The massive growth of data has brought vigorous vitality to the Internet of Things (IoT). It has also brought new challenges, such as confidentiality privacy protection and redundant data transmission. Concerning this regard, data aggregation serves as an efficient technique to minimize the transmission frequency among massive objects in smart grid (SG). By aggregating a large amount of encrypted data from distributed IoT devices, the proposed framework enables efficient and privacy-preserving analytics at the edge. With the advent of the postquantum era, a good aggregation scheme must provide quantum resistance while ensuring the secure aggregation of ciphertext power data. However, the excessive overhead limits anti-quantum algorithms from being widely used in SG, where resource devices are limited. Therefore, it is an important part of the current private data security aggregation technology to find a low cost and lightweight inverse quantum algorithm to achieve user data security aggregation. In this article, we propose an improved NTRU-based cryptosystem with multidimensional coding, referred to as multidimensional coding NTRU (MC-NTRU), and use the lattice batch signature technique, which improves the efficiency of the scheme while satisfying the anti-quantum attack. Based on these, we design the multidimensional auditable lattice-based privacy-preserving data aggregation scheme (MA-PPDA) for privacy data on resource-limited IoT devices, such as SGs. In addition to this, the scheme achieves fault tolerance (FT) of the scheme by adding zeros and random numbers to the user data. The comparative study against the existing approaches demonstrates that the proposed scheme not only adheres to critical security aspects, including user privacy, data confidentiality, integrity, and authenticity, but also decreases both communication and computational burdens on the system. This makes our scheme particularly apt for IoT environments characterized by constrained device resources.

With the rapid development of cloud-based data sharing technologies, enterprises and organizations tend to outsource their local data to cloud servers. They adopt searchable encryption (SE) techniques to access and search encrypted data. However, most existing SE schemes use static ranking strategies based on query-index similarity. These strategies fail to capture users' personalized retrieval preferences and often result in suboptimal search performance. In this article, we present a privacy-preserving data sharing framework with personalized encrypted retrieval (PP-PER) that combines SE technology with federated learning. PP-PER trains user interest models locally on user devices by utilizing historical query behavior. Only encrypted model parameters are uploaded for aggregation, which avoids the centralized collection of users' private data. In addition, we design an attention-based user query update algorithm. The learned personalized features are integrated into the ciphertext query process. This design enables personalized ranking results and improves the user retrieval experience. Furthermore, PP-PER combines matrix factorization with ciphertext-policy attribute-based encryption (CP-ABE). This mechanism ensures secure document key distribution and supports fine-grained access control. Finally, we formalize the security model under a practical threat and leakage setting and provide a theoretical analysis of the proposed scheme. Experimental results on real-world datasets further validated its practicality and effectiveness.

Federated Learning (FL) has demonstrated considerable promise for wind power forecasting through data-driven approaches, yet it faces critical challenges including data scarcity, spatiotemporal heterogeneity, and privacy vulnerabilities inherent in centralized training. Conventional FL methods often neglect variability in client contributions, resulting in compromised forecasting accuracy. To overcome these limitations, this paper introduces a novel multi-weight aggregation federated learning framework that innovatively employs Particle Swarm Optimization (PSO) to dynamically determine aggregation weights. At the client level, a hybrid CNN-BiLSTM-ATT model is deployed as the core local architecture to effectively capture complex spatiotemporal dependencies for each individual wind turbine. The core of our approach, the federated multi-weight aggregation (FedMWA) mechanism, leverages PSO to optimize aggregation weights based on multi-factor contributions from heterogeneous clients, enabling more effective integration of distributed knowledge while preserving data privacy. This PSO-driven aggregation strategy represents a significant departure from traditional static weighting schemes, substantially enhancing the global model’s representational capability. Experimental results demonstrate significant performance gains, with average reductions of 4.45% in NMAE and 3.66% in NRMSE compared to baseline methods. The framework maintains robust performance across diverse power outputs and fluctuation scenarios while ensuring data privacy through localized learning.

Enforcing commitments under time-dependent constraints is a fundamental challenge in distributed and blockchain-based systems operating without centralized trust. Existing approaches often face a critical trade-off: they incur linear verification costs, rely on trusted parties, or fail to prevent strategic abort behavior by malicious committers. To address this problem, we propose a Verifiable Timed-Release Adaptor Signature (VTRAS) scheme. VTRAS integrates time-lock puzzles with Non-Interactive Zero-Knowledge (NIZK) proofs into the adaptor signature framework, enabling a witness to be cryptographically bound to a hard relation for a predetermined duration while supporting verifiable recovery. This design allows any committed puzzle to be efficiently verified as well-formed and correctly bound to a witness before acceptance, thereby preventing abort-based and invalid-puzzle attacks. As a concrete application, we construct a decentralized auction protocol that leverages VTRAS to enforce post-auction payments without trusted intermediaries. The protocol enables the seller to complete payment extraction after a time delay if the buyer defaults, ensuring fairness in asynchronous and adversarial settings. We formally analyze the correctness, soundness, and privacy of VTRAS, and evaluate its practical performance on the Ethereum Virtual Machine (EVM). Experimental results demonstrate that VTRAS achieves constant on-chain verification cost, reducing gas consumption by over 90% compared to threshold-based alternatives. These results establish VTRAS as an economically viable security primitive for time-critical decentralized applications.

We consider the problem of optimizing a complex function f(・) given a sample consisting of M input-output pairs (argument, function value), S = {(xn,yn)}_N=1^M , where x ∈ \mathbb{R}^{d} and y ∈ \mathbb{R} . We assume that the sample is available only in a distributed manner at K entities, where each entity controls a sample part Sk and S = ∪_k=1^K Sk. The problem is addressed employing a machine learning approach. Our goal is to generate a surrogate of the function based on the given sample, and to perform the optimization on the surrogate function. Due to the distributed nature of the sample we employ Federated Machine Learning (FML) to generate the surrogate function by joint efforts of all K entities that possess the sample parts. We call the entities involved in FML clients, and we assume that certain clients participate in FML in a malicious manner by poisoning training data or the model parameters they generate locally. Systematic experiments have been performed to study the characteristics of the surrogate function and its capabilities to provide reliable analysis of the original function depending on the sample available for training, and impacts of the training data poisoning, as well as locally generated model parameters.

Adaptor signature can guarantee the atomicity of cross-chain transactions without relying on trusted third parties, but most adaptor signature algorithms fail to protect the privacy of user identities. Moreover, most of these schemes are not based on domestic cryptographic algorithms, resulting in poor autonomy and controllability. To address these issues, a ring adaptor signature scheme based on the SM2 digital signature algorithm was proposed. The signer first generates a pre-signature using a ring signature; after the pre-signature is verified, it is combined with a secret value to adapt to a complete signature; any user with the pre-signature and its complete signature can extract the secret value. This ring adaptor signature scheme uses the properties of ring signatures to protect the privacy of the signer's identity in the pre-signature stage, while using adaptor signatures to ensure the atomicity of transactions, and combines the excellent characteristics of both identity privacy protection and transaction atomicity. Its security can be reduced to the discrete logarithm problem. Under the random oracle model, it was proved that the proposed scheme satisfies pre-signature adaptability, pre-signature correctness, existential unforgeability under chosen message attacks, witness extractability, and anonymity. Theoretical analysis shows that the time and communication overhead of this ring adaptor signature scheme is comparable to that of existing ring adaptor signature schemes. This scheme greatly expands the application of domestic cryptographic signature algorithms in the field of ring adaptor signatures, which is especially pertinent to the field of privacy protection in cross-chain payments.

Federated learning (FL) enables the collaborative model training while preserving data privacy, making it particularly attractive for large-scale Internet of Things (IoT) systems. However, in practical deployments, data collected by distributed clients is often nonindependent and identically distributed (Non-IID), which amplifies the vulnerability of FL to poisoning attacks. Among them, label-flipping attacks (LFA) are especially stealthy, as they can induce targeted misclassification without noticeably affecting overall accuracy, posing serious risks to safety-critical IoT applications. In this article, we propose NeuDFL, a lightweight and robust defense framework against LFA under Non-IID settings. Unlike many existing defenses that primarily rely on the gradient analysis over the full-model update space or auxiliary clean datasets, NeuDFL exploits lightweight classwise parameter statistics extracted from the final fully connected layer (FCL). By leveraging the cumulative and task-aligned nature of model parameters, NeuDFL enables reliable identification of attacked classes and filters malicious clients via an adaptive statistical threshold, improving robustness to data heterogeneity while incurring low computational overhead. Extensive experiments on multiple datasets demonstrate that NeuDFL offers an effective and efficient defense against LFA, providing a robust solution for FL in complex real-world environments.

Short-term load forecasting is essential for the construction of smart grids and optimizing power dispatch. This paper innovatively proposes a federated learning framework that fuses dynamic Fisher personalization with adaptive constraints to address the three major challenges faced by traditional methods: the contradiction between privacy and utility, the rigidity of personalization strategy, and the difficulty of convergence due to differential privacy. This scheme intelligently screens parameters through layer-by-layer Fisher information and combines it with adaptive constraints to enhance noise resistance. It significantly improves prediction performance while guaranteeing privacy. Experiments show that, compared to the FedAvg+DP benchmark method, this scheme reduces RMSE, MAE, MSE, and MAPE by 38.72

We present Mask, an efficient and tunable algorithm for hiding volume patterns in multimaps. Volume-pattern leakage, referring to the observable size of data returned by a query, enables adversaries to infer sensitive dataset information, posing severe privacy risks in multimap scenarios. Designed to address this issue, Mask focuses on balancing storage/query overhead and privacy (a key tradeoff in this field), allowing users to define parameters for random data distribution across buckets to realize fine-grained control over storage and query performance, integrating Bloom filters with a bounded cache for efficient indexing (optimizing performance under skewed query workloads); extended to MaskIO for reduced client-side storage, it obfuscates query indexes and uploads them to the server, achieving constant-bounded client storage overhead, and experiments show that Mask outperforms the bucket-based peer Veil with 2 & times; - 4 & times; higher performance and a lower stash ratio (SR).

The rapid development of electric vehicle (EV) charging transactions has caused serious privacy and security issues, including sensitive data leakage, tampering risks and lack of audit mechanisms. Traditional privacy protection technologies (such as encrypted transmission and access control) have insufficient scalability, single-point vulnerability and efficiency bottlenecks in high-concurrency scenarios. To address these issues, this paper proposes a privacy protection scheme for energy transaction data based on blockchain and homomorphic encryption. The proposed scheme integrates data segmentation and homomorphic encryption to enable decentralized storage and secure encrypted computations, ensuring end-to-end privacy protection throughout the transaction lifecycle. To enhance efficiency, edge computing is introduced, optimizing data processing and system responsiveness. Furthermore, blockchain technology is leveraged to provide immutable transaction records, ensuring transparency, auditability, and compliance. Experimental results show that the proposed scheme effectively solves the problems of data leakage and tampering, improves the security and privacy protection capabilities of transaction data, and ensures the reliability and traceability of transactions.

As electric vehicle (EV) utilization in energy trading expands, protecting user identity privacy during transactions has become a critical area of research. Current privacy protection schemes face significant challenges, including the risks of identity leakage, insufficient protection against double signatures, and low verification efficiency. To address these challenges, this document proposes a lightweight identity privacy protection scheme based on blockchain technology from the consortium. The scheme introduces an improved linkable ring signature algorithm, which guarantees identity anonymity while effectively preventing double signature attacks. In addition, this paper develops a batch aggregation signature and verification algorithm designed to significantly improve verification efficiency in highly concurrent environments. Theoretical analysis and simulation experiments demonstrate that the proposed scheme outperforms existing solutions in security, computational efficiency, and communication overhead. Compared to existing solutions, this approach provides substantial improvements, making it a promising approach to secure user privacy within the rapidly evolving domain of EV energy trading.

The PBFT (Practical Byzantine Fault Tolerance) consensus algorithm was confronted with several challenges in the blockchain sharding environment, including high communication complexity, random leader election, node insecurity, and insufficient participation incentives. To address these issues, a reputation-driven blockchain sharding consensus and incentive mechanism was proposed. Reputation mechanisms were utilized to evaluate and incentivize node behavior, thereby enhancing system trust and promoting active participation. Aggregated signature technology was employed to merge consensus node signatures, and node voting weights were determined based on reputation assessments, which reduced communication overhead and improved consensus equality and security. Additionally, a deposit and incentive mechanism was introduced under reputation management, where initial entry standards were set and gradually increased over time according to reputation values. Differentiated incentive models, based on node roles, were designed to balance stable income and dynamic rewards, thereby continuously improving network stability and security. Simulation results demonstrated that the proposed R-PBFT consensus algorithm outperformed the traditional PBFT, maintaining lower latency and higher throughput, which effectively enhanced system performance. Furthermore, compared to proportional incentive allocation based solely on reputation, the proposed incentive mechanism achieves higher consensus participation rates.

This paper presents a new class of erasure-correcting codes (ECCs) aimed at enhancing cryptographic security of certain encryption schemes. The proposed ECCs employ integer arithmetic to encode and decode data bits and can correct all data bytes, each affected by exactly two erasures.. In the enhanced encryption scheme, the ciphertext produced by the initial encryption undergoes further processing. The enhancement leverages specific fragmentation, the proposed ECCs and a simulated noisy channel. For legitimate users, the simulated noisy channel functions as a binary erasure channel, while for an attacker without the secret key, it acts as a channel with random deletions. Security notation and evaluation follow the traditional approach, assessing the attacker’s advantage in distinguishing between two ciphertexts versus random guessing. This evaluation employs dedicated analysis based on information-theoretic findings on the capacity of certain deletion channels and is supported by illustrative numerical examples.

In the digital age, the importance of data sharing has significantly increased as it accelerates the release of value from data elements. Nevertheless, data confronts substantial security threats, including potential leakage during outsourcing and sharing procedures. To ensure the security of shared data, plaintext data is often replaced by encrypted data for sharing, and searchable encryption algorithms are used to improve the efficiency of sharing. However, due to the inherent limitations of searchable encryption schemes, existing secure data-sharing approaches frequently encounter inefficient search capabilities and privacy violations. This paper proposes a secure data-sharing scheme for encrypted data that integrates blockchain with an overlapping clustering technique to tackle these challenges. The proposed scheme combines blockchain with an attribute-based searchable encryption method to guarantee data transparency, trustworthiness, and confidentiality. Furthermore, the scheme significantly enhances search efficiency and accuracy by incorporating overlapping clustering and keyword relevance-based ranking strategies. Experimental results show that this scheme effectively enhances search efficiency and provides robust privacy protection.

The growing adoption of IoT applications underscores the need for advanced data fusion and information acquisition techniques, driving demand for secure, privacy-preserving querying of integrated IoT data. Existing schemes like searchable encryption are practical but leak access patterns, while leakage-free methods using Oblivious RAM or cryptographic techniques incur significant resource overhead. In this paper, we propose PQBL, a framework for privacy-preserving, trusted data integration and search, leveraging distributed trust against malicious attackers. Our query scheme combines function secret sharing and blockchain to enable efficient, privacy-preserving searches on encrypted IoT data. To improve search efficiency, we introduce a compressed RAMBO Bloom Filter for keyword trapdoors. Formal security analysis shows that PQBL leaks no search patterns and is secure against Privacy under Selective Chosen-Plaintext Attacks. Extensive experiments on the PQBL prototype validate its effectiveness and efficiency.

Federated Learning is a widely used distributed machine learning framework that allows clients to collaboratively train a global model by uploading local gradients while keeping data stored locally, thus protecting user privacy. However, attackers can still infer local data from gradients. Recently, integrating differential privacy into FL has become a popular approach to ensure strong privacy guarantees. This paper proposes a Decentralized Federated Learning Framework with Enhanced Privacy and Optimized Fairness (DFL-EPOF). First, noise is added to local parameters before uploading, and a local differential privacy mechanism ensures data privacy. An adaptive privacy budget allocation strategy, based on data sensitivity, dynamically controls noise levels to balance privacy protection and model accuracy. Second, a weighted aggregation method based on clients' data volume, trustworthiness, and participation frequency is used to optimize fairness, ensuring balanced contributions. Finally, a decentralized blockchain-based architecture is implemented to enhance transparency and immutability, ensuring reliable model updates and data transmission. Experimental results show that DFL-EPOF improves privacy protection, fairness, and system robustness, balancing privacy and accuracy effectively.

Secure neural network inference provides privacy guarantees for both the client and the server, and is an integral approach in Machine Learning as a Service Setting (MLaaS). However, the multilayer structure in the neural network introduces frequent activation function calculations, which causes large overhead. Most of the prior secure inference systems focused on designing cryptographic protocols to improve computational efficiency, but high computing and communication overhead are still bottlenecks in practicality. In this work, we refocus on the potential of shallow neural networks and propose a model with only one trainable layer to reduce the required computation. Our main contributions are in three-fold: 1) introduce training-free weights and formally prove their contribution in the model expressivity; 2) design the Self Enhanced Module that is more suitable for shallow models as an alternative for the activation function; and 3) propose a linear layer with multiscale and normalization property, named Nested & Norm Conv. We conduct extensive experiments on visual datasets and the results demonstrate the proposed single-layer trainable model holds promise as a viable platform for secure inference in practical applications.

Due to the rapid emergence of malware and its greater harm, the successful execution of malware often brings incalculable losses. Consequently, the detection of malware has become increasingly crucial. The sequence of API calls in software embodies substantial behavioral information, offering significant advantages in the identification of malicious activities. Meanwhile, the capability of automatic feature extraction by deep learning can better mine the features of API call sequences. In the current research, API features remain underutilized, resulting in suboptimal accuracy in API detection. In this paper, we propose a deep-learning-based method for detecting malware using API call sequences. This method transforms the API call sequence into a grayscale image and performs classification in conjunction with sequence features. By leveraging a range of deep-learning algorithms, we extract diverse behavioral information from software, encompassing semantic details, time-series information, API call frequency data, and more. Additionally, we introduce a specialized neural network framework and assess the impact of pixel size on classification effectiveness during the grayscale image-mapping process. The experimental results show that the accuracy of our classification method is as high as 99%. Compared with other malware-detection techniques, especially those based on API call sequences, our method maps API call sequences to gray image analysis and has higher detection accuracy.

The e-vote is regarded as a way to express the opinion that the voters ask for. Actually, the e-vote could be applied wildly like questionnaire, survey and feedback. Moreover, the coexistences of efficiency and security as well as transparency and privacy could be considered as building blocks in the e-vote system. The blockchain could provide a public access board to reduce the storage costs for the field consisted of the vote group manager (GM) with its vote assistants (VA). Particularly the $k$ -times anonymous authentication ( $k$ -TAA) could also be a practical approach to preserve voters' privacy and reduce the computation costs during the vote process. However, the e-vote scheme with pure $k$ -TAA strategy could damage either the supervision of voting or the efficiency and consistency of authentication process. What's more, the impacts of dishonest voters couldn't be stopped until the vote end. To tackle these problems, we apply the accumulator technology to add or revoke the voters at any time and extend the framework of $k$ -TAA with the update process for the e-vote on blockchain for supervision ( $ k$ -TEVS). In our scheme, the voter updates his membership witness and proves the fact that he is still a valid member with respective VA under the latest accumulator value. What's more, this witness update operation is not contained in the authentication process, which means that the authentication process is still constant and efficient. Moreover, our add or delete update process with signature of knowledge needs only one pairing operation. For the security, we prove that the relaxed anonymity still holds in the $ k$ -TEVS framework. Finally, We implement $ k$ -TEVS scheme, the Emura's work [1] and the Huang's work [2] for comparison. Then we make time cost and communication cost experiments, which present the feasibility and practicality of this scheme

As the core application of blockchain technology, smart contracts have been widely used in many fields such as finance, supply chain, and copyright management. Smart contracts are prone to various vulnerabilities that attackers can exploit to steal or freeze funds. Traditional vulnerability detection methods rely heavily on complex rules defined by experts, which are difficult to adapt to the explosion of smart contracts. Some recent studies of neural network-based vulnerability detection methods rely on contract source code, and the accuracy of bytecode-level vulnerability detection methods is low. To overcome the limitations of existing methods, we propose CV-IRTL, a new method for smart contract vulnerability detection. Specifically, CV-IRTL designs a vulnerability detection framework for smart contracts based on inverted residual network architecture and transfer learning. In particular, CV-IRTL enables vulnerability detection at the bytecode level, simplifies data preprocessing, utilizes transfer learning to better capture vulnerability characteristics and effectively address dataset imbalances. We have extensively tested CV-IRTL on a dataset containing six vulnerabilities. The experimental results show that the macro average F1-score is 90.75%, and the overall false positive rate is 9.6%, which is better than representative methods in performance.