Pengchuan Wang

Natural language processing models are widely acknowledged for their strong data fitting capabilities, diverse application scenarios, and adaptable learning methodologies. However, these models, including the large language models, exhibit sensitivity to adversarial example attacks. These examples are slightly perturbed from the pristine text but mislead the model classification. Nevertheless, the existing attack methods primarily focus on the attack effectiveness without semantics-preservation considered. Moreover, the trade-off between evasion effectiveness and concealment of perturbed texts is less investigated. In this study, we propose a multi-objective adversarial text generation framework (MOATG) that simultaneously optimizes attack success rate, imperceptibility, and semantic similarity. Tailored objective functions and dominance relations are designed for character-, word-, and sentence-level perturbations. MOATG is evaluated against five baselines across five benchmark datasets. Experimental results show that MOATG achieves a 5.38% average improvement in attack success rate and reduces word error rate by 1.48%, demonstrating its effectiveness in balancing attack strength and stealth.

In the rapidly evolving landscape of cloud-edge computing, efficient resource scheduling across Kubernetes clusters is essential for optimizing microservice deployment. Traditional scheduling methods, e.g., heuristic and meta-heuristic algorithms, often struggle with the dynamic and heterogeneous nature of cloud-edge environments, relying on fixed parameters and lacking adaptability. We propose and implement DRKC, a novel deep reinforcement learning-based approach that addresses these challenges by improving resource utilization and balancing workloads. We model the scheduling problem as a Markov decision process, enabling DRKC to automatically learn optimal scheduling policies from real-time system data without relying on predefined heuristics. The work synthesizes state information from multiple clusters, using multidimensional resource awareness to effectively respond to changing conditions. We evaluate our performance in three Kubernetes clusters with thirteen nodes and ninety-six test applications with different resource requirements. Experimental results validate the effectiveness of DRKC in enhancing overall resource efficiency and achieving superior load balancing across cloud-edge environments.

As a central subdivision of the low-altitude economy industry, industrial and consumer drones have broad market application prospects and are becoming the primary focus of the low-altitude economy; however, with increasing aircraft density, effective planning of reasonable flight paths and avoiding conflicts between flight paths have become critical issues in UAV clustering. Current UAV path planning often concentrates on 2D and 3D realistic scenes, which do not meet the actual requirements of realistic spherical paths. This paper has proposed a Gradient-Based Optimization algorithm based on the State Transition function (STGBO) to address the spherical path planning problem for UAV clusters. The state transition function is applied on the scale of medium and high-dimensional cities, balancing the stability and efficiency of the algorithm. Through evolution and comparisons with many mainstream meta-heuristic algorithms, STGBO has demonstrated superior performance and effectiveness in solving Medium-Altitude Unmanned Aerial Vehicle (MUAV) path planning problems on three-dimensional spherical surfaces, contributing to the development of the low-altitude economy.

The rapid development of the Industrial Internet of Things (IIoT) and its application across various sectors has led to increased interconnectivity and data sharing between devices and sensors. While this has brought convenience to users, it has also raised concerns about information security, including data security and identity authentication. IIoT devices are particularly vulnerable to attacks due to their lack of robust key management systems, efficient authentication processes, high fault tolerance, and other issues. To address these challenges, technologies such as blockchain and the formal analysis of security protocols can be utilized. And blockchain-based Industrial Internet of Things (BIIoT) is the new direction. These technologies leverage the strengths of cryptography and logical reasoning to provide secure data communication and ensure reliable identity authentication and verification, thereby becoming a crucial support for maintaining the security of the Industrial Internet. In this paper, based on the theory of the strand space attack model, we improved the Fiber Channel Password Authentication Protocol (FACP) security protocol in the network environment based on symmetric cryptography and asymmetric cryptography. Specifically, in view of the problem that the challenge value cannot reach a consensus under the symmetric cryptography system, and the subject identity cannot reach a consensus under the asymmetric cryptography system, an improved protocol is designed and implemented to meet the authentication requirements, and the corresponding attack examples are shown. Finally, the effectiveness and security of the protocol were verified by simulating different networking environments. The improved protocol has shown an increase in efficiency compared with the original protocol across three different network configurations. There was a 6.43% increase in efficiency when centralized devices were connected to centralized devices, a 5.81% increase in efficiency when centralized devices were connected to distributed devices, and a 6.32% increase in efficiency when distributed devices were connected to distributed devices. Experimental results show that this protocol can enhance the security and efficiency of communication between devices and between devices and nodes (servers, disks) in commonly used Ethernet passive optical network (EPON) environments without affecting the identity authentication function.

Recent research has shown that large natural language processing (NLP) models have been vulnerable to a security threat called backdoor attacks. The current privately-triggered backdoor attack achieves the attack target by modifying the label of the poisoned data to a specified label. Still, the result often ignores the consistency of the sample semantics and the label, thus failing to achieve stealthy attacks on system users and system deployers. To address this issue, we use human-written text perturbations, which are less likely to be detected as machine-manipulation and removed by defense systems, as a backdoor trigger and apply a mild adversarial perturbation to the poisoned samples before implanting the backdoor, thereby addressing the semantic inconsistency caused by modifying the labels of the poisoned data. Furthermore, we use learnable backdoor triggers to improve the stealth of backdoor attacks while avoiding the conflict between textual adversarial perturbations and backdoor implantation. Experiments show that our attack achieves close to 100% attack success rates on SST-2, OLID and AG’s News datasets without affecting the utility of existing NLP models.