Abstract Extracting Chinese Cyber Threat Intelligence (CTI) under increasingly complex advanced persistent threat scenarios is crucial, yet challenging due to domain-specific term ambiguity and frequent long, nested entities. To address polysemy, nested-label conflicts, and cross-sentence semantic discontinuity, we propose an enhanced Transformer-based entity recognition method formulated as a pointer network. On the encoder side, we build a RoBERTa model with Rotary Positional Embeddings. To handle complex positions and boundaries of heterogeneous entity types, we introduce tokenization compensation and positional-parameter compression to sharpen boundary sensitivity. In the decoder, we refine GlobalPointer and model recognition as 2D head–tail span matching, enabling direct detection of overlapping and nested entities. To mitigate long-tail bias, we introduce an entity-frequency-aware dynamic threshold and a reweighted zero-boundary log-loss to improve recall for rare entities. Experiments demonstrate an overall F1 improvement of 6.32% over baselines on Chinese CTI datasets, with absolute gains reaching 19.7% specifically on nested and long entities. These results validate the model’s effectiveness in Chinese-specific named entity recognition and its utility for high-accuracy automated CTI analysis.

Abstract Social network alignment (SNA) aims to match corresponding users across different platforms, playing a critical role in cross-platform behavior analysis, personalized recommendations, security, and privacy protection. Traditional methods based on attribute and structural features face significant challenges due to the sparsity, heterogeneity, and dynamic nature of social networks, resulting in limited accuracy and efficiency. Recent advances in graph representation learning (GRL) provide promising solutions to these issues by leveraging deep learning to extract network features, effectively addressing sparsity, integrating heterogeneous data, and adapting to network dynamics. This paper presents a comprehensive survey of SNA methods based on GRL. We first introduce key definitions and outline a framework for SNA using GRL. Next, we systematically review state-of-the-art advancements in both static and dynamic networks, considering homogeneous and heterogeneous settings, including emerging approaches integrating large language models (LLMs). We further conduct an in-depth comparative analysis, highlighting the effectiveness of different GRL-based methods, with a particular emphasis on LLM-enhanced techniques. Finally, we discuss open challenges and outline potential future research directions in this rapidly evolving field.

Path planning constitutes a critical component of penetration testing for 5G-ICPS networks. The diversity of interfaces and protocols necessitates deep analysis of vulnerability exploitation methods and cross-protocol combination strategies, significantly increasing attack path planning complexity. Furthermore, dynamic network slice configurations and physical-information coupling effects drive continuous topological evolution, causing state-space explosion and challenging path planning under uncertainty with incomplete information. To address these issues, we construct a temporal attack graph modeling 5G-ICPS attack processes, and design a GraphSAGE-based environment representation encoder. This encoder undergoes multi-tiered self-supervised pre-training, employing node-level and graph-level training to encode diverse reinforcement learning environments across attack scenarios into fixed-dimensional vector representations. This achieves decoupling from underlying topology, vulnerability specifics, and security configurations, effectively mitigating state-space explosion in large-scale networks. Subsequently, we cluster highly similar vulnerabilities and filter invalid attack actions using three typical 5G-ICPS attack constraints, compressing the agent’s exploration space. Especially, we design a customized reward function that dynamically incentivizes/penalizes actions based on compromised assets. Experimental results demonstrate significant improvements: penetration testing invalid action rates decrease from 22.3% to 7.5%, while average steps to achieve attack targets reduce by >54%. These advancements effectively reduce penetration testing costs and increase attack success rates.

Symbiotic radio (SR) is a promising technology designed to meet the increasing demand for spectrum-efficient communication. However, the small size of backscatter devices (BDs), which are typically equipped with a single antenna, poses challenges in achieving sufficient diversity or spatial multiplexing, thereby hindering the advancement of SR. To address this issue, we introduce fluid antennas (FAs) into SR, enabling devices to dynamically adjust their positions to create a favorable wireless environment and overcome spatial constraints, thereby achieving significant diversity gains. In this paper, we investigate the uplink performance of FA-assisted SR (FA-SR). First, we propose a novel collaborative cancellation channel estimation scheme based on least squares regression (CC-LSR) for scenarios with imperfect channel state information (CSI). We then derive tight lower bound expressions for the channel capacity under both perfect and imperfect CSI cases and formulate the corresponding weighted sum channel capacity (WSCC) optimization problems. The positions of the FAs and the combining vectors are jointly optimized to maximize the lower bound of the WSCC. To solve these problems, we develop joint optimization methods for both perfect and imperfect CSI scenarios using chaotic sequence-based adaptive particle swarm optimization (CSA-PSO). Nevertheless, the high computational complexity of joint optimization poses challenges for practical implementation. To this end, we propose a progressive optimization framework (POF) tailored to both perfect and imperfect CSI scenarios, in which the original problem is divided into three subproblems that are progressively solved to find locally optimal solutions. Numerical results demonstrate that POF significantly reduces computational complexity with minimal performance loss compared to joint optimization methods, particularly under imperfect CSI conditions.

While virtualization and resource pooling empower cloud networks with structural flexibility and elastic scalability, they inevitably expand the attack surface and challenge cyber resilience. Reinforcement Learning (RL)-based defense strategies have been developed to optimize resource deployment and isolation policies under adversarial conditions, aiming to enhance system resilience by maintaining and restoring network availability. However, existing approaches lack robustness as they require retraining to adapt to dynamic changes in network structure, node scale, attack strategies, and attack intensity. Furthermore, the lack of Human-in-the-Loop (HITL) support limits interpretability and flexibility. To address these limitations, we propose CyberOps-Bots, a hierarchical multi agent reinforcement learning framework empowered by Large Language Models (LLMs). Inspired by MITRE ATT&CK's “Tactics-Techniques” model, CyberOps-Bots features a two-layer architecture: (1) An upper-level LLM agent with four mod ules—ReAct planning, IPDRR-based perception, long-short term memory, and action/tool integration—performs global awareness, human intent recognition, and tactical planning; (2) Lower-level RL agents, developed via heterogeneous separated pre-training, execute atomic defense actions within localized network regions. This synergy preserves LLM adaptability and interpretability while ensuring reliable RL execution. Experiments on real cloud datasets show that, compared to state-of-the-art algorithms, CyberOps-Bots maintains network availability 68.5% higher and achieves a 34.7% jumpstart performance gain when shifting the scenarios without retraining. To our knowledge, this is the first study to establish a robust LLM-RL framework with HITL support for cloud defense.

Cell-free massive multiple-input multiple-output (CF massive MIMO) is regarded as a promising technology for next-generation wireless communication systems. However, relying on a single central processing unit (CPU) in CF massive MIMO systems is not scalable in practical networks, requiring the introduction of multiple CPUs for more efficient and feasible transmission. In this paper, we investigate a CF massive MIMO system with multiple CPUs. To obtain flexible and cost-efficient deployment, we propose to use wireless x-haul links instead of wired ones. More specifically, we assume that both the fronthaul links from the APs to the corresponding CPU and the backhaul links between CPUs operate under millimeter wave (mmWave) networks. Taking into account a tradeoff between the degree of centralized coordination and the signal overhead on the backhaul links, we consider four levels of multiple CPUs cooperation schemes from fully centralized to fully distributed. In addition, we propose a binary search method to allocate the backhaul capacities for maximizing the sum spectral efficiency (SE). Simulation results show that mmWave backhaul amplifies the compression noise introduced by mmWave fronthaul, leading to a more pronounced impact on the SE of systems. In this case, the centralized processing scheme can generate more compression noise due to the larger data overhead on the backhaul link, making the distributed processing scheme a superior processing scheme, especially when dealing with a large number of APs or significant distances between CPUs.

Cell-free massive multiple-input multiple-output (CF massive MIMO) is regarded as a promising technology for next-generation wireless communication systems. However, conventional CF massive MIMO systems typically employ fixed-position antennas (FPAs) at access points (APs), which limits the exploitation of spatial degrees of freedom (DoFs) for antenna position optimization. To address this issue, we propose the use of fluid antennas (FAs) in place of FPAs, enabling more DoFs at APs and leading to a novel FA-aided CF massive MIMO (FA-CF) architecture. In this paper, we investigate the uplink spectral efficiency (SE) of FA-CF systems with imperfect channel state information (CSI). We design a minimum mean-square error (MMSE)-based channel estimation scheme to estimate the aggregated channels between APs and user equipments (UEs). We further derive achievable SE expressions for both centralized and distributed processing schemes, including fully centralized processing (FCP), large-scale fading decoding (LSFD), and equal-gain decoding processing (EGDP). Moreover, we formulate a mean-square error (MSE) minimization problem based on the signal transmission model. To solve this problem, we develop an efficient algorithm that combines orthogonal matching pursuit (OMP) with binary search to jointly optimize FA positions and the combining matrix. In addition, we propose a protective weak-ordering (PWO) strategy to enhance the SE of the FCP scheme. Numerical results demonstrate that FA-CF significantly outperforms conventional CF systems in terms of SE, even with a limited number of antennas, and maintains strong robustness under imperfect CSI or heavy UE loads by adaptively adjusting antenna positions. These results highlight FA-CF as a promising architecture offering enhanced SE and robustness for future wireless systems, particularly in scenarios where large-scale AP deployment is infeasible or cost-constrained.

Abstract The rapid evolution of cyber threats demands efficient entity extraction from Cyber Threat Intelligence (CTI) reports to support proactive analysis and sharing. Current methods for CTI extraction falter due to a lack of domain knowledge, which can lead to the overlooking of critical entities. Moreover, the hallucinations in LLM’s outputs result in insufficient accuracy. To address these limitations, we propose a zero-shot, self-improving NER method for CTI via knowledge injection. The framework consists of four modules: a domain knowledge extractor, a reliable data annotator, a high-consistency annotation filter, and a self-retrieval reasoner. The domain knowledge extractor enhances LLM comprehension of specialized threat intelligence, while the others work in a multi-stage reasoning process to mitigate hallucinations by generating, filtering, and reasoning upon high-consistency data. These modules collaborate to improve the model’s entity recognition ability through continuous in-context learning. Experimental results show that under strict zero-shot conditions, the proposed method achieves F1 scores of 67.7%, 61.41%, 74.56%, and 65.83% on the LLM-TIKG, APT-NER, LADDER, and CDTier datasets, respectively. This represents an improvement of 7.66% over the average F1 score of other baseline methods, demonstrating superior adaptability in low-resource security scenarios.

Cell-free massive multiple-input multiple-output symbiotic radio (CF-SR) has recently been introduced as a promising solution for the Internet of Things (IoT), offering cost advantages and more uniform coverage performance for user devices. However, most previous studies assume perfect hardware, which is impractical in IoT systems. In this article, we investigate the uplink performance of CF-SR systems in the presence of hardware impairments (HWIs). We adopt two novel schemes for hybrid combining, namely hybrid maximum ratio (HMR) and hybrid local minimum mean square error (HL-MMSE), both of which effectively enhance the spectral efficiency (SE) of the backscattering link. We derive closed-form expressions for the achievable SE of both conventional MR and HMR combining schemes, taking into account the imperfect channel state information (CSI) and HWIs. The simulation results that both the direct and backscattering links are primarily limited by HWIs including multiplicative and additive distortions from the device side. We apply the differential evolution (DE) algorithm for power control to maximize the minimum SE, and the results show that the DE algorithm improves the minimum SE by approximately 52%, avoiding further degradation of the SE of the weakest device due to the impact of HWIs.

Reconfigurable intelligent surface (RIS)-aided cell-free massive multiple-input multiple-output (CF-mMIMO) technology has tremendous potential to revolutionize wireless communications by dynamically adapting wireless channels to boost average rate and energy efficiency (EE) of Internet of Things (IoT) networks for meeting the specifications of ultra-reliable and low-latency communications (URLLC). In this paper, we study the downlink harvested energy (HE), uplink rate, and total EE of the wireless-powered RIS-aided CF-mMIMO communication system with hardware impairments under finite blocklength. IoT devices harvest energy from the energy signals transmitted from access points (APs) during the downlink and use it for the uplink pilot and data transmission. Specifically, based on the unique characteristics of the channel fading model and the RIS deployment location, we propose a novel RIS phase shift design according to the line-of-sight (LoS) components of channels. Furthermore, we derive the average HE and uplink rate in closed form with a two-layer decoding method. We also validate the effectiveness of the proposed RIS phase shift design and the derived closed-form expressions by Monte Carlo simulations. Moreover, it is interesting to find that local minimum mean squared error (L-MMSE) combining is recommended to meet the requirements of URLLC, including communication reliability and delay. More notably, the numerical results show that the RIS-aided system with impaired hardware exhibits even superior performance, compared to the system with ideal hardware and more APs but lacking the assistance of RISs.

Cell-free symbiotic radio (CF-SR) is a promising technology to meet the demands of good quality-of-service and spectrum-efficient communications. However, the introduction of SR brings additional interference terms, which can seriously degrade the performance of the CF-SR systems. To suppress the interference, we adopt a rate-splitting (RS) transmission scheme to CF-SR. In this paper, we derive downlink spectral efficiency (SE) expressions of the CF-SR system with RS. Furthermore, in a conventional two-phase (TP) channel estimation scheme, the direct link causes heavy interference to the backscatter link, consequently diminishing the accuracy of the backscatter-link channel estimation. To this end, we propose a collaborative cancellation (CC) channel estimation scheme, which can eliminate the interference from the direct link and thus improve the accuracy of the backscatter-link channel estimation. Moreover, we derive the novel closed-form SE expressions under the CC channel estimation scheme using maximum ratio (MR) precoding. Simulation results show that the normalized mean square error (NMSE) of the CC channel estimation is consistently better than the one of the TP channel estimation, both on the direct and backscatter links. Furthermore, the advantages of the CC channel estimation scheme on the backscatter link can be further amplified in scenarios with a sufficient number of pilots. In addition, simulation results demonstrate that both the CC channel estimation scheme and the RS transmission scheme can provide significant improvements.

Reconfigurable intelligent surface (RIS) has the potential to revolutionize wireless communications by dynamically controlling wireless channels to boost spectral efficiency (SE) and energy efficiency (EE), towards meeting the advanced specifications of Internet of Things (IoT) networks. In this context, we study the downlink harvested energy (HE), uplink SE and total EE of the RIS-aided cell-free massive multiple-input multiple-output (CF-mMIMO) system with wireless power transfer (WPT) technology. IoT devices harvest energy from the energy signals transmitted from access points (APs) during the downlink and use it for the uplink pilot and data transmission. Based on the unique characteristics of the channel fading model and the RIS deployment location, we put forward a novel RIS phase shift design scheme according to the line-of-sight (LoS) components of channels and verify its effectiveness. Furthermore, we derive the average HE and uplink SE in closed form with a two-layer decoding method (i.e., the maximal ratio combining (MRC) at APs is called first-layer decoding and the large-scale fading decoding (LSFD) at CPU is called second-layer decoding.) under the assumptions of both perfect and imperfect statistical channel state information (CSI). The results verify the derived closed-form expressions by Monte-Carlo simulations. Increasing the number of RIS elements further improves the uplink SE and total EE with the two-layer decoding. Since the statistical CSI is unknown in practical scenarios, we propose an acquisition method for the statistical CSI applicable to this system. Simulation results validate the efficiency of the proposed statistical CSI acquisition method. Furthermore, it is interesting to find that better statistical CSI estimation can be achieved with more coherent blocks of pilot.

Cell-free massive multiple-input multiple-output symbiotic radio (CF-mMIMO-SR) is a promising technology to address the requirements of high-rate and spectrum-efficient communication for the Internet of Things (IoT). However, in the conventional CF-mMIMO-SR system aided by backscatter devices (BDs), the backscatter link is impacted by double fading without any supplementary compensation, resulting in significantly low spectral efficiency (SE) on the backscatter link. To address this issue, we propose the usage of reconfigurable intelligent surfaces (RISs) instead of BD for symbol-level reflection on the backscatter link, leading to a novel RIS-aided CF-mMIMO-SR (RIS-CF-SR) system. In this paper, we conduct a comprehensive analysis of the RIS-CF-SR system considering different levels of cooperation among the access points (APs). Specifically, we analyze the uplink SEs of four different implementations with arbitrary linear processing on both the direct and backscatter links. Moreover, we investigate different signal cancellation schemes based on full or local channel state information (CSI) to improve the SE of the backscatter link. Through the simulation results, we find that RISs can significantly improve the SE of the backscatter link due to the large number of reflection elements, whereas additional appropriate signal processing schemes are required for the direct link. More specifically, from Level 1 to Level 3, RIS-CF-SR does not have significant advantage in SE over BD-CF-SR on the direct link. At Level 4, RIS-CF-SR can outperform BD-CF-SR on the direct link with the MMSE combining scheme.

This article considers the uplink of a cell-free massive multiple-input–multiple-output (MIMO) symbiotic radio (CF-mMIMO-SR) system. Conventional combining schemes cannot be used directly to detect the signal of the direct link due to its heavy suppression for the backscatter link. To this end, we propose two hybrid combining schemes, including the hybrid maximum ratio (MR) and hybrid local minimum mean square error (L-MMSE) combining schemes, which take into account the superposition of the two local channel estimation vectors at the access points (APs). When the number of APs goes to infinite, the asymptotic spectral efficiency (SE) of CF-mMIMO-SR with different combining schemes is analyzed. We prove that the conventional combining schemes tend to cause the effective signal of the backscatter link to disappear while the hybrid combining schemes can obtain good performance on the backscatter link. Meanwhile, the performance gap between the two on the direct link is small. In addition, we derive closed-form expressions with the conventional MR combining and hybrid MR combining schemes over independent Rayleigh fading channels. Moreover, we derive the achievable uplink SE expressions with an effective signal-to-interference-and-noise ratio (SINR) for a finite number of antennas. Simulation results verify our theoretical analysis and demonstrate that hybrid combining schemes perform much better on the backscatter link than conventional combining schemes. Specifically, compared to the hybrid MR scheme, the hybrid L-MMSE scheme offers a huge improvement in 95% likely SE, and has negligible performance loss on the direct link.

Remote sensing image scene classification (RSISC) is important for analyzing and interpreting remote sensing images (RSIs). However, the intraclass difference and interclass similarity problems caused by complex backgrounds and variable scales bring great challenges for the effective classification of RSI scenes. In this letter, we solve the above problem mainly from two aspects. First, a multilayer feature fusion (MLFF) module is proposed to improve the classification performance of the networks by adaptively fusing multilayer semantic information. Then, to learn more discriminative local fine-grained, a regional growth-based key area (RGKA) fine location algorithm is proposed to accurately obtain local key areas. Finally, a two-branch network is utilized to complete the classification. Experiments are conducted on three publicly available datasets. The experimental results show that the proposed method outperforms most state-of-the-art methods and can considerably improve the accuracy of RSISC.

Many people use mobile devices as the basic sensing units since they are cheap and convenient. Several incentive mechanisms have been proposed in past literature to incentivize worker participation in such sensing units. However, existing auction-based privacy-preserving incentive mechanisms only consider asymptotically truthful property apart from exactly dominant strategy truthful property while maximizing crowdsourcers' revenue. Past research has shown that a single mechanism model is difficult to solve this problem, so we designed a novel hybrid mechanism with differential privacy to strike the balance between utility property and truthful bidding property. The hybrid mechanism, denoted as PQHM, consists of price-privacy auction-based mechanism (denoted as PPAM) and quality-privacy auction-based mechanism (denoted as QPAM). PPAM can get great social utility and QPAM can get truthful bidding property. We could randomize between running the two mechanisms according to a probability distribution (defined as λ) to guarantee the reasonable utility for the crowdsourcers, and give each bidder a strict positive incentive to report truthfully. This paper found that the hybrid incentive mechanism has the characteristics of differential private, exactly truthful, individual rationality, reasonable platform profitability, and calculation efficiency. Furthermore, the selection probability λ of PQHM is linearly controllable.