INFOCOM 2024 论文列表

Existing Deep Learning (DL)-based network Intrusion Detection System (IDS) is able to characterize sequence semantics of traffic and discover malicious behaviors. Yet DL models are often nonlinear and highly non-convex functions that are difficult for in-network deployment. In this paper, we present RIDS, a hardware-friendly Recurrent Neural Network (RNN) model that is co-designed with programmable switches. As its core, RIDS is powered by two tightly-coupled components: (i) rLearner, the RNN learning module with in-network deployability as the first-class requirement; and (ii) rEnforcer, the concrete pipeline design to realize rLearner-generated models inside the network dataplane. We implement a prototype of RIDS and evaluate it on our physical testbed. The experiments show that RIDS could satisfy both detection performance and high-speed bandwidth adaptation simultaneously, when none of the other existing approaches could do so. Inspiringly, RIDS realizes remarkable intrusion/malware detection effect (e.g., ~99% F1 score) and model deployment (e.g., 100 Gbps per port), while only imposing nanoseconds of latency.

We study a scheduling problem for a base-station transmitting status information to multiple user-equipments (UE) with the goal of minimizing the total expected Age-of-Information (AoI). Such a problem can be formulated as a Restless Multi-Armed Bandit (RMAB) problem and solved asymptotically-optimally by a low-complexity Whittle index policy, if each UE’s sub-problem is Whittle indexable. However, proving Whittle indexability can be highly non-trivial, especially when the value function cannot be derived in closed-form. In particular, this is the case for the AoI minimization problem with stochastic arrivals and unreliable channels, whose Whittle indexability remains an open problem. To overcome this difficulty, we develop a sufficient condition for Whittle indexability based on the notion of active time (AT). Even though the AT condition shares considerable similarity to the Partial Conservation Law (PCL) condition, it is much easier to understand and verify. We then apply our AT condition to the stochastic-arrival unreliable-channel AoI minimization problem and, for the first time in the literature, prove its Whittle indexability. Our proof uses a novel coupling approach to verify the AT condition, which may also be of independent interest to other large-scale RMAB problems.

Network fluctuations can cause unpredictable degradation of the user’s quality of experience (QoE) on real-time video streaming. The intrinsic property of real-time video streaming, which generates delay-sensitive and chunk-based video frames, makes the situation even more complicated. Although previous approaches have tried to alleviate this problem by controlling the video bitrate based on the current network capacity estimate, they do not take into account the explicit queueing delay experienced by the video frame in determining the bitrate of upcoming video frames. To tackle this problem, we propose a new real-time video streaming system, Exstream, that can adapt to dynamic network conditions with the help of video bitrate control method and bandwidth estimation method designed to support real-time video streaming environments. Exstream explicitly estimates the queueing delay experienced by the video frame based on the transmission time budget that each frame can maximally utilize, which depends on the frame generation interval, and adjusts the bitrate of newly generated video frames to suppress the queueing delay level close to zero. Our comprehensive experiments demonstrate that Exstream achieves lower frame delay than four existing systems, Salsify, WebRTC, Skype, and Hangouts without frequent video frame skip.

The multi-armed bandits (MAB) framework is widely used for sequential decision-making under uncertainty, finding applications in various domains, including computer and communication networks. To address the increasing complexity of real-world systems and their operational requirements, researchers have proposed and studied various extensions to the basic MAB framework. In this paper, we focus on an adversarial MAB problem inspired by real-world systems with combinatorial semi-bandit arms, switching costs, and anytime cumulative arm selection constraints. To tackle this challenging problem, we introduce the Block-structured Follow-the-Regularized-Leader (B-FTRL) algorithm. Our approach employs a hybrid Tsallis-Shannon entropy regularizer in arm selection and incorporates a block structure that divides time into blocks to minimize arm switching costs. The theoretical analysis shows that B-FTRL achieves a reward regret bound of $O\left( {{T^{\frac{{2a - b + 1}}{{1 + a}}}} + {T^{\frac{b}{{1 + a}}}}} \right)$ and a switching regret bound of $O\left( {{T^{\frac{1}{{1 + a}}}}} \right)$, where a and b are tunable algorithm parameters. By carefully selecting the values of a and b, we are able to limit the total regret to O(T2/3) while satisfying the arm selection constraints in expectation. This outperforms the state-of-the-art regret bound of O(T3/4) and expected constraint violation bound o(1), which are derived in less challenging stochastic reward environments. Additionally, we provide a high-probability constraint violation bound of $O(\sqrt T )$. To validate the effectiveness of the proposed BFTRL algorithm, numerical results are presented to demonstrate its superiority in comparison to other existing methods.

Caching can be leveraged to significantly improve network performance and mitigate congestion. However, characterizing the optimal tradeoff between routing cost and cache deployment cost remains an open problem. In this paper, for a network with arbitrary topology and congestion-dependent nonlinear cost functions, we aim to jointly determine the cache deployment, content placement, and hop-by-hop routing strategies, so that the sum of routing cost and cache deployment cost is minimized. We tackle this mixed-integer nonlinear problem starting with a fixed-routing setting, and then generalize to a dynamic-routing setting. For the fixed-routing setting, a Gradient-combining Frank-Wolfe algorithm with $\left( {\frac{1}{2},1} \right)$-approximation is presented. For the general dynamic-routing setting, we obtain a set of KKT conditions, and devise a distributed and adaptive online algorithm based on these conditions. We demonstrate via extensive simulation that our algorithms significantly outperform a number of baselines.

With the recent surge of research interests in Large Language Models (LLMs), a natural question that arises is how pre-trained LLMs can be fine-tuned to tailor to specific needs of enterprises and individual users, while preserving the privacy of data used in the fine-tuning process. On the one hand, sending private data to cloud datacenters for fine-tuning is, without a doubt, unacceptable from a privacy perspective. On the other hand, conventional federated learning requires each client to perform local training, which is not feasible for LLMs with respect to both computation costs and communication overhead, since they involve billions of model parameters. In this paper, we present Titanic, a new distributed training paradigm that allows LLMs to be fine-tuned in a privacy-preserving fashion directly on the client devices where private data is produced, while operating within the resource constraints on computation and communication bandwidth. Titanic first optimally selects a subset of clients with an efficient solution to an integer optimization problem, then partitions an LLM across multiple client devices, and finally fine-tunes the model with no or minimal losses in training performance. A primary focus in the design of Titanic is its feasibility in real-world systems: it is first and foremost designed for production-quality systems, featuring a model-agnostic partitioning mechanism that is fully automated. Our experimental results show that Titanic achieves superior training performance as compared to conventional federated learning, while preserving data privacy and satisfying all constraints on local computation and bandwidth resources.

The centralization of web services has raised concerns about critical single points of failure, such as content hosting, name resolution, and certification. To address these issues, the "Decentralized Web" movement advocates for de-centralized alternatives. Distributed Hash Tables (DHTs) have emerged as a key component facilitating this movement, as they offer efficient key/value indexing. The InterPlanetary File System (IPFS) exemplifies this approach by leveraging DHTs for data indexing and distribution. A critical finding of previous studies is that DHT PUT performance for record storage is unacceptably slow, sometimes taking minutes to complete and hindering the adoption of delay-intolerant applications. To address this challenge, this research paper presents three significant contributions. First, we present the design of Optimistic Provide, an approach to accelerate DHT PUT operations in Kademlia-based IPFS networks while maintaining full backward compatibility. Second, we implement and deploy the mechanism and see its usage in the de-facto IPFS deployment, Kubo. Third, we evaluate its effectiveness in the IPFS and Filecoin DHTs. We confirm that we enable sub-second record storage from North America and Europe for 90% of PUT operations while reducing networking overhead by over 40% and maintaining record availability.

Given the high cost associated with directly measuring the traffic matrix (TM), researchers have dedicated decades to devising methods for estimating the complete TM from low-cost link loads by solving a set of heavily ill-posed linear equations. Today’s increasingly intricate networks present an even greater challenge: the routing matrix within these equations can no longer be deemed reliable. To address this challenge, we, for the first time, employ a flow-based generative model for TM estimation by establishing an invertible correlation between TM and link loads, oblivious of the routing matrix. We demonstrate that the lost information within the ill-posed equations can be independently segregated from the TM. Our model collaboratively learns the invertible correlations between TM and link loads as well as the distribution of the lost information. As a result, our model can unbiasedly reverse-transform the link loads to the true TM. Our model has undergone extensive experiments on two real-world datasets. Surprisingly, even without knowledge of the routing matrix, it significantly outperforms six representative baselines in deterministic and noisy routing scenarios regarding estimation accuracy and distribution similarity. Particularly, if the actual routing matrix is absent, our model can improve the performance of the best baseline by 41% ∼ 58%.

Bluetooth Low Energy (BLE) is a prevalent technology in various applications due to its low power consumption and wide device compatibility. Despite its numerous advantages, the encryption methods of BLE often expose devices to potential attacks. To fortify security, we investigate the application of Physical-layer Key Generation (PKG), a promising technology that enables devices to generate a shared secret key from their shared physical environment. We propose a distinctive approach that capitalizes on the inherent characteristics of BLE to facilitate efficient PKG. We harness the constant tone extension within BLE protocols to extract comprehensive physical layer information and introduce an innovative method that employs Legendre polynomial quantization for PKG. This method facilitates the exchange of secret keys with a high key matching rate and a high key generation rate. The efficacy of our approach is validated through extensive experiments on a software-defined radio platform, underscoring its potential to enhance security in the rapidly expanding field of BLE applications.

The increasing pervasiveness of intelligent mobile applications requires to exploit the full range of resources offered by the mobile-edge-cloud network for the execution of inference tasks. However, due to the heterogeneity of such multi-tiered networks, it is essential to make the applications’ demand amenable to the available resources while minimizing energy consumption. Modern dynamic deep neural networks (DNN) achieve this goal by designing multi-branched architectures where early exits enable sample-based adaptation of the model depth. In this paper, we tackle the problem of allocating sections of DNNs with early exits to the nodes of the mobile-edge-cloud system. By envisioning a 3-stage graph-modeling approach, we represent the possible options for splitting the DNN and deploying the DNN blocks on the multi-tiered network, embedding both the system constraints and the application requirements in a convenient and efficient way. Our framework – named Feasible Inference Graph (FIN) – can identify the solution that minimizes the overall inference energy consumption while enabling distributed inference over the multi-tiered network with the target quality and latency. Our results, obtained for DNNs with different levels of complexity, show that FIN matches the optimum and yields over 65% energy savings relative to a state-of-the-art technique for cost minimization.

Blockchain systems widely employ Byzantine fault-tolerant (BFT) protocols to ensure consistency. Improving BFT protocols’ throughput is crucial for large-scale blockchain systems. Frontier protocols face crucial problems: (i) the binary dilemma between leader bottleneck in star-based linear communication and compromised resilience in tree-based sublinear communication; and (ii) 2- or 3-round protocols restrict the phase number of one proposal, thereby limiting the scalability and parallelism of the pipeline. To overcome the above problems, this paper proposes Crackle, the first sector-based pipelined BFT protocol with a sublinear communication complexity, for a throughput improvement of consensus protocol with max resilience of (N-1)/3. We propose a sector-based communication mode to disseminate messages from the leader to a subset of replicas in each phase to accelerate consensus and split the traditional two-round protocol into 2κ phases to increase the basic pipeline scale. When implementing Crackle, we address two technical challenges: (i) to ensure Quorum Certificate (QC) validation during continuous κ phases, we design a voteMap field within each block, and verify QC by the aggregation of continuous κ voteMaps; and (ii) to achieve pipeline decoupling among shorter phases, we propose a vote-appending mechanism that accelerates the leader’s transition to the next phase. We provide comprehensive theoretical proof of the correctness of Crackle, including safety and liveness. Moreover, we implement Crackle based on a public BFT framework and deploy it on 64 cloud servers. Real experimental results reveal that Crackle achieves up to 10.36x higher throughput compared with state-of-the-art BFT protocols such as Kauri and Hotstuff.

Edge intelligence enables resource-demanding Deep Neural Network (DNN) inference without transferring original data, addressing concerns about data privacy in consumer Inter-net of Things (IoT) devices. For privacy-sensitive applications, deploying models in hardware-isolated trusted execution environments (TEEs) becomes essential. However, the limited secure memory in TEEs poses challenges for deploying DNN inference, and alternative techniques like model partitioning and offloading introduce performance degradation and security issues. In this paper, we present a novel approach for advanced model deployment in TrustZone that ensures comprehensive privacy preservation during model inference. We design a memory-efficient management method to support memory-demanding inference in TEEs. By adjusting the memory priority, we effectively mitigate memory leakage risks and memory overlap conflicts, resulting in 32 lines of code alterations in the trusted operating system. Additionally, we leverage two tiny libraries: S-Tinylib (2,538 LoCs), a tiny deep learning library, and Tinylibm (827 LoCs), a tiny math library, to support efficient inference in TEEs. We implemented a prototype on Raspberry Pi 3B+ and evaluated it using three well-known lightweight DNN models. The experimental results demonstrate that our design significantly improves inference speed by 3.13 times and reduces power consumption by over 66.5% compared to non-memory optimization method in TEEs.

To reduce costs and improve performance, video Content Delivery Networks (CDNs) have started to incorporate lightweight edge nodes, e.g., WiFi access points. Because of this, it is necessary for CDNs to intelligently select which video files should be placed at their core data centers vs. these edge nodes. This is more complex than traditional CDN management, as lightweight edge nodes are much more numerous and unstable than data centers. With this in mind, we present SDPush —- a system for managing content placement in edge CDNs. SDPush tackles two problems. First, it is necessary for SDPush to select which files to proactive push. To address this, we build a file popularity prediction model that effectively identifies video files that will receive many views. Second, SDPush should determine how many replicas of each file to push. To address this, we design a model to predict the benefits of pushing particular files (regarding traffic savings) and then formulate the replica decision problem as a lightweight problem, which is solvable within seconds, even for platforms that accommodate millions of daily active users. Through a trace-driven evaluation and a live deployment on a real video platform, we validate SDPush’s effectiveness, offloading peak-period traffic by 12.1% to 23.9% from the data center to edge nodes, thereby reducing the CDN costs.

Fifth generation (5G) and beyond cellular networks are vulnerable to security threats, primarily due to the lack of integrity protection in the Radio Resource Control (RRC) layer. In order to address this problem, we propose a real-time anomaly detection framework that leverages the concept of distributed applications in 5G Open RAN networks. Specifically, we identify Physical Layer (PHY) features that can generate a reliable fingerprint, infer in a novel way the time of arrival of uplink packets lacking integrity protection, and handle cross-layer features. By identifying legitimate message sources and detecting suspicious activities through an Artificial Intelligence (AI) design, we demonstrate that Open RAN-based applications that run at the edge can be designed to provide additional security to the network. Our solution is first validated in extensive emulation environments achieving over 85% accuracy in predicting potential attacks on unseen test scenarios. We then integrate our approach into a real-world prototype with a large channel emulator to assess its real-time performance and costs. Our solution meets the low-latency real-time constraints of 2 ms, making it well-suited for real-world deployments.

The rapid growth of Internet of Things (IoT) applications has spurred the need for efficient data collection mechanisms. Traditional approaches relying on fixed infrastructure have limitations in coverage, scalability, and deployment costs. Unmanned Aerial Vehicles (UAVs) have emerged as a promising alternative due to their mobility and flexibility. In this paper, we aim to minimize the number of UAVs deployed to collect data in IoT networks while considering a delay budget for energy limitation and data freshness. To this end, we propose a novel 3-approximation dynamic-programming-based algorithm called GPUDA to address the challenges of efficient data collection from IoT devices via UAVs for real-world scenarios where the number of UAVs owned by an individual or organization is unlikely to be excessive, improving the best-known approximation ratio of 4. GPUDA is a geometric partition-based method that incorporates data rounding techniques. The experimental results demonstrate that the proposed algorithm requires 35.01% to 58.55% fewer deployed UAVs than the existing algorithms on average.

Container technology has revolutionized service deployment, offering streamlined processes and enabling container orchestration platforms to manage a growing number of container clusters. However, the deployment of containers in distributed edge clusters presents challenges due to their unique characteristics, such as bandwidth limitations and resource constraints. Existing approaches designed for cloud environments often fall short in addressing the specific requirements of edge computing. Additionally, very few edge-oriented solutions explore fundamental changes to the container design, resulting in difficulties achieving backward compatibility.In this paper, we reevaluate the fundamental layer-based structure of containers. We identify that the proliferation of redundant files and operations within image layers hinders efficient container deployment. Drawing upon the crucial insight of enhancing layer reuse and extracting benefits from it, we introduce BREAK, a holistic approach centered on layer structure throughout the entire container deployment pipeline, ensuring backward compatibility. BREAK refactors image layers and proposes an edge-oriented cache solution to enable ubiquitous and shared layers. Moreover, it addresses the complete deployment pipeline by introducing a customized scheduler and a tailored storage driver. Our results demonstrate that BREAK accelerates the deployment process by up to 2.1× and reduces redundant image size by up to 3.11× compared to state-of-the-art approaches.

Sketch-based traffic measurement is a crucial building block for monitoring traffic statistics and ensuring the quality of services of end-host applications. However, existing approaches for building sketches in end-hosts exhibit poor packet processing performance or high CPU consumption. In this paper, we propose MPU, which automatically offloads sketch-based measurement to the emerging hardware, DPU. MPU consists of a sketch analyzer that profiles sketch resource consumption and an optimization framework that formulates the offloading problem and maximizes sketch performance on DPU. We implement MPU on the NVIDIA BlueField DPU. Our testbed results indicate that MPU achieves 85% lower per-packet processing latency and 47% higher traffic measurement accuracy when compared to existing approaches.

Current network functions consume a significant amount of energy and lack the capacity to support more expressive learning models like neuromorphic functions. The major reason is the underlying transistor-based components that require continuous energy-intensive data movements between the storage and computational units. In this research, we propose the use of a novel component, called Memristor, which can colocalize computation and storage, and provide computational capabilities. Building on memristors, we propose the concept of match-compute processing for supporting energy efficient network functions. Considering the analog processing of memristors, we propose a Probabilistic Content Addressable Memory (pCAM) abstraction which can provide analog match functions. pCAM provides deterministic and probabilistic outputs depending upon the closeness of match of an incoming query with the specified network policy. pCAM uses a crossbar array for line rate matrix multiplications on the match outputs. We proposed a match-compute packet processing architecture and developed the programming abstractions for a baseline network function, i.e., Active Queue Management, which drops packets based upon the higher-order derivatives of sojourn times and buffer sizes. The analysis of match-compute processing over a physically fabricated memristor chip showed only 0.01 fJ/bit/cell of energy consumption, which is 50 times less than the traditional match-action processing.

In stream processing applications, an operator is often instantiated into multiple parallel execution instances, referred to as executors, to facilitate large-scale data processing. Due to unpredictable changes in executor workloads, data tuples processed by different executors may exhibit varying latency. In particular, within the same operator, the executor with the maximum latency significantly impacts the end-to-end (E2E) latency of the application. Existing solutions, such as load balancing and horizontal scaling, which involve workload migration, often incur substantial time overhead induced by state migration and synchronization. In contrast, elastically scaling up/down resources of executors rather than moving workloads can not only effectively handle workload fluctuations but also offer rapid adjustments; however, prior works only considered CPU scaling with the assumption of sufficient memory.In this paper, we propose Emma, an elastic multi-resource manager. Emma leverages the resource elasticity of lightweight virtualization containers, e.g., Linux containers, to resize the resource of executors at runtime. The core of Emma is a multi-resource provisioning plan that conducts performance analysis and resource adjustment in real-time. We explore the relationship between resources and performance experimentally and theoretically, guiding the plan to adaptively allocate the appropriate combination of resources to each executor to 1) accommodate the dynamic workload; 2) efficiently utilize resources to enhance the performance of as many executors as possible. Additionally, we propose an online learning method that makes the manager seamlessly adapt to diverse stream applications. We integrate Emma with Apache Samza, and our experiments show that compared to existing solutions, Emma can significantly reduce latency by orders of magnitude in real-world applications.