only once and is discarded after processing. The system employs local stochastic gradient descent (local SGD), where each computation node performs stochastic gradient descent locally using the data it receives from the source and periodically synchronizes with other computation nodes. Since the routing policy of the source determines the availability of data points at each computation node, the performance of the system, i.e., the optimization error obtained by local SGD, depends on the routing policy.In this paper, we study the influence of the routing policy on the performance of stream learning systems. We first derive an upper bound on the optimization error as a function of the routing policy. The upper bound reveals that the routing policy influences the performance through tuning the bias-variance trade-off of the optimization process, and gives rise to a framework for optimizing the routing policy for stream learning systems. By minimizing the upper bound, we propose an optimal static routing policy that achieves the best trade-off for stream learning systems with deterministic data generation process. We then propose a routing policy that can approximate the optimal static routing policy arbitrarily closely for systems where the data points are generated according to a stochastic process with unknown rate. Finally, we conduct simulations using Support Vector Machine as the machine learning task on a real data set, and show that the optimal static routing policy has excellent empirical performance in terms of minimizing the optimization error and the proposed stochastic routing policy closely matches the optimal static routing policy.

As a cost-effective paradigm, Sparse Crowdsensing aims to recruit workers to perform a part of sensing tasks and infer the rest, which has broad applications, including environmental monitoring, urban sensing, etc. In most cases, workers will participate in real time, and thus their sensing data are coming dynamically. Taking full advantage of the online coming data to complete the full sensing map is an important problem for Sparse Crowdsensing. However, for data completion, the importance of data collected from different spatio-temporal areas is usually different and time-varying. For example, the newly obtained data in the center is often more important than the old ones from edges. Moreover, the area importance may also influence the following worker selection, i.e., selecting suitable workers to actively sense important areas (instead of passively waiting for given data) for improving completion accuracy. To this end, in this paper, we propose a framework for online Sparse Crowdsensing, called OS-MCS, which consists of three parts: matrix completion, importance estimation, and worker selection. We start from the dynamically coming data and propose an online matrix completion algorithm with spatio-temporal constraints. Based on that, we estimate the spatio-temporal area importance by conducting a reinforcement learning-based up-to-date model. Finally, we utilize the prophet secretary problem to select suitable workers to sense important areas for accurate completion in an online manner. Extensive experiments on real-world data sets show the effectiveness of our proposals.