In recent years, the frequent leakage of sensitive information in natural language texts has presented significant challenges in text distribution, sharing, and migration. Existing methods for text protection and watermarking are limited, capable of either hide sensitive data or prevent unauthorized copying and tampering, but not both simultaneously. To address these limitations, this paper proposes a masking-watermarking cooperative framework designed to hide text-sensitive information, prevent unintentional data leakage, and ensure content ownership verification and tampering prevention. The framework introduces three novel techniques: a variable autoencoder to ensure diverse watermark generation, an improved transformer to enhance the adaptability of dynamic masks, and a dual discriminator for joint verification of text and watermarks. A comprehensive evaluation was conducted, covering text similarity, masking flexibility, and the imperceptibility of text masking, as well as watermark classification recognition and robustness against various attacks. The proposed framework achieved a score of 0.98 on the SBERT metric, demonstrating its effectiveness in achieving imperceptible text masking and robust watermark embedding.

The surge in blockchain transaction volume has corresponds with a rise in phishing incidents. Despite the widespread use of graph representation learning in fraud detection, these techniques often fail to fully exploit the directed multi-graph structure and temporal dynamics of blockchain transaction graphs. This oversight, coupled with the neglect of network tail nodes, impairs the accuracy of phishing detection and distorts graph integrity. This study introduces TDM-GAT, a graph neural network that integrates directionality, edge attributes, and temporal information through transaction flow serialization and functional time encoding. This approach significantly enhances the precision of phishing account identification in blockchain networks. Additionally, a PageRank-based biased sampling strategy is implemented to address the long-tailed distribution of graph nodes, ensuring balanced node participation during learning. Evaluations on three distinct datasets show that TDM-GAT outperforms with an AUC exceeding 95% and an accuracy close to 90%, demonstrating a clear advantage in phishing account detection.

In this work, we propose a method that leverages CLIP feature distillation, achieving efficient 3D segmentation through language guidance. Unlike previous methods that rely on multi-scale CLIP features and are limited by processing speed and storage requirements, our approach aims to streamline the workflow by directly and effectively distilling dense CLIP features, thereby achieving precise segmentation of 3D scenes using text. To achieve this, we introduce an adapter module and mitigate the noise issue in the dense CLIP feature distillation process through a self-cross-training strategy. Moreover, to enhance the accuracy of segmentation edges, this work presents a low-rank transient query attention mechanism. To ensure the consistency of segmentation for similar colors under different viewpoints, we convert the segmentation task into a classification task through label volume, which significantly improves the consistency of segmentation in color-similar areas. We also propose a simplified text augmentation strategy to alleviate the issue of ambiguity in the correspondence between CLIP features and text. Extensive experimental results show that our method surpasses current state-of-the-art technologies in both training speed and performance.

Graph Neural Networks (GNNs) are commonly used for node classification tasks in real-world scenarios. However, research has found that GNN-based classification can significantly drop in performance when the graph structure is perturbed. Current methods, such as adjusting the graph structure to reduce dependency or strengthen protection against perturbations, still have not completely solved this problem. A recent approach using Bayesian inference was introduced to handle perturbations, but this method often faces high uncertainty, which can negatively affect classification results. In this paper, we introduce a new model called WalkBayes, which combines Bayesian inference with random-walk-based label correction to improve the robustness of GNNs in handling complex perturbations. WalkBayes uses random walks to refine label correction, making better use of the graph structure to increase label propagation accuracy and reduce uncertainty. Our experiments on four graph datasets show that WalkBayes performs better than other models in GNN node classification tasks under both random and sparse perturbations to the graph structure.

Federated learning (FL) is a distributed machine learning approach that helps autonomous vehicles train models together without a single organization holding all the data. FL ensures data confidentiality, but it also introduces vulnerabilities to backdoor attacks in autonomous driving systems. These attacks pose a specific threat to the accuracy of traffic sign recognition, potentially leading to the misclassification of signs and subsequent traffic accidents. To overcome these challenges, we propose a novel framework for backdoor attacks called Covert and Persistent Backdoor Attacks (CPBA), designed specifically for traffic sign recognition in autonomous driving systems. This framework utilizes a Generative Adversarial Network (GAN) to generate specific covert triggers for each sample. Minimizing the differences in feature vectors ensures that the backdoored images retain visual similarity to the original ones, making them nearly indistinguishable from human observers. Furthermore, CPBA addresses the dynamic nature of autonomous driving systems by selectively targeting model parameters that are infrequently updated and more stable, ensuring sustained effectiveness of the backdoor despite fluctuations in compromised vehicle participation and the influence of benign updates. Experimental evaluations demonstrate that CPBA maintains robustness against five distinct defense mechanisms on two publicly accessible traffic sign recognition datasets.

Summary The security threats from malicious web pages have become a hot topic for cyber security. One goal pursued by current research is to identify malicious web pages quickly, accurately, and efficiently. Considering the high detection costs and potential dimensionality curse of malicious webpage detection, in this article, we proposes a detection framework based on feature optimization and hybrid classification. It provides three properties: more new malicious webpage features, information gain‐based feature selection method, and integrating multiple machine learning method. A comprehensive experimental evaluation demonstrates that the proposed framework has remarkable advantages in aspects of detection accuracy and detection performance.

Summary Malicious webpage detection is a crucial work in both theory and practical environment. In practical applications, static detection methods are usually regarded as a priority choice, which can quickly detect unknown malicious web pages and avoid a costly in‐depth analysis. However, existing solution of static detection typically has the following problems. For example, a single static detection may lead to a higher false positive rate, and the integrated detection usually has a lower detection efficiency. In this article, we propose an efficient webpage static detection framework, especially considering both the detection efficiency and the detection accuracy. Then, on the basis of the extended feature sets from URL, HTML, and JavaScript, we introduce an optimized naive Bayesian algorithm, in which a novel amplification factor strategy is proposed. Finally, a webpage threat assessment model oriented to general machine learning is presented to achieve the refined detection. Three main properties are provided: high detection efficiency, high detection accuracy, and better applicability. Furthermore, the comprehensive experimental results and comparative analysis is given to show the advantages of the proposed framework.

Abstract In recent years, with the rapid development of mobile social networks and services, the research of mobile malicious webpage detection has become a hot topic. Most of the existing malicious webpage detection systems are deployed on desktop systems and servers. Due to the limitation of network transmission delay and computing resources, these existing solutions fail to provide the real‐time and lightweight properties for mobile webpage detection. In this paper, we propose an advanced mobile malicious webpage detection framework based on deep learning and edge cloud. Inspired by the idea of edge computing, a multidevice load optimization approach is first introduced to improve detection efficiency. Second, an automatic extraction approach based on deep learning model features is presented to enhance detection accuracy. Furthermore, detection systems can be flexibly deployed on edge nodes and servers, thus providing the properties of resource optimization deployment and real‐time detection. Finally, comparative analysis and performance evaluation are presented to show the detection efficiency and accuracy of the proposed framework.

In recent years, with the rapid development and popularization of e-commerce, the applications of e-coupons have become a market trend. As a typical bar code technique, QR codes can be well adopted in e-coupon-based payment services. However, there are many security threats to QR codes, including the QR code tempering, forgery, privacy information leakage and so on. To address these security problems for real situations, in this paper, we introduce a novel fragment coding-based approach for QR codes using the idea of visual cryptography. Then, we propose a QR code scheme with high security by combining the fragment coding with the commitment technique. Finally, an enhanced QR code-based secure e-coupon transaction framework is presented, which has a triple-verification feature and supports both online and offline scenarios. The following properties are provided: high information confidentiality, difficult to tamper with and forge, and the ability to resist against collusion attacks. Furthermore, the performance evaluation of computing and communication overhead is given to show the efficiency of the proposed framework.

The Internet of Things (IoT) is becoming a backbone of sensing infrastructure to several mission-critical applications such as smart health, disaster management, and smart cities. Due to resource-constrained sensing devices, IoT infrastructures use Edge datacenters (EDCs) for real-time data processing. EDCs can be either static or mobile in nature, and this article considers both of these scenarios. Generally, EDCs communicate with IoT devices in emergency scenarios to evaluate data in real-time. Protecting data communications from malicious activity becomes a key factor, as all the communication flows through insecure channels. In such infrastructures, it is a challenging task for EDCs to ensure the trustworthiness of the data for emergency evaluations. The current communication security pattern of “communication before authentication” leaves a “black hole” for intruders to become part of communication processes without authentication. To overcome this issue and to develop security infrastructures for IoT and distributed Edge datacenters, this article proposes a user-centric security solution. The proposed security solution shifts from a network-centric approach to a user-centric security approach by authenticating users and devices before communication is established. A trusted controller is initialized to authenticate and establishes the secure channel between the devices before they start communication between themselves. The centralized controller draws a perimeter for secure communications within the boundary. Theoretical analysis and experimental evaluation of the proposed security model show that it not only secures the communication infrastructure but also improves the overall network performance.

In recent years, with the rapid development and popularization of QR code-based services, the research of QR codes has become a hot topic. Because QR codes are easy to use, they are well adopted in e-coupon-based payment services. However, there are many security threats to QR codes, including QR code forgery, privacy information leakage and so on. To address these security problems, in this paper, we first propose a novel fragment coding-based approach for QR codes using the idea of visual cryptography. Second, we propose a QR code scheme with a high security by combining the fragment coding with commitment technique. Then, an enhanced QR code-based secure e-coupon transaction framework is presented, which has a triple verification feature. This framework can provide at least the following properties: high information confidentiality, difficult to tamper with and forge, and the ability to resist collusion attacks. Finally, security analysis and performance evaluation are presented to show the security and efficiency of the proposed framework.

Ensuring the security and privacy of vehicular ad hoc networks (VANETs) and related services such as secure payment has been the focus of recent research efforts. Existing secure payment solutions generally require stable and reliable network connection. This is, however, a challenge in a VANET setting. Drive-thru Internet, a secure payment solution for VANETs, involves a great number of fast-moving vehicles competing for connections/communications simultaneously. Thus, service providers may find it challenging to provide real-time payment services or may have to sacrifice the confidentiality and the authenticity of payment vouchers for usability. In this article, we propose a secure installment payment framework for drive-thru Internet deployment in a VANET setting. The framework also provides the capability to embody properties such as confidentiality of payment vouchers, offline signature verification, periodical reconciliation, and installment payment. Performance evaluation and security analysis demonstrate the utility of the framework in a VANET setting.

Abstract Opportunistic vehicular ad hoc networks have attracted enormous attention from both industry and academia in recent years. However, the presence of selfish behaviors of nodes could cause a severe threat to well designed opportunistic routing scheme, and even jeopardize the whole network. This paper presents a privacy‐preserving distance‐based incentive scheme, particularly to address issues of the nodes' selfish behavior and the location privacy. The proposed protocol adopts secure multiparty computation and homomorphic encryption methods to realize three properties, i.e., the confidentiality of nodes location information, the integrity of the message carried distance, and the correctness of reputation computation. Furthermore, this incentive scheme can satisfy the following security requirements, such as mutual authentication, non‐repudiation, and conditional privacy preservation, and it can also stimulate the active helpers and suppress harmful behaviors fairly and reasonably. The security analysis and performance evaluation show that the proposed framework is secure, efficient, and practical. Copyright © 2015 John Wiley & Sons, Ltd.

Opportunistic routing has been extensively studied and utilized in networks with high dynamics and large scales, e.g., city-wide vehicle networks. The extensive use of nodes' local information, i.e., the routing metrics, in such routings can cause severe security and privacy problems. Existing solutions of anonymous routing can introduce undesired overhead and fail to provide the confidentiality of the routing metric. In this paper, we propose an advanced framework for opportunistic routings, providing following properties: the confidentiality of nodes' routing metric, anonymous authentication and an efficient key agreement for pair-wise secret communication. A comprehensive evaluation, including security analysis, efficiency analysis and simulation evaluation, is presented to show the security and feasibility of the proposed framework.

Vehicular ad hoc networks have attracted a lot of attention in recent years, in either vehicle-to-vehicle or vehicle-to-infrastructure scenarios. In this paper, we focus on the latter, particularly for vehicles to upload to roadside units, the so-called drive-thru Internet, in a secure and efficient manner. Due to the ad hoc nature and wireless communications, traditional certificate-based security schemes are either infeasible or inefficient in this scenario. Thus we propose a certificateless approach to secure upload in a drive-thru Internet. We discuss the attack model and the desired security properties, and how to achieve these properties through the proposed certificateless scheme. We implement and evaluate the proposed scheme, and also investigate how to mitigate the security overhead through the separation of security association and data transfer in a drive-thru Internet.