INFOCOM 2024 Papers

As live streaming services skyrocket, Crowdsourced Cloud-edge service Platforms (CCPs) have surfaced as pivotal intermediaries catering to the mounting demand. Despite the role of stream scheduling to CCPs’ Quality of Service (QoS) and throughput, conventional optimization strategies struggle to enhancing CCPs’ revenue, primarily due to the intricate relationship between resource utilization and revenue. Additionally, the substantial scale of CCPs magnifies the difficulties of time-intensive scheduling. To tackle these challenges, we propose Seer, a proactive revenue-aware scheduling system for live streaming services in CCPs. The design of Seer is motivated by meticulous measurements of real-world CCPs environments, which allows us to achieve accurate revenue modeling and overcome three key obstacles that hinder the integration of prediction and optimal scheduling. Utilizing an innovative Preschedule-Execute-Re-schedule paradigm and flexible scheduling modes, Seer achieves efficient revenue-optimized scheduling in CCPs. Extensive evaluations demonstrate Seer’s superiority over competitors in terms of revenue, utilization, and anomaly penalty mitigation, boosting CCPs revenue by 147% and expediting scheduling 3.4× faster.

To meet stringent performance requirements, communication networks are becoming increasingly programmable and flexible, supporting fast and frequent adjustments. However, reconfiguring networks in a dependable and transiently consistent manner is known to be algorithmically challenging. This paper revisits the fundamental problem of how to update the routes in a network in a (transiently) loop-free manner, considering both the Strong Loop-Freedom (SLF) and the Relaxed Loop-Freedom (RLF) property.We present two fast algorithms to solve the SLF and RLF problem variants exactly, to optimality. Our algorithms are based on a parameterized integer linear program which would be intractable to solve directly by a classic solver. Our main technical contribution is a lazy cycle breaking strategy which, by adding constraints lazily, improves performance dramatically, and outperforms the state-of-the-art exact algorithms by an order of magnitude on realistic medium-sized networks. We further explore approximate algorithms and show that while a relaxation approach is relatively slow, with a local search approach short update schedules can be found, outperforming the state-of-the-art heuristics.On the theoretical front, we also provide an approximation lower bound for the update time of the state-of-the-art algorithm in the literature. As a contribution to the research community, we made all our code and implementations publicly available.

Federated learning (FL) enables distributed training via periodically synchronizing model updates among participants. Communication overhead becomes a dominant constraint of FL since participating clients usually suffer from limited bandwidth. To tackle this issue, top-k based gradient compression techniques are broadly explored in FL context, manifesting powerful capabilities in reducing gradient volumes via picking significant entries. However, previous studies are primarily conducted on the raw gradients where massive spatial redundancies exist and positions of non-zero (top-k) entries vary greatly between gradients, which both impede the achievement of deeper compressions. Top-k may also degrade the performance of trained models due to biased gradient estimations. Targeting the above issues, we propose FedTC, a novel transform coding based compression framework. FedTC transforms gradients into a new domain with more compact energy distributions, which facilitates reducing spatial redundancies and biases in subsequent sparsification. Furthermore, non-zero entries across clients from different rounds become highly aligned in the transform domain, motivating us to partition the gradients into smaller entry blocks with various alignment levels to better exploit these alignments. Lastly, positions and values of non-zero entries are independently compressed in a block-wise manner with our customized designs, through which a higher compression ratio is achieved. Theoretical analysis and extensive experiments consistently demonstrate the effectiveness of our approach.

Cross-shard Transactions (CTXs) widely exist in sharded blockchains. CTXs have to endure large confirmation latency because they need to participate in consensus in both their source and destination shards. To diminish CTXs, plenty of state-of-the-art blockchain protocols have been proposed. For example, in BrokerChain [1], some intermediary broker accounts can help turn CTXs into intra-shard transactions through their voluntary liquidity services. Thereby, the original CTXs can be confirmed in blockchain shards quickly. However, we found that BrokerChain is impractical for a sharded blockchain because it does not consider how to recruit a sufficient number of broker accounts. Thus, blockchain clients do not have the motivation to provide token liquidity for others. To address this challenge, we design Broker2Earn, which is essentially a decentralized finance (DeFi) protocol that works as an incentive mechanism for blockchain users who choose to become brokers. Via participating in Broker2Earn, brokers can earn native revenues when they collateralize their tokens to the protocol. Furthermore, Broker2Earn can also benefit the sharded blockchain since it can efficiently spend each staked liquidity provided by brokers on diminishing CTXs. We formulate the core module of Broker2Earn into a revenue-maximization problem, which is proven NP-hard. To solve this problem, we design an online approximation algorithm using the relax-and-rounding technique. We also rigorously analyze the approximation ratio of our online algorithm. Finally, we conduct extensive experiments using real-world Ethereum transactions on both a transaction-driven simulator and an open-source blockchain testbed. The evaluation results show that the proposed Broker2Earn protocol demonstrates a near-optimal performance that outperforms other baselines, in terms of broker revenues and the usage of system liquidity.

This paper presents COPILOT, a ML-based approach that allows vehicles requiring ubiquitous high bandwidth connectivity to identify the most suitable road side units (RSUs) through proactive handoffs. By cooperatively exchanging the data obtained from local 3D Lidar point clouds within adjacent vehicles and with coarse knowledge of their relative positions, COPILOT identifies transient blockages to all candidate RSUs along the path under study. Such cooperative perception is critical for choosing RSUs with highly directional links required for mmWave bands, which majorly degrade in the absence of LOS. COPILOT proposes three modules that operate in an inter-connected manner: (i) As an alternative to sending raw Lidar point clouds, it extracts and transmits low-dimensional intermediate features to lower the overhead of inter-vehicle messaging; (ii) It utilizes an attention-mechanism to place greater emphasis on data collected from specific vehicles, as opposed to nearest neighbor and distance-based selection schemes, and (iii) it experimentally validates the outcomes using an outdoor testbed composed of an autonomous car and Talon AD7200 60GHz routers emulating the RSUs, accompanied by the public release of the datasets. Results reveal COPILOT yields upto 69.8% and 20.42% improvement in latency and throughput compared to traditional reactive handoffs for mmWave networks, respectively.

Graph convolutional network (GCN) has achieved enormous success in learning structural information from unstructured data. As graphs become increasingly large, distributed training for GCNs is severely prolonged by frequent cross-worker communications. Existing efforts to improve the training efficiency often come at the expense of GCN performance, while the communication overhead persists. In this paper, we propose PSC-GCN, a holistic pipelined framework for distributed GCN training with communication-efficient sampling and inclusion-aware caching, to address the communication bottleneck while ensuring satisfactory model performance. Specifically, we devise an asynchronous pre-fetching scheme to retrieve stale statistics (features, embedding, gradient) of boundary nodes in advance, such that the embedding aggregation and model update are pipelined with statistics transmission. To alleviate communication volume and staleness effect, we introduce a variance-reduction based sampling policy, which prioritizes inner nodes over boundary ones for reducing the access frequency to remote neighbors, thus mitigating cross-worker statistics exchange. Complementing graph sampling, a feature caching module is co-designed to buffer hot nodes with high inclusion probability, ensuring that frequently sampled nodes will be available in local memory. Extensive evaluations on real-world datasets show the superiority of PSC-GCN over state-of-the-art methods, where we can reduce training time by 72%-80% without sacrificing model accuracy.

All-optical circuit-switching (OCS) technology is the key to design energy-efficient and high-performance datacenter network (DCN) architectures for the future. However, existing round-robin based OCS cores perform poorly under realistic workloads having high traffic skewness and high volume of inter-rack traffic. To address this issue, we propose a novel DCN architecture OSSV: a combination of OCS-based core (between ToR switches) and OCS-based reconfigurable edge (between servers and ToR switches). On one hand, the OCS core is traffic agnostic and realizes reconfigurably non-blocking ToR-level connectivity. On the other hand, OCS-based edge reconfigures itself to reshape the incoming traffic in order to jointly minimize traffic skewness and inter-rack traffic volume. Our novel optimization framework can obtain the right balance between these intertwined objectives. Our extensive simulations and testbed evaluation show that OSSV can achieve high performance under diverse DCN traffic while consuming low power and incurring low cost.

To maximize Signal-to-Noise Ratio (SNR), it is necessary to move beyond selecting beams from a codebook. While the state-of-the-art approaches can significantly improve SNR compared to codebook-based beam selection by exploiting the globally-optimal beam, they incur significant beam-training overhead, which limits the applicability to large-scale antenna arrays and the scalability for multiple users. In this paper, we propose FTP, a highly-scalable beam-training solution that can find the globally-optimal beam with minimal beam-training overhead. FTP works by estimating per-path direction along with its complex gain and synthesizes the globally-optimal beam from these parameters. Our design significantly reduces the search space for finding such path parameters, which enables FTP to scale to large-scale antenna arrays. We implemented and evaluated FTP on a mmWave experimental platform with 32 antenna elements. Our results demonstrate that FTP achieves optimal SNR performance comparable with the state-of-the-art while reducing the beam-training overhead by 3 orders of magnitude. Under simulated settings, we demonstrate that the gain of FTP can be even more significant for larger antenna arrays with up to 1024 elements.

Embedding machine learning (ML) models in programmable switches realizes the vision of high-throughput and low-latency inference at line rate. Recent works have made breakthroughs in embedding Random Forest (RF) models in switches for either packet-level inference or flow-level inference. The former relies on simple features from packet headers that are simple to implement but limit accuracy in challenging use cases; the latter exploits richer flow features to improve accuracy, but leaves early packets in each flow unclassified. We propose Jewel, an in-switch ML model based on a fully joint packet-and flow-level design, which takes the best of both worlds by classifying early flow packets individually and shifting to flow-level inference when possible. Our proposal involves (i) a single RF model trained to classify both packets and flows, and (ii) hardware-aware model selection and training techniques for resource footprint minimization. We implement Jewel in P4 and deploy it in a testbed with Intel Tofino switches, where we run extensive experiments with a variety of real-world use cases. Results reveal how our solution outperforms four state-of-the-art benchmarks, with accuracy gains in the 2.0%–5.3% range.

Collaborative Edge Computing (CEC) is a new edge computing paradigm that enables neighboring edge servers to share computational resources with each other. Although CEC can enhance the utilization of computational resources, it still suffers from resource waste. The primary reason is that end-users from the same area are likely to offload similar tasks to edge servers, thereby leading to duplicate computations. To improve system efficiency, the computation results of previously executed tasks can be cached and then reused by subsequent tasks. However, most existing computation reuse algorithms only consider one edge server, which significantly limits the effectiveness of computation reuse. To address this issue, this paper applies computation reuse in CEC networks to exploit the collaboration among edge servers. We formulate an optimization problem that aims to minimize the overall task response time and decompose it into a caching subproblem and a scheduling subproblem. By analyzing the properties of optimal solutions, we show that the optimal caching decisions can be efficiently searched using the bisection method. For the scheduling subproblem, we utilize projected gradient descent and backtracking to find a local minimum. Numerical results show that our algorithm significantly reduces the response time in various situations.

Efficient entanglement routing in Quantum Data Networks (QDNs) is essential in order to concurrently establish as many Entanglement Connections (ECs) as possible, which in turn maximizes the network throughput. In this work, we consider a new class of QDNs with wavelength division multiplexed (WDM) quantum links where each quantum repeater will perform entanglement swapping by measuring two photonic qubits coming from some entangled photon sources directly on the same wavelength. To address unique challenges in achieving a high network throughput in such QDNs, we propose QuRWA to jointly optimize the entanglement routing and wavelength assignment. To this end, we introduce a key concept named Co-Path to improve fault-tolerance: all ELs in a Co-Path set will be assigned the same wavelength and this may serve as backup for some other ELs in the same Co-Path when establishing ECs. We design efficient algorithms to optimize the Co-Path selection and wavelength assignment to maximize resource utilization and fault tolerance. Extensive simulations demonstrate that compared with the methods without introducing Co-Path, QuRWA improves the network throughput by up to 122%.

Decentralized Storage Network (DSN) is an emerging technology that challenges traditional cloud-based storage systems by consolidating storage capacities from independent providers and coordinating to provide decentralized storage and retrieval services. However, current DSNs face several challenges associated with data privacy and efficiency of the proof systems. To address these issues, we propose FileDES ( Decentralized Encrypted Storage), which incorporates three essential elements: privacy preservation, scalable storage proof, and batch verification. FileDES provides encrypted data storage while maintaining data availability, with a scalable Proof of Encrypted Storage (PoES) algorithm that is resilient to Sybil and Generation attacks. Additionally, we introduce a rollup-based batch verification approach to simultaneously verify multiple files using publicly verifiable succinct proofs. We conducted a comparative evaluation on FileDES, Filecoin, Storj and Sia under various conditions, including a WAN composed of up to 120 geographically dispersed nodes. Our protocol outperforms the others in terms of proof generation/verification efficiency, storage costs, and scalability.

The domain name system (DNS) is indispensable to nearly every Internet service. It has been extensively utilized for network activity characterization in passive and active approaches. Compared to the passive approach, active DNS cache probing is privacy-preserving and low-cost, enabling worldwide characterization of remote network activities in different networks. Unfortunately, existing probing-based methods are too coarse-grained to characterize the time-varying features of network activities, substantially limiting their applications in time-sensitive tasks. In this paper, we advance DNSScope, a fine-grained DNS cache probing framework by tackling three challenges: sample sparsity, observational distortion, and cache entanglement. DNSScope synthesizes statistical learning and self-supervised transfer learning to achieve time-varying characterization. Extensive evaluations demonstrate that it can accurately estimate the time-varying DNS query arrival rates on recursive DNS resolvers. Its average mean absolute error is 0.124, as low as one-sixth that of the baseline methods.

The quality of federated learning (FL) is highly correlated with the number and quality of the participants involved. It is essential to design proper contribution evaluation mechanisms. Shapley Value (SV)-based techniques have been widely used to provide fair contribution evaluation. Existing approaches, however, do not support dynamic participants (e.g., joining and departure) and incur significant computation costs, making them difficult to apply in practice. Worse, participants may be incorrectly valued as negative contribution under the Non-IID data scenarios, further jeopardizing fairness. In this work, we propose FairFed to address the above challenges. First, given that each iteration is of equal importance, FairFed treats FL as Multiple Single-stage Cooperative Games, and evaluates participants by each iteration for effectively coping with dynamic participants and ensuring fairness across iterations. Second, we introduce Cooperative Shapley Value (CSV) to rectify negative values of participants to improving the fairness while preserving true negative values. Third, we prove if participants are Strategically Equivalent, the number of participant combinations can be sharply reduced from exponential to polynomial, thus significantly reducing the computational complexity of CSV. Experimental results show that FairFed achieves up to 25.3 × speedup and reduces deviations by three orders of magnitude to two state-of-the-art approximation approaches.

Drones have demonstrated their pivotal role in various applications such as search-and-rescue, smart logistics, and industrial inspection, with accurate localization playing an indispensable part. However, in high dynamic range and rapid motion scenarios, traditional visual sensors often face challenges in pose estimation. Event cameras, with their high temporal resolution, present a fresh opportunity for perception in such challenging environments. Current efforts resort to event-visual fusion to enhance the drone’s sensing capability. Yet, the lack of efficient event-visual fusion algorithms and corresponding acceleration hardware causes the potential of event cameras to remain underutilized. In this paper, we introduce EventBoost, an acceleration platform designed for drone-based applications with event-image fusion. We propose a suit of novel algorithms through software-hardware co-design on Zynq SoC, aimed at enhancing real-time localization precision and speed. EventBoost achieves enhanced visual fusion precision and markedly elevated processing efficiency. The performance comparison with two state-of-the-art systems shows EventBoost achieves 24.33% improvement in accuracy with 30 ms latency on resource-constrained platforms.

Deep neural network (DNN) based website fingerprinting (WF) attacks can achieve an attack success rate (ASR) of over 90%, seriously threatening the privacy of Tor users. At present, adversarial example (AE) based defenses have demonstrated great potential to defend against WF attacks. However, existing AE-based defenses require knowing a complete traffic trace for adversarial perturbation calculation, which is unrealistic in practice. Moreover, they may become ineffective once adversarial training (AT) is adopted by attackers. To mitigate these two problems, we propose a defense called ALERT. It generates adversarial perturbations without knowing traffic traces, and can effectively resist AT-aided WF attacks. The key idea of ALERT is to produce universal perturbations that vary from user to user. We conduct extensive experiments to evaluate ALERT. In the closed world, ALERT significantly surpasses four representative WF defenses, including the state-of-the-art (SOTA) defense AWA. Specifically, ALERT reduces the ASR of the SOTA DF attack to 12.68% and uses only 20.13% of communication bandwidth. In the open world, ALERT uses only 19.91% of bandwidth, reduces the True Positive Rate (TPR) of the DF attack to 37.46%, obviously outperforming the other defenses.

The Domain Name System (DNS) is a critical piece of Internet infrastructure with remarkably complex properties and uses, and accordingly has been extensively studied. In this study we contribute to that body of work by organizing and analyzing records maintained within the DNS as a bipartite graph. We find that relating names and addresses in this way uncovers a surprisingly rich structure. In order to characterize that structure, we introduce a new graph decomposition for DNS name-to-IP mappings, which we term elemental decomposition. In particular, we argue that (approximately) decomposing this graph into bicliques — maximally connected components — exposes this rich structure. We utilize large-scale censuses of the DNS to investigate the characteristics of the resulting decomposition, and illustrate how the exposed structure sheds new light on a number of questions about how the DNS is used in practice and suggests several new directions for future research.

Network intrusion detection (NID) aims to identify unusual network traffic patterns (distribution shifts) that require NID systems to evolve continuously. While prior art emphasizes fully supervised annotated data-intensive continual learning methods for NID, semi-supervised continual learning (SSCL) methods require only limited annotated data. However, the inherent class imbalance (CI) in network traffic can significantly impact the performance of SSCL approaches. Previous approaches to tackle CI issues require storing a subset of labeled training samples from all past tasks in the memory for an extended duration, potentially raising privacy concerns. The proposed Semisupervised Privacy-preserving Intrusion detection with Drift-aware continual LEaRning (SPIDER) is a novel method that combines gradient projection memory (GPM) with SSCL to handle CI effectively without the requirement to store labeled samples from all of the previous tasks. We assess SPIDER’s performance against baselines on six intrusion detection benchmarks formed over a short period and the Anoshift benchmark spanning ten years, which includes natural distribution shifts. Additionally, we validate our approach on standard continual learning image classification benchmarks known for frequent distribution shifts compared to NID benchmarks. SPIDER achieves comparable performance to fully supervised and semisupervised baseline methods, while requiring a maximum of 20% annotated data and reducing the total training time by 2X.

The online linear optimization paradigm is important to many real-world network applications as well as theoretical algorithmic studies. Recent studies have made attempts to augment online linear optimization with machine-learned predictions of the cost function that are meant to improve the performance of the algorithms. However, they fail to address the critical case in practical systems where the predictions can be arbitrarily bad. In this work, we take the first step to study the problem of online linear optimization with a dynamic number of arbitrarily bad machine-learned predictions per round and propose an algorithm termed OLOAP. Our theoretical analysis shows that, when the qualities of the predictions are satisfactory, OLOAP achieves a regret bound of O(logT), which circumvents the tight lower bound of Ω($\sqrt T $) for the vanilla problem of online linear optimization (i.e., the one without any predictions). Meanwhile, the regret of our algorithm is never worse than O($\sqrt T $) irrespective of the qualities of predictions. In addition, we further derive a lower bound for the regret of the studied problem, which demonstrates that OLOAP is near-optimal. We consider two important network applications and conduct extensive evaluations. Our results validate the superiority of our algorithm over state-of-the-art approaches.