Short-video platforms have grown rapidly by allowing users to browse rich media content through seamless swiping. However, the inherently random nature of swipe behavior creates significant challenges for bandwidth efficiency and playback continuity, often resulting in stalls and unnecessary data transfers. We present OffLoad, a new preloading framework that enhances bandwidth efficiency and playback quality using range-based downloading, which generalizes traditional chunk-based preloading to arbitrary-length segments for finer-grained control. At the core of OffLoad is a two-dimensional watch-time estimation model that jointly captures user preferences and video characteristics. Guided by this estimator, OffLoad introduces a hybrid preloading algorithm that integrates heuristic rules with a learning-based module trained directly on large-scale production data, enabling strong generalization in deployment. Following extensive system-level optimization, OffLoad has been deployed on a commercial short-video platform for more than six months. Our A/B testing results show that OffLoad increases overall user watch-time by 1.1‰, while simultaneously reducing 0.13% rebuffering events and 4.92% of bandwidth consumption.

Millimeter-wave (mmWave) communication technology with high throughput and high reliability attracts much attention in both academic and industrial fields. This technology plays a pivotal role in next-generation communication networks, offering promising solutions for high-speed data transfer. Localization of mobile mmWave communication devices is essential in this context, as it can effectively guide mmWave beam steering, thereby eliminating the need for cumbersome beam alignment processes. However, existing approaches for commercial mmWave communication devices suffer channel state fluctuations and can only work on static devices. To provide accurate localization for mobile mmWave devices, we propose MobiLoc , a neural network-based approach to enhance localization accuracy in mobile scenarios. Our method leverages Channel Frequency Response (CFR) and the angular spectrum for assistance to determine the positions. We first analyze the feasibility of classifying sensing data into different qualities. Then we implement a neural network architecture specifically designed to identify the sensing data with high quality. The effectiveness of our approach is demonstrated through comprehensive experiments conducted on commercial off-the-shelf (COTS) mmWave communication devices. Results show that MobiLoc can increase the localization accuracy significantly and reduce the median angle estimation error of mobile devices to 1.33ˆ with only single items of CFR measurements.

While deep reinforcement learning (DRL)-based congestion control (CC) algorithms outperform traditional TCP methods in dynamic networks, their robustness is challenged by the complex influence caused by state perturbations from competing flows. Existing DRL-based CC models, constrained by their black-box nature, often suffer performance degradation under dynamic network conditions, which may indirectly alter their observations. To address this challenge, we propose Mona, an adversarial attack framework that explores vulnerabilities in existing DRL-based CC schemes while helping improve their robustness. Leveraging a dual-Critic architecture and a global state representation, Mona enables centralized training with an explicit estimation mechanism embedded in the reward function, facilitating targeted attacks without direct access to the victim’s state. Based on this design, Mona can be distributedly deployed to indirectly perturb the victim flow’s observations by injecting a stealthy flow into the bottleneck link, ultimately inducing suboptimal decisions. Experiments demonstrate Mona’s effectiveness across diverse CC models, reducing victim flow throughput by 8.69%–26.67% under various network conditions in both simulated environments and real-world network deployments, and we conduct a preliminary exploration of defense strategies through a minimax-based adversarial training framework that improves model robustness by 20%.

Short video streaming has become a dominant paradigm in digital media, characterized by rapid swiping interactions and diverse media content. A key technical challenge is designing an effective preloading strategy that dynamically selects and prioritizes download tasks from an evolving playlist, balancing Quality of Experience (QoE) and bandwidth efficiency under practical commercial constraints. However, real-world analysis reveals critical limitations of existing approaches: (1) insufficient adaptation of download task sizes to dynamic conditions, and (2) watch-time prediction models that are difficult to deploy reliably at scale. In this paper, we propose DeLoad, a novel preloading framework that addresses these issues by introducing dynamic task sizing and a practical, multi-dimensional watch-time estimation method. Additionally, a Deep Reinforcement Learning (DRL)-enhanced agent is trained to optimize the download range decisions adaptively. Extensive evaluations conducted on an offline testing platform, leveraging massive real-world network data, demonstrate that DeLoad achieves significant improvements in QoE metrics (34.4%-87.4% gain). Furthermore, after deployment on a large-scale commercial short-video platform, DeLoad has increased overall user watch-time by 0.9‰ while simultaneously reducing rebuffering events and 3.76% bandwidth consumption.

In large-scale short-video platforms, CDN resource selection plays a critical role in maintaining users' Quality of Experience (QoE) while controlling escalating traffic costs. To better understand this phenomenon, we conduct in-the-wild network measurements during video playback in a production short-video system. The results reveal that CDNs delivering higher average QoE often come at greater financial cost, yet their connection quality fluctuates even within a single video-underscoring a fundamental and dynamic trade-off between QoE and cost. However, the problem of sustaining high QoE under cost constraints remains insufficiently investigated in the context of CDN selection for short-video streaming. To address this, we propose PIRA, a dynamic resource selection algorithm that optimizes QoE and cost in real-time during video playback. PIRA formally integrating QoE and cost by a mathematical model, and introduce a intra-video control-theoretic CDN resource selection approach which can balance QoE and cost under network dynamics. To reduce the computation overheads, PIRA employs state-space pruning and adaptive parameter adjustment to efficiently solve the high-dimensional optimization problem. In large-scale production experiments involving 450,000 users over two weeks, PIRA outperforms the production baseline, achieving a 2.1% reduction in start-up delay, 15.2% shorter rebuffering time, and 10% lower average unit traffic cost, demonstrating its effectiveness in balancing user experience and financial cost at scale.

Low Power Wide Area Networks (LPWANs) have emerged as a promising technology for facilitating large-scale, cost-effective connections through low-power, long-range communications. Nevertheless, the deployment of existing LPWANs is impeded by severe packet collisions. In this paper, we present OrthoRa, an innovative technology that significantly enhances the concurrency of low-power, long-range LPWAN transmissions. The cornerstone of OrthoRa lies in a groundbreaking design named Orthogonal Scatter Chirp Spreading Spectrum (OSCSS), which facilitates orthogonal packet transmissions while ensuring low signal-to-noise ratio (SNR) communication within LPWANs. Utilizing OrthoRa, different nodes can transmit packets encoded with unique orthogonal scatter chirps, enabling the receiver to decode collided packets from various nodes. We provide a theoretical validation of OrthoRa, demonstrating its capacity for high concurrency in low SNR communication. To surmount practical challenges inherent in real network deployments, we address the detection of multiple packets in collisions, the identification of scatter chirps for each packet’s decoding, and the precise synchronization of packets under Carrier Frequency Offset. We implemented OrthoRa on the HackRF One platform and conducted extensive performance evaluations. The results corroborate that OrthoRa amplifies network throughput and concurrency by a factor of 50 compared to LoRa and significantly outstripping the state-of-the-art in terms of robustness against collision time offset.

The current deployment of extremely shallow-shared-buffer switches in data center networks has posed challenges to widely adopted ECN-based congestion control algorithms, leading to the issue of ECN failure. Switches may not allocate sufficient buffer space for each port, leading to the possibility that the ECN marking threshold exceeds the buffer limit per port. This results in excessive packet loss during bursts, even before the ECN markings take effect. To address this problem, we propose BBQ, an automatic ECN tuning system based on reinforcement learning. BBQ ensures that the ECN threshold does not exceed the buffer capacity allocated to the port, thus avoiding the ECN failure issue. Besides, BBQ is designed to adapt to switches with varying buffer sizes ensuring generalization. We validate the effectiveness of BBQ through experiments conducted with shallow buffering and high bursts. The results show that BBQ efficiently controls the packet loss rate of incast flows to within 3%, 1.2 times lower than State-of-the-Arts in shallow-buffered environments.

WiFi backscatter localization is a promising technology for the Internet of Things. However, existing works cannot work well for large-scale and low-cost tags with commodity WiFi devices. We present Willow, which provides accurate localization for parallel backscatter tags with commodity WiFi devices. We design a packet-level orthogonal backscatter modulation method to generate multiple orthogonal backscatter signals and support in-band backscatter with ambient WiFi. We show that backscatter signals can be effectively extracted even under strong in-band interference. To work in real WiFi traffic, we propose adaptive packet selection-based modulation to guarantee the orthogonality of backscatter signals. For parallel localization, we propose an iterative inter-tag interference cancellation method and a location filtering method to remove location ambiguity. We theoretically analyze the effectiveness of our method in supporting parallel tags. We prototype Willow tags using low-cost hardware and implement Willow AP on commodity WiFi NIC AX200. Through extensive experiments, we show that Willow achieves a median localization error of 27 cm and supports 51 parallel tags, which is 2× and 17× better than the state-of-the-art method.

In modern network environments, the Multipath QUIC (MPQUIC) protocol significantly enhances data transmission reliability and efficiency by leveraging multiple paths. However, the challenge lies in developing an effective scheduling algorithm that can adapt to dynamic network conditions. Existing heuristic scheduling algorithms are tailored to specific environments, while learning-based algorithms lack the capability for fine-grained scheduling. To address this, we propose SSRL (RL-based Scheduler Switcher), a framework that dynamically switches among heuristic scheduling algorithms based on real-time network condition recognition. SSRL combines the advantages of both heuristic and learning-based algorithms, thereby enhancing MPQUIC's performance by reducing latency and improving bandwidth utilization while consuming less reorder buffer. We also design a scheduler selection model that leverages LSTM and Double DQN, enabling SSRL to understand network conditions better and make more effective scheduling decisions. We implement SSRL using Pytorch and conduct extensive evaluations with the NS3 network emulator. The results show that SSRL increases throughput by 23% and reduces RTT by 10%.

Network feedback, representing the impact of congestion control actions on the network, is crucial for evaluating the advantages of the actions taken. Previous reinforcement learning (RL)-based congestion control algorithms use average performances over fixed periods to measure network feedback, which fails to accurately capture the impact of an action and leads to performance degradation. In this paper, we propose Zygos, which accurately estimates network feedback. This accurate feedback can benefit other RL-based congestion control algorithms. Zygos contains a distribution-based reward correction mechanism that leverages a RL model to generate relevance distributions for the sequence rewards of each state-action pair, and then aggregates the rewards by weighted average. Zygos also adopts metagradient RL to capture network feedback offset patterns, thereby updating the relevance generation model during the training of the congestion control algorithm. Experiments show that the RL congestion control method using Zygos achieves an average 20–30% improvement in throughput and 20% decrease in delay compared to the original method, highlighting substantial enhancements in RL-based congestion control algorithms.

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×.

As the development of the Internet of Things, location service plays a more important role in mobile computing. To provide location service for the already deployed devices and objects, we present LiTag, a visible light-based localization and posture estimation solution with commercial off-the-shelf (COTS) cameras. The core of LiTag is based on the design of a chip-less and battery-less optical tag which can show different color patterns from different observation directions. After capturing a photo containing the tag, LiTag can calculate the tag position and posture by combining the color pattern and the geometric relation in camera imaging. To solve the localization ambiguity, we propose an ambiguity-avoidance method based on a projection relationship. LiTag can work with a single camera without calibration, which significantly reduces the calibration overhead and deployment costs. We implement LiTag and evaluate its performance extensively. Results show that LiTag can provide the tag position with a median error of 1$cm$in the 2D plane, a median error of 5$cm$in the 3D space, and posture estimation with a median error of$0.8^{\circ }$. We believe that LiTag has high potential to provide a low-cost and easy-to-use solution for ubiquitous localization and posture estimation with widely deployed cameras.

Low Power Wide Area Networks (LPWANs) have been shown promising in connecting large-scale low-cost devices with low-power long-distance communication. However, existing LPWANs cannot work well for real deployments due to severe packet collisions. We propose OrthoRa, a new technology which significantly improves the concurrency for low-power long-distance LPWAN transmission. The key of OrthoRa is a novel design, Orthogonal Scatter Chirp Spreading Spectrum (OSCSS), which enables orthogonal packet transmissions while providing low SNR communication in LPWANs. Different nodes can send packets encoded with different orthogonal scatter chirps, and the receiver can decode collided packets from different nodes. We theoretically prove that OrthoRa provides very high concurrency for low SNR communication under different scenarios. For real networks, we address practical challenges of multiple-packet detection for collided packets, scatter chirp identification for decoding each packet and accurate packet synchronization with Carrier Frequency Offset. We implement OrthoRa on HackRF One and extensively evaluate its performance. The evaluation results show that OrthoRa improves the network throughput and concurrency by 50× compared with LoRa.

Low Power Wide Area Networks (LPWAN) have become one of the key techniques to provide long-range, low-power communication for large-scale devices in the Internet of Things. However, LPWAN devices in real deployments (e.g., in buildings and basements) suffer from low-quality links due to signal attenuation, leading to coverage holes and significant deployment overhead. In this work, we propose Ostinato to enable communication for weak links and to enhance the coverage for real deployments of COTS LoRa. The key idea of Ostinato is to transform the original packet to a pseudo packet with repeated symbols and to concentrate the energy of multiple symbols to enhance the signal SNR. To address practical challenges, we reverse engineer the entire coding and modulation process of LoRa and propose a method to generate repeated symbols on COTS LoRa by manipulating input data bits. Thus, Ostinato can be directly used for widely deployed LoRa nodes without hardware modification. We achieve weak packet detection, synchronization, and effective decoding on the receiver side by concentrating energy from multiple symbols with phase offsets. We implement Ostinato on Software Defined Radio (SDR) platform and extensively evaluate its performance. The evaluation results show that Ostinato achieves an 8.5 dB gain on receiving sensitivity and 2.88× gain on the coverage compared with COTS LoRa.

LoRa, as a representative of Low Power Wide Area Network technology, has attracted significant attention from both academia and industry. However, the current understanding of LoRa is far from complete, and implementations have a large performance gap in SNR and packet reception rate. This article presents a comprehensive understanding of LoRa physical layer protocol (PHY) and reveals the fundamental reasons for the performance gap. We present the first full-stack LoRa PHY implementation with a provable performance guarantee. We enhance the demodulation to work under extremely low SNR (-20 dB) and analytically validate the performance, where many existing works require SNR > 0. We derive the order and parameters of decoding operations, including dewhitening, error correction, deinterleaving, and so on, by leveraging LoRa features and packet manipulation. We implement a complete real-time LoRa on the GNU Radio platform and conduct extensive experiments. Our method can achieve (1) a 100% decoding success rate while existing methods can support at most 66.7%, (2) -142 dBm sensitivity, which is the limiting sensitivity of the commodity LoRa, and (3) a 3,600-m communication range in the urban area, even better than commodity LoRa under the same setting.

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 ×.

Visible Light Positioning (VLP) has attracted much research effort recently. Most existing VLP approaches require special designed light or receiver, collecting light information or strict user operation (e.g., horizontally holding the mobile phone). This incurs a high deployment, maintenance and usage cost. We present RainbowLight, a low-cost ambient light 3D localization approach that is easy to deploy in today's buildings. Our key finding is that light through a chip of polarizer and birefringence material produces specific interference and light spectrum at different directions to the chip. We derive a model to characterize the relation for direction, light interference, and spectrum. Exploiting the model, RainbowLight calculates the direction to a chip after taking a photo containing the chip. With multiple chips, RainbowLight designs a direction intersection based method to derive the location. We implement RainbowLight and extensively evaluate its performance in various environments. The evaluation results show that RainbowLight achieves an average localization error of 3.3 cm in 2D and 9.6 cm in 3D for light on, and an error of 7.4 cm in 2D and 20.5 cm in 3D for light off scenario in the daytime.

LoRa is recently a rising star in Low Power Wide Area Network (LPWAN) family to provide low power and long range communication for large number of devices in Internet of Things. LoRa is based on Chirp Spread Spectrum (CSS) and uses chirp frequency shift to encode data. It has been shown that collision significantly degrades LoRa performance in practice. We propose FlipLoRa, a new mechanism to disentangle LoRa collisions, which allows concurrent transmission of multiple packets. The key idea of FlipLoRa is to utilize the quasi-orthogonality between upchirp and downchirp. FlipLoRa encodes packets with interleaved upchirps and downchirps instead of only using upchirps as in LoRa. We then propose a novel method to disentangle chirps and decode multiple collided packets. To evaluate the performance, we formally prove the quasi-orthogonality and analyze its applicable conditions. We validate the performance improvement by theoretical analysis. Further, we implement FlipLoRa on software-defined radio and extensively evaluate its performance for real LoRa networks. The evaluation results show that FlipLoRa can improve the throughput by 3.84x over LoRa physical layer.

The development of Internet of Things calls for ubiquitous and low-cost localization and posture estimation. We present LiTag, a visible light based localization and posture estimation solution with COTS cameras. The core of LiTag is based on the design of a chip-less and battery-less optical tag which can show different color patterns from different observation directions. After capturing a photo containing the tag, LiTag can calculate the tag position and posture by combining the color pattern and the geometry relation between the camera image plane and the real world. Unlike existing marker-based visible localization and posture estimation approaches, LiTag can work with a single camera without calibration, which significantly reduces the calibration overhead and deployment costs. We implement LiTag and evaluate its performance extensively. Results show that LiTag can provide the tag position with a median error of 1.6 cm in the 2D plane, a median error of 12 cm in the 3D space, and posture estimation with a median error of 1°. We believe that LiTag has a high potential to provide a low-cost and easy-to-use solution for ubiquitous localization and posture estimation with existing widely deployed cameras.

Visible Light Positioning (VLP) has attracted much research effort recently. Most existing VLP approaches require special designed light or receiver, collecting light information or strict user operation (e.g., horizontally holding mobile phone). This incurs a high deployment, maintenance and usage overhead. We present RainbowLight, a low cost ambient light 3D localization approach easy to deploy in today's buildings. Our key finding is that light through a chip of polarizer and birefringence material produces specific interference and light spectrum at different directions to the chip. We derive a model to characterize the relation for direction, light interference and spectrum. Exploiting the model, RainbowLight calculates the direction to a chip after taking a photo containing the chip. With multiple chips, RainbowLight designs a direction intersection based method to derive the location. We implement RainbowLight and extensively evaluate its performance in various environments. The evaluation results show that RainbowLight achieves an average localization error of 3.3 cm in 2D and 9.6 cm in 3D for light on, and an error of 7.4 cm in 2D and 20.5 cm in 3D for light off scenario in daytime.

Most existing VLP approaches require special designed light or receiver, collecting light information or strict user operation (e.g., horizontally holding mobile phone). This incurs a high deployment, maintenance and usage overhead. In this demo, we present RainbowLight, a low cost ambient light 3D localization approach easy to deploy in today's buildings. Our key finding is that light through a chip of polarizer and birefringence material produces specific interference and light spectrum at different directions to the chip. We derive a model to characterize the relation for direction, light interference and spectrum. Exploiting the model, RainbowLight calculates the direction to a chip after taking a photo containing the chip. With multiple chips, RainbowLight designs a direction intersection based method to derive the location. This demo shows our prototype of implementation, with simple photo capturing and deriving location of camera on mobile phone. The evaluation results show that RainbowLight achieves an average localization error of 3.3 cm in 2D and 9.6 cm in 3D for light on, and an error of 7.4 cm in 2D and 20.5 cm in 3D for light off scenario in daytime.

Device-to-device (D2D) communication is widely used for mobile devices and Internet of Things (IoT). Authentication and key agreement are critical to build a secure channel between two devices. However, existing approaches often rely on a pre-built fingerprint database and suffer from low key generation rate. We present GeneWave, a fast device authentication and key agreement protocol for commodity mobile devices. GeneWave first achieves bidirectional initial authentication based on the physical response interval between two devices. To keep the accuracy of interval estimation, we eliminate time uncertainty on commodity devices through fast signal detection and redundancy time cancellation. Then we derive the initial acoustic channel response (ACR) for device authentication. We design a novel coding scheme for efficient key agreement while ensuring security. Therefore, two devices can authenticate each other and securely agree on a symmetric key. GeneWave requires neither special hardware nor pre-built fingerprint database, and thus it is easy-to-use on commercial mobile devices. We implement GeneWave on mobile devices (i.e., Nexus 5X and Nexus 6P) and evaluate its performance through extensive experiments. Experimental results show that GeneWave efficiently accomplish secure key agreement on commodity smartphones with a key generation rate 10x faster than the state-of-the-art approach.