The rapid growth of Large Language Models (LLMs) raises significant challenges for deployment in resource-constrained environments. Existing compression approaches, such as low-rank decomposition and quantization, are typically applied independently, which limits their effectiveness and fails to exploit their complementarity. To address this issue, we present a training-free framework for joint compression that integrates low-rank decomposition with mixed-precision quantization. We formulate the allocation of layer-wise rank and bit-width as a combinatorial optimization problem, guided by an input-aware sensitivity metric to allocate resources where they yield the highest accuracy retention. We further develop a sample-aware low-rank decomposition scheme with theoretical guarantees, and introduce a unified difference matrix to mitigate the coupled errors from structural approximation and quantization. Extensive experiments on diverse LLM architectures and datasets demonstrate that our method achieves state-of-the-art compression, reducing model size to 20% of the original while preserving inference accuracy. The code is available at https://github.com/zzzzzjq0126/HALO.git

Semantic communication is a transformative approach for efficient data transmission in bandwidth-constrained IoT networks, particularly for device-to-device (D2D) communication and real-time systems. To address the challenges of high-quality data reconstruction in low-bandwidth scenarios, such as IoT-enabled underwater or wireless networks, this article proposes a novel framework for ultraefficient image and video frame compression. The framework employs an optimized feature extraction method to capture essential semantic information from images or video frames, converting them into compact grayscale representations to minimize data volume for bandwidth-constrained devices. At the receiver, a dual-decoding strategy reconstructs structural details using lightweight semantic reconstruction techniques, followed by color attribute recovery, ensuring high visual quality in real-time applications. This approach enhances transmission efficiency, reduces latency, and maintains information integrity in resource-limited IoT environments. Experimental results on the Kodak dataset show a 7.02% compression ratio (PSNR = 13 dB), with a 9.76% improvement in bandwidth efficiency compared to existing methods. For 4K images and video frames, the framework achieves a 98.35% MS-SSIM retention rate under SNR conditions of 1–15 dB, demonstrating robust performance for real-time IoT communication.

As High-speed rail becomes a popular way to travel, users have a high demand for streaming services. In High-speed rail scenarios, users move fast and base stations handover frequently. Most of the existing network bandwidth prediction algorithms and bitrate selection algorithms are proposed based on low-speed scenarios. These algorithms are difficult to adapt to high-speed mobile scenarios. To solve this problem, this paper proposes an adaptive streaming media transmission method using edge computing, High-speed rail status and cross-layer information (EHCI) in the 5G network environment. Firstly, a QoE model and a coordinated transmission architecture using edge computing, High-speed rail operation status and cross-layer information are proposed. Secondly, a media transcoding algorithm and rate selection algorithm are proposed. Finally, the simulation experiment is carried out in this paper. Simulation results demonstrate that the method proposed in this paper can well improve the QoE of High-speed rail passengers, and is helpful to the study of the optimized transmission of streaming media in High-speed rail scenarios.

With the rapid development of the Internet of Things (IoT), the continuous emergence of cyberattacks have brought great threat to the security of the network. Intrusion Detection System (IDS) which can identify malicious network attacks has become a strong tool to ensure network security. Many deep learning-based approaches have been used in IDS. However, most of these researches ignore the internal structural characteristics of the network traffic, and cannot accurately learn the key features of the malicious traffic. Thus, they have a low accuracy in classifying different kinds of network attacks. In this paper, we build an intrusion detection model DAL (Dense-Attention-LSTM, DAL), in which dense dilated convolutions is used to extract the underlying features of the network traffic. Then, attention mechanism is utilized to capture key features which represent the structural characteristics of traffic data. Moreover, CuDNN-based long short-term memory network is used to learn time-related information of the traffic while accelerating the convergence of the model. Finally, global maxpooling is adopted to compress data and to improve the generalization capabilities of the proposed model. Experimental results on UNSW-NB15 dataset show that the binary classification accuracy of the proposed model is up to 92.65%. Further, it can also identify various attacks with the accuracy of 81.28%. The performance of our model is better than some competing machine learning methods and some deep learning methods. We published our code at https://github.co-m/cKiNg37/IDS-model-DAL.

Vehicular edge computing has been a promising paradigm to offer low-latency and high reliability vehicular services for users. Nevertheless, for compute-intensive vehicle applications, most previous researches cannot perform them efficiently due to both the inadequate of infrastructure construction and the computing resource bottleneck of the edge server. Motivated by the fact that there is a large number of outside parked vehicles with rich and underutilized resources in the urban area, we propose the idea of parking edge computing, which makes use of the parked vehicles to assist edge servers in offloaded task handling. Specifically, on-street and off-street parked vehicles are first organized into parking clusters to act as virtual edge servers, participating in offloaded tasks execution in our framework. Second, a novel task scheduling algorithm is designed to jointly decide edge server selection and resource assignment. Furthermore, a local task scheduling policy is proposed as well, which reasonably allocates parked vehicles to perform the tasks with the aim of further improving task offloading performance. Finally, a time-related trajectory prediction model based on the random forest model is built, which helps to send back output result accurately. Our framework not only requires no additional infrastructure investment but also provides adequate computing resources. Simulation results based on a real city map and realistic traffic situations demonstrate that our framework provides more efficient and stable offloading services, especially in a large number of task requests condition.

Recent breakthrough progress of wireless energy transfer technology and rechargeable lithium battery technology emerge the wireless rechargeable sensor networks(WRSNs). In WRSNs, how to schedule the mobile charger to efficiently replenish energy for sensor nodes is very challenging. However, most of current existing WRSNs mobile energy replenishment schemes either cannot adapt to the dynamic and diversity energy consumption of sensors in actual environment or leave out of consideration of the fairness of charging response, which may result in sensor nodes failure due to energy depletion and low charging performance. Particularly, the nodes failure issue will get worse when there are a large number of charging requirements in the network. In this paper, we explore the node energy depletion problem in mobile charging for WRSNs and propose a node failure avoidance online charging scheme(NFAOC). To avoid the nodes failure due to energy depletion, NFAOC compares the current maximum tolerable charging delay of each request node with its shortest waiting time for charging, and then always chooses the nodes which make the least number of other request nodes suffer from energy depletion as the charging candidates. Simulation results demonstrate that NFAOC can effectively solve the node energy depletion problem with lower charging latency and charging cost in comparison with other current existing online charging schemes.

Uncertain data are inherent in mobile crowd sensing applications, and the objects that they correspond to are usually vaguely specified. In order to improve performance, we often increase the number of features. However, the more features are used, the more redundancy and cost are involved correspondingly. Therefore, the number of features we selected for a specified application is a tradeoffs between the accuracy and the cost. In this paper, we model such tradeoffs between accuracy and cost as an optimization problem. Moreover, for investigating this problem, we propose to model the sensing with multiple features under a hypercube structure. In our scheme, each feature of uncertain objects is represented as a component of the vertex's coordinate in hypercube. At the same time, we prefer to define the edges between vertices with relative entropy rather than Euclidean distance. Because the former one could accurately measures the difference between two probability distributions of data. We evaluate our proposed schemes with real data of a crowd sensing recognition case, which are collected by smartphones with sensors.

The delay tolerant mobile sensor network (DTMSN) is a new type of sensor network for pervasive information gathering. Although similar to conventional sensor networks in hardware components, DTMSN owns some unique characteristics such as sensor mobility, intermittent connectivity, etc. Therefore, traditional data gathering methods can not be applied to DTMSN. In this paper, we propose an efficient motion tendency-based data delivery scheme (MTAD) tailored for DTMSN. By using sink broadcast instead of GPS, MTAD obtains the information about the nodal motion tendency with small overhead. The information can then be used to evaluate the node's effective delivery ability and provide guidance for message transmission. MTAD also employs the message survival time to effectively manage message queues. Our simulation results show that, compared with other DTMSN data delivering approaches, MTAD achieves not only a relatively longer network lifetime but also a higher message delivery ratio with lower transmission overhead and data delivery delay.

The basic operation of delay tolerant sensor network (DTSN) is to finish pervasive data gathering in networks with intermittent connectivity, while the publish/subscribe (Pub/Sub for short) paradigm is used to deliver events from a source to interested clients in an asynchronous way. Recently, to extend a Pub/Sub system in DTSN has become a promising topic. However, due to the unique characteristic of frequent partitioning in DTSN, to extend a Pub/Sub system in DTSN is a considerably difficult and challenging problem, and there is no good solution to it in existing works. To adapt Pub/Sub systems to DTSN, we propose CED, a community-based event delivery protocol. In our design, event delivery is based on several unchanged communities, which are formed by sensor nodes in the network according to their connectivity. CED consists of two components: event delivery and queue management. In event delivery, events in a community are delivered to mobile subscribers once a subscriber comes into the community, for improving the data delivery ratio. The queue management employs both the event successful delivery time and the event survival time to decide whether an event should be delivered or dropped for minimizing the transmission overhead. The effectiveness of CED is demonstrated through comprehensive simulation studies.

The basic operation of delay tolerant mobile sensor network (DTMSN) is for pervasive data gathering in networks with intermittent connectivity, where traditional data gathering methods can not be applied. In this paper, an efficient mobility prediction-based adaptive data gathering protocol (MPAD) based on the random waypoint mobility model tailored for DTMSN is proposed. In MPAD, a node independently makes decision to replicate messages and send them to the neighbor sensor nodes with a higher probability of meeting the sink node. MPAD consists of two components for data transmission and queue management. Data transmission makes decisions on when and where to transmit data messages according to the node delivery probability, and the queue management employs the message survival time to decide whether the message should be transmitted or dropped for minimizing the transmission overhead. Simulation results show that the proposed MPAD achieves the longer network lifetime and the higher message delivery ratio with the lower transmission overhead and data delivery delay than some other previous solutions designed for DTMSN, such as direct transmission, flooding and message fault tolerance-based data delivery protocol (FAD).