With the rapid development of 6G, RF-based wireless sensing technology is evolving into a core pillar of the ISAC framework, transitioning from single-target, static scenarios to multi-target and collaborative sensing. However, constrained by inherent vulnerabilities such as channel openness, wireless sensing faces severe security threats. This paper systematically reviewed the research progress in wireless sensing security over the past decade at home and abroad, analyzing existing work from three dimensions: technological evolution, threat mechanisms, and defense strategies. In terms of the evolution of wireless sensing technology, the development history of sensing platforms, sensing features and sensing methods was sorted out. In terms of threat mechanisms, this paper conducted an in-depth analysis of typical attack paradigms including privacy leakage, adversarial attacks, and channel manipulation, revealing multi-level attack entry points ranging from signal acquisition and feature reconstruction to model interference. From the perspective of defense strategies, this paper summarized the development trend from passive defense to active detection, and from single-mechanism protection to cross-layer collaborative defense. Finally, this paper prospected the future development trends from the aspects of 6G ISAC security, multimodal sensing security, and lightweight trusted sensing, aiming to provide a theoretical reference for the construction of secure and reliable wireless sensing systems.

With the accelerated advancement of the digital transformation of power systems, the number of edge devices has shown exponential growth trend. However, the container images and firmware of power edge devices are facing increasingly severe security threats, and these vulnerabilities pose great hidden risks to the safe and stable operation of power systems. Therefore, how to effectively evaluate the vulnerabilities of container images and firmware and detect their security loopholes in advance is an urgent issue to be solved in current power system security. Existing studies have not yet formed an effective paradigm compatible with the vulnerability assessment of container images and firmware for power edge devices, making it difficult to establish unified evaluation standards. In view of this, this paper proposed a security assessment method of vulnerability indicators for container images and firmware of power edge devices, constructed a three-level vulnerability indicator system of “basic-image-firmware”, which fully covered the common risks and exclusive characteristics of container images and firmware, and combined the Bayesian optimization method to build an evaluation method based on the XGBoost model. To verify the effectiveness of the proposed indicator system and evaluation method, this paper collected and aggregated 295 experimental samples of power firmware and container images for testing. The experimental results show that the accuracy rate of the proposed method reaches 96%, indicating that this method has high reliability and effectiveness in practical applications.

With the widespread adoption of cloud computing and data-sharing services, large volumes of sensitive data are outsourced to cloud platforms, exposing such data to the risks of unauthorized access and privacy leakage1. However, existing ciphertext search schemes struggle to simultaneously hide an access policy and retrieve keywords efficiently. To address this challenge, this paper proposed an efficient ciphertext search scheme for attribute-based policy hiding. The scheme ensured access policy hiding while maintaining high keyword retrieval efficiency. The scheme constructed a policy-hiding mechanism by embedding random vectors into attribute ciphertexts. This mechanism effectively protected user access intentions and sensitive data. This paper designed a trapdoor aggregation algorithm by compressing multiple matching operations into a single linear pairing. The algorithm achieved constant-level complexity for rapid retrieval. Furthermore, this paper designed a cloud-edge-end collaborative computing algorithm by outsourcing the major computational overhead of ciphertext decryption. This algorithm significantly reduced the computational burden on an end user and improved local decryption efficiency. Scheme analysis demonstrates that the proposed scheme guarantees the correctness of both search and decryption. Under standard hardness assumptions, it achieves policy hiding and ensures semantic security, thereby resisting sensitive information leakage and inference attacks. Experimental results further show that the proposed scheme attains a lower computational overhead and faster response times across various attribute dimensions.

With the rapid development of quantum computing technologies, traditional digital signature schemes based on large integer factorization and discrete logarithm problems are facing potential security risks. Code-Based cryptographic systems, whose security relies on computationally hard problems such as syndrome decoding, are widely regarded as an important approach toward post-quantum cryptography. However, existing code-based signature schemes still provide relatively limited support for deniable semantics.To address this issue, this paper proposed a deniable ring signature scheme based on the HQC code structure. The proposed construction systematically adapted Stern-Like Σ protocols, OR composition, and the Fiat-Shamir transformation to the linear consistency relation induced by HQC public keys, and used this relation as the underlying statement of the zero-knowledge proof. On this basis, a signature-context-based tag binding mechanism was further introduced, through which signer confirmation and non-signer disavowal algorithms were incorporated as independent layers into the ring signature structure, thereby achieving complete deniable semantics. Finally, under the random oracle model, the unforgeability, anonymity, traceability, and non-frameability of the proposed scheme were analyzed, and its computational cost, communication overhead, and functional properties were evaluated. The analysis results show that the proposed scheme can uniformly achieve ring anonymity and deniable semantics under security assumptions based on code-based hard problems. However, its signature length and computational cost are affected by the number of proof rounds and the response encoding length, making it more suitable for anonymous authentication scenarios with small ring sizes where post-quantum security and subsequent responsibility clarification are prioritized.

With the increasing frequency and sophistication of cyberattacks, cyber threat intelligence (CTI) has gradually become an essential instrument for threat mitigation. Nevertheless, existing CTI is primarily disseminated through online platforms in standardized formats, whereas a substantial volume of unstructured intelligence remains scattered across blogs, forums, and personal websites without a unified quality assessment framework. To enable automated evaluation of large-scale fragmented intelligence, this paper proposed a threat intelligence quality assessment method based on the dynamic fusion of subjective and objective indicators. Specifically, twelve indicators were designed from the perspectives of statistical characteristics and content analysis; temporal modeling was employed to compute composite quality scores for intelligence sources across different time points; an autoencoder integrated with a Transformer architecture was used to derive indicator weights in an unsupervised manner; and a temporal smoothing mechanism was introduced to suppress abrupt score variations caused by short-term fluctuations. Extensive experiments were conducted to evaluate indicator correlation, model performance, and practical effectiveness on a dataset of 810911 articles collected from 347 websites. The results demonstrate that the proposed indicators exhibit strong discriminative power, enabling multi-dimensional quality assessment, and that the evaluation outcomes achieve high consistency with actual intelligence quality.

At present, traditional requirement modeling relies on manual efforts, is inefficient, and prone to semantic gaps, while the quality of unified modeling language model generation by large language model (LLM) remains unstable. To address these issues, this study proposed a quality assessment framework for LLM-generated statecharts by combining a structured requirements modeling description with a state-diagram quality dimension function. The structured requirements modeling description used five core components: system overview, pattern hierarchy, state definitions, transition logic, and safety constraints, aiming to convert natural language requirements into structured semantic inputs for LLM to generate state diagrams. State-diagram quality dimension function constructed a formalized evaluation system for statecharts across four dimensions, including executability, syntactic quality, semantic quality, and practical quality. This framework overcomes the limitations of existing research in behavioral model quality assessment, which relies on single metrics and lacks systematic verification mechanisms. Experimental evaluation is conducted using two large models, GPT-4.0 and DeepSeek-R1, on 100 cross-domain requirements documents. Results demonstrate that GPT-4.0 achieves an average semantic coverage of 82%, while DeepSeek-R1 reaches 91%, confirming the applicability of LLM in requirements-driven modeling. The analysis further reveals distinct strengths and characteristics of different models across various quality metrics. This study provides a systematic methodology for evaluating LLM-generated models and offers practical references for model selection and optimization in real-world applications.

Federated learning (FL) as a privacy-preserving distributed machine learning paradigm has been widely applied across various fields. However, traditional single-server architectures face issues of excessive server communication overhead as user devices increase. Although multi-edge-server architectures alleviate the single-point load pressure, their limited coverage cannot meet the demands of large-scale FL for data diversity and model generalization capability. In recent years, the hierarchical federated learning (HFL) architecture has been introduced, which incorporates edge servers in different regions to perform intermediate aggregation while a cloud server handles global aggregation. This approach effectively reduces the cloud server’s load and is better suited for large-scale FL training across multiple regions. Nevertheless, HFL still faces challenges such as privacy leakage and poisoning attacks, especially in non-independent and identically distributed (non-IID) data scenarios. Moreover, the unreliability of edge servers in the intermediate layer is often overlooked, which can easily lead to collusion and crash risks, thereby threatening privacy security and affecting model training. To address these issues, this paper proposed a secure and reliable aggregation scheme based on the HFL architecture. Firstly, by using Shamir’s secret sharing and a masking mechanism, the scheme not only enhanced privacy protection during the training phase but also significantly improves the system’s tolerance to collusion and crash at edge servers. Secondly, the paper integrated Gaussian mixture model clustering and a scaled dot-product attention mechanism to design an anti-poisoning protocol under privacy protection. Theoretical analysis proves that the proposed scheme offers strong reliability in terms of privacy protection and resilience against edge server collusion and crash risks. Extensive experiments show that the scheme effectively resists poisoning attacks on various datasets and under non-IID settings, improving model accuracy by 5% to 20% compared to state-of-the-art methods.

Plaintext transmission of model updates in federated large language model fine-tuning may lead to privacy leakage, whereas full homomorphic encryption incurs substantial communication and computational overhead. To address these issues, this paper proposed a communication-efficient privacy-preserving federated learning method. It should be noted that data heterogeneity does not change the computational complexity of a single CKKS encryption or decryption operation. Its impact was mainly reflected in the inconsistency of sensitive parameter locations and LoRA rank structures identified by different clients, which enlarged the ciphertext index set to be aligned and the reconstruction matrix used for aggregation, thereby indirectly increasing encryption, communication, and matrix decomposition overhead. To address this issue, centered kernel alignment (CKA) was first employed to quantify the representational redundancy of model layers. Under a unified upper bound on the number of encrypted columns per layer, the encryption quotas of redundant layers were adaptively reduced to eliminate unnecessary ciphertexts. Then, a three-way hybrid negotiation mechanism based on order-preserving encryption was designed to integrate public high-frequency columns, globally highly sensitive columns, and client-specific long-tail columns within fixed layer-wise quotas, thereby achieving ciphertext index alignment. Finally, CKKS-based aggregation was performed on the selected sensitive subset, and randomized singular value decomposition (RSVD) was adopted to accelerate the low-rank reconstruction of heterogeneous LoRA updates. Experimental results demonstrate that the proposed method effectively reduces communication and computational overhead while maintaining model utility. Ablation experiments demonstrate that the CKA-based layer-wise quota mechanism effectively reduces communication overhead caused by redundant encryption, the three-way hybrid negotiation mechanism alleviates ciphertext expansion under heterogeneous data distributions, and the RSVD-based reconstruction method significantly reduces the computational overhead of heterogeneous LoRA aggregation. In terms of privacy protection, no successful inversion attacks are observed when the encryption ratio reaches 0.150% or higher under the DAGER attack samples and attack configuration used in this study.

Confidential Computing provides a securely isolated execution environment for model inference, but it uses the entire VM as the basic isolation unit, resulting in vulnerability to lateral movement in multi-tenant and multi-service environments. A compromised VM component can expose other modules to security risks. To address this, we proposed a fine-grained confidential computing framework for model inference. It leveraged AMD SEV-SNP to establish a trusted environment, implemented hardware-level security boundaries within confidential VMs, and integrated the NestedSGX mechanism to achieve intra-VM fine-grained isolation. Additionally, we explored a multi-enclave fine-grained isolation framework based on application partitioning. Through evaluation on multi-enclave applications, we tested the framework’s security benefits and performance overhead in different scenarios. In terms of security benefits, the multi-enclave fine-grained isolation framework can substantially reduce the trusted computing base and minimize the impact of failures on applications. In terms of performance, the overhead is within 16% for large-scale cross-domain transfers. This work provides a practical approach for building fine-grained trusted execution frameworks in VM-level confidential environments and derives actionable strategies for multi-enclave partitioning.

The lattice-based G+G digital signature is a novel digital signature scheme. It adopts Gaussian convolution technology to replace the flooding or rejection sampling methods commonly used in traditional lattice-based digital signatures, thereby offering a new approach for the design and implementation of lattice-based digital signatures with the Fiat-Shamir transform. However, software experiments have revealed that the G+G scheme can hardly be correctly implemented using its originally recommended parameters. Existing security analyses indicate that the G+G scheme either fails in practical implementation, or works normally but cannot resist ratio attacks based on the truncated Cauchy distribution. Therefore, secure and efficient parameter setting has become a crucial problem for the further research and application of the G+G scheme. By investigating the properties of the private key derivation matrix Σ involved in the G+G scheme, this paper pointed out that the validity of the private key should essentially be verified by judging the positive definiteness of matrix Σ, so as to ensure the G+G signature scheme can resist the ratio attack while being correctly implemented. On this basis, this paper presented a set of parameter adjustment methods for the G+G signature scheme and the recommended specific parameters after adjustment, and then analyzed the impact of parameter adjustment on public key size and signature length. Software experiments confirm the effectiveness of the proposed parameter adjustment methods and the implementation correctness of the G+G scheme with adjusted parameters.