As supply chain finance businesses expand, traditional risk assessment systems, which rely heavily on manual processes and static rule-based frameworks, are increasingly unable to keep up with the complexity and dynamism of modern risk patterns. This often leads to delayed responses and inefficiencies in risk management. To address key challenges such as difficulties in integrating heterogeneous data, low detection rates for hidden risks, and limited ability to capture dynamic risk patterns, this paper introduces a novel Risk-Aware Feature Network (RAFN) driven by an adaptive attention mechanism. The RAFN model is designed with a dual-channel architecture to process numerical and categorical data separately, employs gated linear units to dynamically merge heterogeneous data streams, and incorporates a multi-head attention mechanism with dynamic coefficients to focus on risk-sensitive features adaptively. Experiments conducted on both public and proprietary datasets show that RAFN outperforms mainstream algorithms, achieving a 1.73%-5.81% improvement in accuracy, recall, and F1-score, while maintaining a strong balance between specificity and recall. Furthermore, this study proposes a closed-loop risk management framework based on RAFN, which integrates “smart contract triggering, off-chain model evaluation, and on-chain consensus validation.” This approach offers an efficient technical solution to break down data silos and enhance the precision of risk identification in supply chain finance, paving the way for more effective and reliable risk control systems.

Over the course of more than a decade, blockchain technology has made significant advancements and found applications in various domains. Smart contract, as an integral component of blockchain technology, plays a pivotal role in ensuring the security and robustness of blockchain’s development and diverse applications. Currently, smart contract vulnerabilities have caused millions of dollars in economic losses. Due to the inherent immutability of blockchain technology, once smart contracts are deployed on the blockchain, effecting changes becomes a formidable task. Most of the vulnerability detection tools currently available employ traditional security technologies, which require high expertise and have unsatisfactory detection results. In recent years, deep learning technologies have emerged. Although they do not require extensive expert knowledge, they do require a large amount of labeled data for training. The biggest issue in this field is the lack of a large-scale, accurately annotated public dataset. Hence, we propose a method for detecting smart contract vulnerabilities by leveraging federated learning and BiLSTM, called FASCVD. Our approach not only utilizes federated learning technology to aggregate multiple small datasets while ensuring data privacy but also introduces a bidirectional information extraction technique based on BiLSTM, thereby significantly enhancing the accuracy of vulnerability detection. The experimental results show that our method has already surpassed the best existing methods in terms of accuracy, precision, recall, F1-score, and so on, with an accuracy rate of 95.04%.

ABSTRACT Background With the popularization and application of blockchain technology, smart contracts, as one of the underlying important technologies, have naturally attracted the attention of all parties. The vulnerabilities in smart contracts will lead to information leakage, asset theft, and other problems. Motivation Existing smart contract vulnerability detection tools mostly detect vulnerabilities through a set expert mode, relying more on professional knowledge. Traditional smart contract vulnerability detection methods based on deep learning rarely pay attention to syntactic information and semantic information at the same time, and their accuracy is low. Although the method based on graph neural network alleviates this problem to some extent, it suffers from the problem of too many nodes. Methods In this paper, we propose SNOW, an advanced method for detecting smart contract vulnerabilities, which leverages statement‐level joint feature graph and hybrid graph neural network to enhance the performance and efficacy of identifying smart contract vulnerabilities. Our proposed method consists of three parts. First, we generate a new graph representation called the Joint Feature Graph (JFG), which more effectively captures code information. Next, we introduce a hybrid graph neural network designed to extract JFG graph vectors more efficiently. Finally, we classify the graph vectors. Results We have conducted extensive experiments on two datasets and compared various existing methods. The results show that our method is superior to the current state‐of‐art method in many indexes such as accuracy and precision.

A vulnerability or error in a smart contract will lead to serious consequences including loss of assets and leakage of user privacy. Established smart contract vulnerability detection tools define vulnerabilities through symbolic execution, fuzz testing, and other methods requiring extremely specialized security knowledge. Even so, with the development of vulnerability exploitation techniques, vulnerability detection tools customized by experts cannot cope with the deformation of existing vulnerabilities or unknown vulnerabilities. The vulnerability detection based on machine learning developed in recent years studies vulnerabilities from different dimensions and designs corresponding models to achieve a high detection rate. However, these methods usually only focus on some features of smart contracts, or the model itself does not have universality. Experimental results on the publicly large-scale dataset SmartBugs-Wild demonstrate that this paper's method not only outperforms existing methods in several metrics, but also is scalable, general, and requires less domain knowledge, providing a new idea for the development of smart contract vulnerability detection.

Although there are some protection mechanisms in federated learning, its training process is still vulnerable to some powerful attacks, such as invisible backdoor attacks. Existing research work focuses more on how to prevent attacks in distributed training scenarios and improve the security of the FL training process, but it lacks consideration of utility and robustness, especially when the learning model of FL suffers from stealth backdoor attacks. This paper proposes an improved FL defense scheme IPCADP based on user‐level differential privacy and variational autoencoders technology. The scheme can control and protect the privacy attribute of the image and can also eliminate the triggers that exist in the poisoned image. The experimental results show that compared with some existing defense schemes, IPCADP can defend against invisible backdoor attacks and improve the classification accuracy of the main task, while mitigating the impact of attacks on model robustness and stability. To a certain extent, the balance and unity of security, utility, and robustness are realized.

Android Apps have become the most important mobile applications in the evolving mobile IoT systems, whose security and privacy are confronted with ever more challenges, since such mobile devices as smartphones involve too much personal privacy information. Meanwhile, the developers prefer to put core functions (e.g., encryption function and T9 search function) of Android applications in the native layer for execution efficiency. However, there are no automated security analysis tools to protect the security and privacy of the Android native layer, especially for those dynamically loaded third-party SO libraries. In order to solve the previous problem, which is confusing, we propose a novel and scalable system, called SoProtector, to prevent privacy from leaking via the analysis of data flow between the Java and native layers. For detection of the malicious function implanted in the SO libraries, SoProtector realizes a real-time engine. We derive the malware features via three steps: 1) present binary files in native family as a grayscale image; 2) with use of the ARM instructions set reversely obtain the code of the SO file and using Python to obtain the opcode sequence; and 3) each file is transformed as the form of assembly language by IDA Pro, which includes a gdl file as an accompaniment. Our experiment, which involved 3400 applications, demonstrates that SoProtector is able to detect more sinks, sources, and smudges. It effectively inspects and blocks at least 82% of the applications that are loading malicious third-party SO dynamically, and it has relatively low overhead in the meantime, compared to most of the existing static analysis tools (e.g., FlowDroid and AndroidLeaks).

With the wide deployment of wireless sensor networks in smart industrial systems, lots of unauthorized attacking from the adversary are greatly threatening the security and privacy of the entire industrial systems, of which node replication attacks can hardly be defended, since it is conducted in the physical layer. To solve this problem, we propose a secure random key distribution (SRKD) scheme, which provides a new method for the defense against the attack. Specifically, we combine a localized algorithm with a voting mechanism to support the detection and revocation of malicious nodes. We further change the meaning of the parameter s to help prevent the replication attack. Furthermore, the experimental results show that the detection ratio of replicate nodes exceeds 90% when the number of network nodes reaches 200, which demonstrates the security and effectiveness of our scheme. Compared with existing state-of-the-art schemes, the SRKD scheme also has good storage and communication efficiency.

Recommendation systems have been widely used in large e-commerce websites, but cold start and data sparsity seriously affect the accuracy of recommendation. To solve these problems, we propose SSL-SVD, which works to mine the sparse trust between users and improve the performance of the recommendation system. Specifically, we mine sparse trust relationships by decomposing trust impact into fine-grained factors and employing the Transductive Support Vector Machine algorithm to combine these factors. Then, we incorporate both social trust and sparse trust information into the SVD++ model, which can effectively utilize the explicit and implicit influence of trust for rating prediction in the recommendation system. Experiments show that our SSL-SVD increases the trust density degree of each dataset by more than 65% and improves the recommendation accuracy by up to 4.3%.