INFOCOM 2024 Papers

Bloom Filters are a space-efficient data structure used for the testing of membership in a set that errs only in the False Positive direction. However, the standard analysis that measures this False Positive rate provides a form of worst case bound that is both overly conservative for the majority of network applications that utilize Bloom Filters, and reduces accuracy by not taking into account the actual state (number of bits set) of the Bloom Filter after each arrival. In this paper, we more accurately characterize the False Positive dynamics of Bloom Filters as they are commonly used in networking applications. In particular, network applications often utilize a Bloom Filter that “recycles”: it repeatedly fills, and upon reaching a certain level of saturation, empties and fills again. In this context, it makes more sense to evaluate performance using the average False Positive rate instead of the worst case bound. We show how to efficiently compute the average False Positive rate of recycling Bloom Filter variants via renewal and Markov models. We apply our models to both the standard Bloom Filter and a "two-phase" variant, verify the accuracy of our model with simulations, and find that the previous analysis’ worst-case formulation leads to up to a 30% reduction in the efficiency of Bloom Filter when applied in network applications, while two-phase overhead diminishes as the needed False Positive rate is tightened.

Machine learning has been adopted for efficient cooperative spectrum sensing. However, it incurs an additional security risk due to attacks leveraging adversarial machine learning to create malicious spectrum sensing values to deceive the fusion center, called adversarial spectrum attacks. In this paper, we propose an efficient framework for detecting adversarial spectrum attacks. Our design leverages the concept of the distance to the decision boundary (DDB) observed at the fusion center and compares the training and testing DDB distributions to identify adversarial spectrum attacks. We create a computationally efficient way to compute the DDB for machine learning based spectrum sensing systems. Experimental results based on realistic spectrum data show that our method, under typical settings, achieves a high detection rate of up to 99% and maintains a low false alarm rate of less than 1%. In addition, our method to compute the DDB based on spectrum data achieves 54%–64% improvements in computational efficiency over existing distance calculation methods. The proposed DDB-based detection framework offers a practical and efficient solution for identifying malicious sensing values created by adversarial spectrum attacks.

Online data analytics with local privacy protection is widely adopted in real-world applications. Despite numerous endeavors in this field, significant gaps in utility and functionality remain when compared to its offline counterpart. This work demonstrates that private data analytics can be conducted online without excess utility loss, even at a constant factor. We present an optimal, streamable mechanism for local differentially private sparse vector estimation. The mechanism enables a range of online analytics on streaming binary vectors, including multi-dimensional binary, categorical, or set-valued data. By leveraging the negative correlation of occurrence events in the sparse vector, we attain an optimal error rate under local privacy constraints, only requiring streamable computations during the input’s data-dependent phase. Through experiments with both synthetic and real-world datasets, our proposals have been shown to reduce error rates by 40% to 60% compared to SOTA approaches.

Emerging reconfigurable optical communication technologies allow to enhance datacenter topologies with demand-aware links optimized towards traffic patterns. This paper studies the algorithmic problem of jointly optimizing topology and routing in such demand-aware networks to minimize congestion, along two dimensions: (1) splittable or unsplittable flows, and (2) whether routing is segregated, i.e., whether routes can or cannot combine both demand-aware and demand-oblivious (static) links.For splittable and segregated routing, we show that the problem is generally 2-approximable, but APX-hard even for uniform demands induced by a bipartite demand graph. For unsplittable and segregated routing, we establish upper and lower bounds of O (log m/ log log m) and Ω (log m/ log log m), respectively, for polynomial-time approximation algorithms, where m is the number of static links. We further reveal that under un-/splittable and non-segregated routing, even for demands of a single source (resp., d estina tion), the problem cannot be approximated better than $\Omega \left({\frac{{{c_{\max }}}}{{{c_{\min }}}}}\right)$ unless P=NP, where cmax(resp., cmin) denotes the maximum (resp., minimum) capacity. It remains NP-hard for uniform capacities, but is tractable for a single commodity and uniform capacities.Our trace-driven simulations show a significant reduction in network congestion compared to existing solutions.

Feature-only partition of large graph data in distributed Graph Neural Network (GNN) training offers advantages over commonly adopted graph structure partition, such as minimal graph preprocessing cost and elimination of cross-worker subgraph sampling burdens. Nonetheless, performance bottleneck of GNN training with feature-only partitions still largely lies in the substantial communication overhead due to cross-worker feature fetching. To reduce the communication overhead and expedite distributed training, we first investigate and answer two key questions on convergence behaviors of GNN model in feature-partition based distribute GNN training: 1) As no worker holds a complete copy of each feature, can gradient exchange among workers compensate for the information loss due to incomplete local features? 2) If the answer to the first question is negative, is feature fetching in every training iteration of the GNN model necessary to ensure model convergence? Based on our theoretical findings on these questions, we derive an optimal communication plan that decides the frequency for feature fetching during the training process, taking into account bandwidth levels among workers and striking a balance between model loss and training time. Extensive evaluation demonstrates consistent results with our theoretical analysis, and the effectiveness of our proposed design.

Since the advent of decentralized financial applications based on blockchains, new attacks that take advantage of manipulating the order of transactions have emerged. To this end, order fairness protocols are devised to prevent such order manipulations. However, existing order fairness protocols adopt time-consuming mechanisms that bring huge computation overheads and defer the finalization of transactions to the following rounds, eventually compromising system performance. In this work, we present Auncel, a novel consensus protocol that achieves both order fairness and high performance. Auncel leverages a weight-based strategy to order transactions, enabling all transactions in a block to be committed within one consensus round, without cost computation and further delays. Furthermore, Auncel achieves censorship resistance by integrating the consensus protocol with the fair ordering strategy, ensuring all transactions can be ordered fairly. To reduce the overheads introduced by the fair ordering strategy, we also design optimization mechanisms, including dynamic transaction compression and adjustable replica proposal strategy. We implement a prototype of Auncel based on HotStuff and construct extensive experiments. Experimental results show that Auncel can increase the throughput by 6× and reduce the confirmation latency by 3× compared with state-of-the-art order fairness protocols.

Recent years have witnessed a surge of headphones (including in-ear headphones) usage in works and communications. Because of the privacy-preserve property, people feel comfortable having confidential communication wearing headphones and pay little attention to speech leakage. In this paper, we present an end-to-end eavesdropping system, mmEar, which shows the feasibility of launching an eavesdropping attack on headphones leveraging a commercial mmWave radar. Different from previous works that realize eavesdropping by sensing speech-induced vibrations with reasonable amplitude, mmEar focuses on capturing the extremely faint vibrations with a low signal-to-noise ratio (SNR) on the surface of headphones. Toward this end, we propose a faint vibration emphasis (FVE) method that models and amplifies the mmWave responses to speech-induced vibrations on the In-phase and Quadrature (IQ) plane, followed by a deep denoising network to further improve the SNR. To achieve practical eavesdropping on various headphones and setups, we propose a cGAN model with a pretrain-finetune scheme, boosting the generalization ability and robustness of the attack by generating high-quality synthesis data. We evaluate mmEar with extensive experiments on different headphones and earphones and find that most of them can be compromised by the proposed attack for speech recovery.

Integrated Sensing and Communication (ISAC) is gaining a tremendous amount of attention from both academia and industry. Recent work has brought communication capability to sensing-oriented mmWave radars, enabling more innovative applications. These solutions, however, either require hardware modifications or suffer from limited data rates. This paper presents Talk2Radar, which builds a faster communication channel between smartphone speakers and mmWave radars, without any hardware modification to either commodity smartphones or off-the-shelf radars. In Talk2Radar, a smartphone speaker sends messages by playing carefully designed sounds. A mmWave radar acting as a data receiver captures the emitted sounds by detecting the sound-induced smartphone vibrations, and then decodes the messages. Talk2Radar characterizes smartphone speakers for speaker-to-mmWave radar communication and addresses a series of technical challenges, including modulation and demodulation of extremely weak sound-induced vibrations, multi-speaker concurrent communication and human motion suppression. We implement and evaluate Talk2Radar in various practical settings. Experimental results show that Talk2Radar can achieve a data rate of up to 400bps with an average BER of less than 5%, outperforming the state-of-the-art by approximately 33×.

Today’s video streaming service providers have exploited cloud-edge collaborative networks for geo-distributed video delivery. The existing content delivery network (CDN) scheduling and adaptive bitrate algorithms may not fully utilize edge resources or lack a global control to optimize resource sharing. The emerging super-resolution (SR) approach can unleash the potential of leveraging computation resources to compensate for bandwidth consumption, by producing high-quality videos from low-resolution contents. Yet the uncertain SR resource sensitivity and its interplay with bitrate adaptation are underexplored. In this work, we propose Rosevin, the first resource scheduler that jointly decides the bitrates and fine-grained resource allocation to perform SR at the edge, which can learn to optimize the long-term QoE for distributed end users. To handle the time-varying and complex space of decisions as well as a non-smooth objective function, Rosevin realizes a novel online combinatorial learning algorithm, which nicely integrates convex optimization theories and online learning techniques. In addition to theoretically analyzing its performance, we implement an SR-assisted video streaming prototype of Rosevin and demonstrate its advantages over several video delivery benchmarks.

Facial images play a crucial role in many web and security applications, but their uses come with notable privacy risks. Despite the availability of various face anonymization algorithms, they often fail to withstand advanced attacks while struggling to maintain utility for subsequent applications. We present two novel face anonymization algorithms that utilize feature operations to overcome these limitations. The first algorithm utilizes perturbation and matching of high-level features, whereas the second algorithm enhances this approach by also incorporating perturbation of low-level features along with regularization. These algorithms significantly enhance the utility of anonymized images while ensuring differential privacy. Additionally, we introduce a task-based benchmark to enable fair and comprehensive evaluations of privacy and utility across different algorithms. Through experiments, we demonstrate that our algorithms outperform others in preserving the utility of anonymized facial images in classification tasks while effectively protecting against a wide range of attacks.

Federated learning (FL) is widely used in edge environments as a privacy-preserving collaborative learning paradigm. However, edge devices often have heterogeneous computation capabilities and data distributions, hampering the efficiency of co-training. Existing works develop staleness-aware semi-asynchronous FL that reduces the contribution of slow devices to the global model to mitigate their negative impacts. But this makes data on slow devices unable to be fully leveraged in global model updating, exacerbating the effects of data heterogeneity. In this paper, to cope with both system and data heterogeneity, we propose a clustering and two-stage aggregation-based Efficient Asynchronous Federated Learning (EAFL) framework, which can achieve better learning performance with higher efficiency in heterogeneous edge environments. In EAFL, we first propose a gradient similarity-based dynamic clustering mechanism to cluster devices with similar system and data characteristics together dynamically during the training process. Then, we develop a novel two-stage aggregation strategy consisting of staleness-aware semi-asynchronous intra-cluster aggregation and data size-aware synchronous inter-cluster aggregation to efficiently and comprehensively aggregate training updates across heterogeneous clusters. With that, the negative impacts of slow devices and Non-IID data can be simultaneously alleviated, thus achieving efficient collaborative learning. Extensive experiments demonstrate that EAFL is superior to state-of-the-art methods.

The demand for virtual private networks (VPNs) that provide confidentiality, integrity, and authenticity of communications is growing every year. IPsec is one of the oldest and most widely used VPN protocols, implemented between the internet protocol (IP) layer and the data link layer of the Linux kernel. This implementation method, known as bump-in-the-stack, has the advantage of being able to transparently apply IPsec to traffic without changing the application. However, its throughput efficiency (Gbps/core) is worse than regular Linux communication. Therefore, we chose the bump-in-the-wire (BITW) architecture, which handles IPsec in hardware separate from the host. Our proposed BITW architecture consists of inline cryptographic accelerators implemented in field-programmable gate arrays and a programmable switch that connects multiple such accelerators. A VPN gateway implemented with our architecture is transparent and improves the throughput efficiency by 3.51 times and power efficiency by 3.40 times over a VPN gateway implemented in the Linux kernel. It also demonstrates excellent scalability, and has been confirmed to scale to a maximum of 386.24 Gbps per tunnel, exceeding state-of-the-art technology in maximum throughput and efficiency per tunnel. In multiple-tunnels use cases, the proposed architecture improves the energy efficiency by 2.49 times.

In recent years, WiFi-based gesture recognition (WGR) has gained popularity due to its privacy-preserving nature and the wide availability of WiFi infrastructure. However, existing WGR systems suffer from scalability issues, i.e., requiring extensive data collection and re-training for each new gesture class. To address these limitations, we propose OneSense, a one-shot WiFi-based gesture recognition system that can efficiently and easily adapt to new gesture classes. Specifically, we first propose a data enrichment approach based on the law of signal propagation in physical world to generate virtual gestures, enhancing the diversity of the training set without extra overhead of real sample collection. Then, we devise an aug-meta learning (AML) framework to enable efficient and scalable few-short learning. This framework leverages two pre-training stages (i.e., aug-training and meta-training) to improve the model’s feature extraction and generalization abilities, and ultimately achieves accurate one-shot gesture recognition through fine-tuning. Experimental results demonstrate that OneSense achieves 93% one-shot gesture recognition accuracy, which outperforms the state-of-the-art approaches. Moreover, it maintains high recognition accuracy when facing new environments, user locations, and user orientations. Furthermore, the proposed AML framework reduces 86%+ pre-training latency compared to conventional meta-learning method.

Nowadays, point-of-sale (POS) terminals are no longer limited to wired connections, and many of them rely on Wi-Fi for data transmission. While Wi-Fi provides the convenience of wireless connectivity, it also introduces significant security risks. Previous research has explored Wi-Fi-based eavesdropping methods. However, these methods often rely on limited environmental robustness of Channel State Information (CSI) and require invasive Wi-Fi hardware, making them impractical in real-world scenarios. In this work, we present SThief, a practical Wi-Fi-based eavesdropping attack that leverages beamforming feedback information (BFI) exchanged between POS terminal and access points (APs) to keystroke inference on POS keypads. By capitalizing on the clear-text transmission characteristics of BFI, this attack demonstrates a more flexible and practical nature, surpassing traditional CSI-based methods. BFI is transmitted in the uplink, carrying downlink channel information that allows the AP to adjust beamforming angles. We exploit this channel information to keystroke inference. To enhance the BFI series, we use maximal ratio combining (MRC), ensuring efficiency across various scenarios. Additionally, we employ the Connectionist Temporal Classification method for keystroke inference, providing exceptional generalization and scalability. Extensive testing validates SThief’s effectiveness, achieving an impressive 81% accuracy rate in inferring 6-digit POS passwords within the top-100 attempts.

Efficient discovery of router interfaces on the IPv6 Internet is critical for network measurement and cybersecurity. However, existing solutions commonly suffer from inefficiencies due to a lack of initial probing targets (seeds), ultimately exhibiting limitations on large-scale IPv6 networks. Therefore, it is imperative to develop a methodology that enables the efficient collection of IPv6 router interfaces with limited resources, considering the impracticality of conducting a brute-force exploration across the extensive IPv6 address space. In this paper, we introduce Treestrace, an innovative asynchronous prober specifically designed for this purpose. Without prior knowledge of the networks, this tool incrementally adjusts search directions, automatically prioritizing the survey of IPv6 address spaces with a higher concentration of IPv6 router interfaces. Furthermore, we have developed a carefully crafted architecture optimized for probing performance, allowing the tool to probe at the highest theoretically possible rate without requiring excessive computational resources. Real-world tests show that Treestrace outperforms state-of-the-art works on both seed-based and seedless tasks, achieving at least a 5.57-fold efficiency improvement on large-scale IPv6 router interface discovery. With Treestrace, we discovered approximately 8 million IPv6 router interface addresses from a single vantage point within several hours.