INFOCOM 2021 Papers

Motivated by the recent trend of integrating artificial intelligence into the Internet-of-Things (IoT), we consider the problem of scheduling packets from multiple sensing sources to a central controller over a wireless network. Here, packets from different sensing sources have different values or degrees of importance to the central controller for intelligent decision making. In such a setup, it is critical to provide timely and valuable information for the central controller. In this paper, we develop a parameterized maximum-weight type scheduling policy that combines both the AoI metrics and Upper Confidence Bound (UCB) estimates in its weight measure with parameter η. Here, UCB estimates balance the tradeoff between exploration and exploitation in learning and are critical for yielding a small cumulative regret. We show that our proposed algorithm yields the running average total age at most by O(N2η). We also prove that our proposed algorithm achieves the cumulative regret over time horizon T at most by O(NT/η+ √{NTlogT} ). This reveals a tradeoff between the cumulative regret and the running average total age: when increasing η, the cumulative regret becomes smaller, but is at the cost of increasing running average total age. Simulation results are provided to evaluate the efficiency of our proposed algorithm.

Multi-user panoramic video streaming demands 4~6× bandwidth of a regular video with the same resolution, which poses a significant challenge on the wireless scheduling design to achieve desired performance. On the other hand, recent studies reveal that one can effectively predict the user's Field-of-View (FoV) and thus simply deliver the corresponding portion instead of the entire scenes. Motivated by this important fact, we aim to employ autoregressive process for motion prediction and analytically characterize the user's successful viewing probability as a function of the delivered portion. Then, we consider the problem of wireless scheduling design with the goal of maximizing application-level throughput (i.e., average rate for successfully viewing the desired content) and service regularity performance (i.e., how often each user gets successful views) subject to the minimum required service rate and wireless interference constraints. As such, we incorporate users' successful viewing probabilities into our scheduling design and develop a scheduling algorithm that not only asymptotically achieves the optimal application-level throughput but also provides service regularity guarantees. Finally, we perform simulations to demonstrate the efficiency of our proposed algorithm using a real dataset of users' head motion.

Wi-Fi-enabled devices such as smartphones periodically search for available networks by broadcasting probe requests which encapsulate MAC addresses as the device identifiers. To protect privacy (user identity and location), modern devices embed random MAC addresses in their probe frames, the so-called MAC address randomization. Such randomization greatly hampers statistical analysis such as people counting and trajectory inference. To mitigate its impact while respecting privacy, we propose Espresso, a simple, novel and efficient approach which establishes probe request association under MAC address randomization. Espresso models the frame association as a flow network, with frames as nodes and frame correlation as edge cost. To estimate the correlation between any two frames, it considers the multimodality of request frames, including information elements, sequence numbers and received signal strength. It then associates frames with minimum-cost flow optimization. To the best of our knowledge, this is the first piece of work that formulates the probe request association problem as network flow optimization using frame correlation. We have implemented Espresso and conducted extensive experiments in a leading shopping mall. Our results show that Espresso outperforms the state-of-the-art schemes in terms of discrimination accuracy (> 80%) and V-measure scores (> 0.85).

Clients (e.g., browsers) and servers require public key certificates to establish secure connections. When a client accesses a server, it needs to check the signature, expiration time, and revocation status of the certificate to determine whether the server is reliable. The existing solutions for checking certificate status either have a long update cycle (e.g., CRL, CRLite) or violate clients' privacy (e.g., OCSP, CCSP), and these solutions also have the problem of trust concentration. In this paper, we present PROCESS, an online privacy-preserving on-chain certificate status service based on the blockchain architecture, which can ensure decentralized trust and provide privacy protection for clients. Specifically, we design Counting Garbled Bloom Filter (CGBF) that supports efficient queries and BlockOriented Revocation List (BORL) to update CGBF timely in the blockchain. With CGBF, we design a privacy-preserving protocol to protect clients' privacy when they check the certificate statuses from the blockchain nodes. Finally, we conduct experiments and compare PROCESS with another blockchain-based solution to demonstrate that PROCESS is suitable in practice.

In recent years, to sustain the resource-intensive computational needs for training deep neural networks (DNNs), it is widely accepted that exploiting the parallelism in large-scale computing clusters is critical for the efficient deployments of DNN training jobs. However, existing resource schedulers for traditional computing clusters are not well suited for DNN training, which results in unsatisfactory job completion time performance. The limitations of these resource scheduling schemes motivate us to propose a new computing cluster resource scheduling framework that is able to leverage the special layered structure of DNN jobs and significantly improve their job completion times. Our contributions in this paper are three-fold: i) We develop a new resource scheduling analytical model by considering DNN's layered structure, which enables us to analytically formulate the resource scheduling optimization problem for DNN training in computing clusters; ii) Based on the proposed performance analytical model, we then develop an efficient resource scheduling algorithm based on the widely adopted parameter-server architecture using a sum-of-ratios multi-dimensional-knapsack decomposition (SMD) method to offer strong performance guarantee; iii) We conduct extensive numerical experiments to demonstrate the effectiveness of the proposed schedule algorithm and its superior performance over the state of the art.

Real-time sampling and estimation of autoregressive Markov processes is considered in random access channels. Two classes of policies are studied: (i) oblivious policies in which decision making is independent of the source realizations, and (ii) non-oblivious policies in which sources are observed causally for decision making. In the first class, minimizing the expected time-average estimation error is equivalent to minimizing the expected age of information (AoI). Lower and upper bounds are provided for the achievable estimation error in this class and age-based threshold policies are shown to provide a two-fold improvement compared to the state-of-the-art. In the second class, an error-based threshold policy is proposed: a transmitter becomes active when its error exceeds a threshold in which case it transmits probabilistically following slotted ALOHA. A closed-form expression is derived for the estimation error as a function of the peak age, the transmission delay, a term which we call the silence delay, as well as the source realization. It is analyzed approximately by considering the underlying source as a discretized Wiener process. The proposed threshold policy provides a three-fold improvement compared to oblivious policies and its performance is close to that of centralized greedy scheduling.

Among numerous indoor localization systems, WiFi fingerprint-based localization has been one of the most attractive solutions, which is known to be free of extra infrastructure and specialized hardware. To push forward this approach for wide deployment, three crucial goals on delightful deployment ubiquity, high localization accuracy, and low maintenance cost are desirable. However, due to severe challenges about signal variation, device heterogeneity, and database degradation root in environmental dynamics, pioneer works usually make a trade-off among them. In this paper, we propose iToLoc, a deep learning based localization system that achieves all three goals simultaneously. Once trained, iToLoc will provide accurate localization service for everyone using different devices and under diverse network conditions, and automatically update itself to maintain reliable performance anytime. iToLoc is purely based on WiFi fingerprints without relying on specific infrastructures. The core components of iToLoc are a domain adversarial neural network and a co-training based semi-supervised learning framework. Extensive experiments across 7 months with 8 different devices demonstrate that iToLoc achieves remarkable performance with an accuracy of 1.92m and > 95% localization success rate. Even 7 months after the original fingerprint database was established, the rate still maintains > 90%, which significantly outperforms previous works.

Emerging deep learning-based video analytics tasks demand computation-intensive neural networks and powerful computing resources on the cloud to achieve high accuracy. Due to the latency requirement and limited network bandwidth, edge-cloud systems adaptively compress the data to strike a balance between overall analytics accuracy and bandwidth consumption. However, the degraded data leads to another issue of poor tail accuracy, which means the extremely low accuracy of a few semantic classes and video frames. Autonomous robotics applications especially value the tail accuracy performance but suffer using the prior edge-cloud systems.We present Runespoor, an edge-cloud video analytics system to manage the tail accuracy and enable emerging robotics applications. We train and deploy a super-resolution model tailored for the tail accuracy of analytics tasks on the server to significantly improves the performance on hard-to-detect classes and sophisticated frames. During online operation, we use an adaptive data rate controller to further improve the tail performance by instantly adjusting the data rate policy according to the video content. Our evaluation shows that Runespoor improves class-wise tail accuracy by up to 300%, frame-wise 90%/99% tail accuracy by up to 22%/54%, and greatly improves the overall accuracy and bandwidth trade-off.

LoRa, as a representative Lower Power Wide Area Network (LPWAN) technology, shows great potential in providing low power and long range wireless communication. Real LoRa deployments, however, suffer from severe collisions. Existing collision decoding methods cannot work well for low SNR LoRa signals. Most LoRa collision decoding methods process collisions offline and cannot support real-time collision decoding in practice. To address these problems, we propose Pyramid, a real-time LoRa collision decoding approach. To the best of our knowledge, this is the first real-time multi-packet LoRa collision decoding approach in low SNR. Pyramid exploits the subtle packet offset to separate packets in a collision. The core of Pyramid is to combine signals in multiple windows and transfers variation of chirp length in multiple windows to robust features in the frequency domain that are resistant to noise. We address practical challenges including accurate peak recovery and feature extraction in low SNR signals of collided packets. We theoretically prove that Pyramid incurs a very small SNR loss (<; 0.56 dB) to original LoRa transmissions. We implement Pyramid using USRP N210 and evaluate its performance in a 20-nodes network. Evaluation results show that Pyramid achieves real-time collision decoding and improves the throughput by 2.11 ×.

Fast Internet-wide network measurement plays an important role in cybersecurity analysis and network asset detection. The vast address space of IPv6, however, makes it infeasible to apply a brute-force approach for scanning the entire network. Even worse, the extremely uneven distribution of IPv6 active addresses results in a low hit rate for active scanning. To address the problem, we propose 6Hit, a reinforcement learning-based target generation method for active address discovery in the IPv6 address space. It first divides the IPv6 address space into different regions according to the structural information of a set of known seed addresses. Then, it allocates exploration resources according to the reward of the scanning on each region. Based on the evaluative feedback from existing scanning results, 6Hit optimizes the subsequent search direction to regions that have a higher density of activity addresses. Compared with other state-of-the-art target generation methods, 6Hit achieves better performance on hit rate. Our experiments over real-world networks show that 6Hit achieves 3.5% - 11.5% hit rate for the eight candidate datasets, which is 7.7% - 630% improvement over the state-of-the-art methods.

Inferring the bandwidth of internal links from the bandwidth of end-to-end paths, so-termed bandwidth tomography, is a long-standing open problem in the network tomography literature. The difficulty is due to the fact that no existing mathematical tool is directly applicable to solve the inverse problem with a set of min-equations. We systematically tackle this challenge by designing a polynomial-time algorithm that returns the exact bandwidth value for all identifiable links and the tightest error bound for unidentifiable links for a given set of measurement paths. When measurement paths are not given in advance, we prove the hardness of building measurement paths that can be used for deriving the global tightest error bounds for unidentifiable links. Accordingly, we develop a reinforcement learning (RL) approach for measurement path construction, that utilizes the special knowledge in bandwidth tomography and integrates both offline training and online prediction. Evaluation results with real-world ISP as well as simulated networks demonstrate that compared to other path construction methods, Random and Diversity Preferred, our RL-based path construction method can build measurement paths that result in much smaller average error bound of link bandwidth.

Edge caching will play a critical role in facilitating the emerging content-rich applications. However, it faces many new challenges, in particular, the highly dynamic content popularity and the heterogeneous caching configurations. In this paper, we propose Cocktail Edge Caching, that tackles the dynamic popularity and heterogeneity through ensemble learning. Instead of trying to find a single dominating caching policy for all the caching scenarios, we employ an ensemble of constituent caching policies and adaptively select the best-performing policy to control the cache. Towards this goal, we first show through formal analysis and experiments that different variations of the LFU and LRU polices have complementary performance in different caching scenarios. We further develop a novel caching algorithm that enhances LFU/LRU with deep recurrent neural network (LSTM) based time-series analysis. Finally, we develop a deep reinforcement learning agent that adaptively combines base caching policies according to their virtual hit ratios on parallel virtual caches. Through extensive experiments driven by real content requests from two large video streaming platforms, we demonstrate that CEC not only consistently outperforms all single policies, but also improves the robustness of them. CEC can be well generalized to different caching scenarios with low computation overheads for deployment.

To cope with increasing network rates and massive traffic volumes, sketch-based methods have been extensively studied to trade accuracy for memory scalability and storage cost. However, sketches are sensitive to hash collisions due to skewed keys in real world environment, and need complicated performance control for line-rate packet streams.We present Jellyfish, a locality-sensitive sketching framework to address these issues. Jellyfish goes beyond network flow-based sketching towards fragments of network flows called subflows. First, Jellyfish splits consecutive packets from each network flow to subflow records, which not only reduces the rate contention but also provides intermediate subflow representations in form of truncated counters. Next, Jellyfish maps similar subflow records to the same bucket array and merges those from the same network flow to reconstruct the network-flow level counters. Real-world trace-driven experiments show that Jellyfish reduces the average estimation errors by up to six orders of magnitude for per-flow queries, by six orders of magnitude for entropy queries, and up to ten times for heavy-hitter queries.

Caches are pervasively used in computer networks to speed up access by reusing previous communications, where various replacement policies are used to manage the cached contents. The replacement policy of a cache plays a key role in its performance, and is thus extensively engineered to achieve a high hit ratio in benign environments. However, some studies showed that a policy with a higher hit ratio in benign environments may be more vulnerable to denial of service (DoS) attacks that intentionally send requests for unpopular contents. To understand the cache performance under such attacks, we analyze a suite of representative replacement policies under the framework of TTL approximation in how well they preserve the hit ratios for legitimate users, while incorporating the delay for the cache to obtain a missing content. We further develop a scheme to adapt the cache replacement policy based on the perceived level of attack. Our analysis and validation on real traces show that although no single policy is resilient to all the attack strategies, suitably adapting the replacement policy can notably improve the attack resilience of the cache.

Network-based intrusion detection and data leakage prevention systems inspect packets to detect if critical files such as malware or confidential documents are transferred. However, this kind of detection requires heavy computing resources in reassembling packets and only well-known protocols can be interpreted. Besides, finding similar files from a storage requires pairwise comparisons. In this paper, we present a new network-based file identification scheme that inspects packets independently without reassembly and finds similar files through inverted indexing instead of pairwise comparison. We use a contents-based chunking algorithm to consistently divide both files and packets into multiple byte sequences, called chunks. If a packet is a part of a file, they would have common chunks. The challenging problem is that packet chunking and inverted-index search should be fast and scalable enough for packet processing. The file identification should be accurate although many chunks are noises. In this paper, we use a small Bloom filter and a delayed query strategy to solve the problems. To the best of our knowledge, this is the first scheme that identifies a specific critical file from a packet over unknown protocols. Experimental results show that the proposed scheme can successfully identify a critical file from a packet.

Sharding can significantly improve the blockchain scalability, by dividing nodes into small groups called shards that can handle transactions in parallel. However, all existing sharding systems adopt complete sharding, i.e., shards are isolated. It raises additional overhead to guarantee the atomicity and consistency of cross-shard transactions and seriously degrades the sharding performance. In this paper, we present Pyramid, the first layered sharding blockchain system, in which some shards can store the full records of multiple shards thus the cross-shard transactions can be processed and validated in these shards internally. When committing cross-shard transactions, to achieve consistency among the related shards, a layered sharding consensus based on the collaboration among several shards is presented. Compared with complete sharding in which each cross-shard transaction is split into multiple sub-transactions and cost multiple consensus rounds to commit, the layered sharding consensus can commit cross-shard transactions in one round. Furthermore, the security, scalability, and performance of layered sharding with different sharding structures are theoretically analyzed. Finally, we implement a prototype for Pyramid and its evaluation results illustrate that compared with the state-of-the-art complete sharding systems, Pyramid can improve the transaction throughput by 2.95 times in a system with 17 shards and 3500 nodes.

Synchronous stochastic gradient descent (S-SGD) with data parallelism is widely used for training deep learning (DL) models in distributed systems. A pipelined schedule of the computing and communication tasks of a DL training job is an effective scheme to hide some communication costs. In such pipelined S-SGD, tensor fusion (i.e., merging some consecutive layers' gradients for a single communication) is a key ingredient to improve communication efficiency. However, existing tensor fusion techniques schedule the communication tasks sequentially, which overlooks their independence nature. In this paper, we expand the design space of scheduling by exploiting simultaneous All-Reduce communications. Through theoretical analysis and experiments, we show that simultaneous All-Reduce communications can effectively improve the communication efficiency of small tensors. We formulate an optimization problem of minimizing the training iteration time, in which both tensor fusion and simultaneous communications are allowed. We develop an efficient optimal scheduling solution and implement the distributed training algorithm ASC-WFBP with Horovod and PyTorch. We conduct real-world experiments on an 8-node GPU cluster of 32 GPUs with 10Gbps Ethernet. Experimental results on four modern DNNs show that ASC-WFBP can achieve about 1.09 × -2.48× speedup over the baseline without tensor fusion, and 1.15× -1.35× speedup over the state-of-the-art tensor fusion solution.

Users usually have different privacy demands when they contribute individual data to a dataset that is maintained and queried by others. To tackle this problem, several personalized differential privacy (PDP) mechanisms have been proposed to render statistical information of the entire dataset without revealing individual privacy. However, existing mechanisms produce query results with low accuracy, which leads to poor data utility. This is primarily because (1) some users are over protected; (2) utility is not explicitly included in the design objective. Poor data utility impedes the adoption of PDP in the real-world applications. In this paper, we present an adaptive personalized differential privacy framework, called AdaPDP. Specifically, to maximize data utility in different cases, AdaPDP adaptively selects underlying noise generation algorithms and calculates the corresponding parameters based on the type of query functions, data distributions and privacy settings. In addition, AdaPDP performs multiple rounds of utility-aware sampling to satisfy different privacy requirements for users. Our privacy analysis shows that the proposed framework renders rigorous privacy guarantee. We conduct extensive experiments on synthetic and real-world datasets to demonstrate the much less utility losses of the proposed framework over various query functions.

Edge computing is promising to provide computational resources for connected vehicles. Resource demands for edge servers vary due to vehicle mobility. It is then challenging to reserve edge servers to meet variable demands. Existing schemes rely on statistical information of resource demands to determine edge server reservation. They are infeasible in practice, since the reservation based on statistics cannot adapt to time-varying demands. In this paper, a spatio-temporal reinforcement learning scheme called DeepReserve is developed to learn variable demands and then reserve edge servers accordingly. DeepReserve is adapted from the deep deterministic policy gradient algorithm with two major enhancements. First, by observing that the spatio-temporal correlation in vehicle traffic leads to the same property in resource demands of CVs, a convolutional LSTM network is employed to encode resource demands observed by edge servers for inference of future demands. Second, an action amender is designed to make sure an action does not violate spatio-temporal correlation. We also design a new training method, i.e., DR-Train, to stabilize the training procedure. DeepReserve is evaluated via experiments based on real-world datasets. Results show it achieves better performance than state-of-the-art approaches that require accurate demand information.