Space-Ground Integrated Networks (SGIN) devices are at risk of hardware Trojan attacks. Currently, existing detection schemes (e.g., deep learning) require a large amount of labeled samples. However, obtaining a high-quality labeled hardware Trojan dataset for SGIN devices is challenging due to the structural complexity of hardware, resulting in poor detection performance. To address this challenge, this paper combines prompt-tuning with LangChain to propose a hardware Trojan detection scheme for SGIN devices without requiring extensive training samples. In our scheme, we transform hardware Trojan detection into a mask prediction problem and design a two-phase prompt-based detection framework. In the first phase, we design 5 prompt patterns with masks and utilize Roberta-large as a large language model (LLM) to predict masks and their confidence. If their confidence is below a given threshold value, the second phase is initiated, where the corresponding original samples are fed into LangChain to optimize detection. To enhance the detection accuracy, we develop a Positional State Tree (PST) to extract the logical parallel structure of SGIN Trojan. Experiments show that our scheme achieves an accuracy of 91.3% in detecting the presence of Trojans and 96.3% in identifying the types of Trojans, respectively.

Mental workload recognition is of great significance in preventing human errors and accidents. This study constructed a multimodal recognition scheme to recognize three mental workload states: underload, moderate load, and overload. Based on driving scenarios, these three states were induced in this study by changing the driving modes and situations. Multimodal recognition of underload, moderate load, and overload was performed using electroencephalography (EEG), electrocardiography (ECG), and pupillometry. In addition, various machine learning methods were used to evaluate the recognition performance of different feature combinations. The results showed that the random forest method, trained using spectral power, pupil diameter, and heart rate variability, achieved the highest recognition accuracy of 83.13% for the three mental workload states. This study provides valuable reference information for multimodal recognition of mental workload states.

Hardware Trojans pose a significant security risk in space-ground integrated network (SGIN) devices. It is widely accepted in academia and industry that detecting hardware Trojans at an early stage, typically in register transfer-level (RTL) hardware design, can effectively protect the SGIN device. However, the few hardware Trojan samples dedicated to SGIN (called sHT) make it difficult to detect them using deep learning. To obtain more sHT samples automatically and quickly, this article proposes a lightweight automatic sample augmentation framework for hardware Trojan (HT-ASAF). In our scheme, we first designed a lightweight neural network called variational autoencoder for hardware Trojan (HT-VAE) to achieve high-generation quality without a large amount of training data. Further, we develop the positional state tree (PST) and introduce a node tuple representation for interconversion between PST and sequence to capture the intricate semantic features of concurrent operations in hardware design to enhance the performance of HT-VAE. To automatically verify the effectiveness of the augmented samples, we established an experimental platform incorporating cluster mapping (CLM), which can reduce the verification complexity. In our experiments, to obtain a small number of the training hardware Trojan samples for SGIN, we added activation mechanisms, such as velocity or altitude, to the existing RTL hardware Trojans samples to simulate the hardware Trojan threats faced by orbit devices. The set of the obtained samples is called sHT dataset. Experimental results on the obtained sHT dataset demonstrate that HT-ASAF can automatically and efficiently augment hardware trojan sample compared to existing augmentation schemes, and it performs well in the downstream task of hardware Trojan detection on SGIN devices.

Effectively identifying the workload level of drivers and providing targeted interaction strategies are essential for enhancing human–machine collaboration. Existing studies on workload recognition have primarily focused on binary classification; however, this approach fails to meet the actual requirements for accurate workload assessment. Therefore, to enhance this identification process, this study proposed a new approach to workload recognition and optimal signal feature patterns based on the multiple resource theory and using electroencephalogram signals and machine learning methods. Fifty-six drivers were randomly divided into four groups, with each group participating in a 1-h simulated driving activity. Different visual and auditory workload levels were induced by varying the difficulty of the auditory N-back task and the complexity of the simulation scenarios. Under dichotomous conditions, the support vector machine classifier, featuring the power spectral densities of brain regions, optimally distinguished between auditory and visual overload with an accuracy of 92.41%, with the frontal lobe being the most significant contributor. Under four-class conditions, classification performance was optimal when using a K-nearest neighbor classifier with wholebrain power spectral density as a feature, achieving an accuracy of 57.07%. the most significant contribution in this case came from the Fp2 electrode. These findings provide a theoretical basis for precisely identifying workload states in interaction systems.

Abstract As an essential part of the intelligent transportation system, vehicular ad hoc network plays a vital role in the application of the Internet of Vehicles and security services. However, as the scale of the network continues to increase, trust‐based access and data sharing between vehicles have also become essential security issues. Given how to allocate vehicle resources safely, efficiently and reliably, this paper proposes a blockchain‐based key data sharing mechanism for the Internet of Vehicles. We first utilize certificateless public key cryptography to alleviate redundancy overhead and key escrow issues. Second, with the help of the security, decentralization, and non‐tampering characteristics of the blockchain, two smart contracts are designed and used to realize vehicle identity management and data access control. The identity chain smart contract solves the security problem of vehicle identity management, and the data chain smart contract ensures reliable transaction verification and efficient resource management. At the same time, an auction method is proposed to facilitate vehicle resource transactions. The experimental results show that our solution can realize secure data sharing of the Internet of Vehicles and ensure the operating efficiency of the system. Regarding system overhead, the proposed scheme has 30.6% and 33% lower private key generation overhead and 3.3% and 20.1% lower resource transaction time than AFSN and BSM under the same conditions. In addition, using our bidding method in data chain smart contracts can have a regulatory effect on market data transactions. In the aspect of resource transaction efficiency, for the change of the number of buyers, the transaction completion rate and transaction time of the proposed scheme are better than those of the contrast scheme. When the number of buyers is 100, the transaction completion rate was 9.2% higher than that of AFSN and 17.3% higher than that of BAM. The transaction time of the scheme is 32.2% and 48.1% lower than that of AFSN and BSM, respectively.

With the construction and improvement of 5G infrastructure, more devices choose to access the Internet to achieve some functions. People are paying more attention to information security in the use of network devices. This makes lightweight block ciphers become a hotspot. A lightweight block cipher with superior performance can ensure the security of information while reducing the consumption of device resources. Traditional optimization tools, such as brute force or random search, are often used to solve the design of Symmetric-Key primitives. The metaheuristic algorithm was first used to solve the design of Symmetric-Key primitives of SKINNY. The genetic algorithm and the simulated annealing algorithm are used to increase the number of active S-boxes in SKINNY, thus improving the security of SKINNY. Based on this, to improve search efficiency and optimize search results, we design a novel metaheuristic algorithm, named particle swarm-like normal optimization algorithm (PSNO) to design the Symmetric-Key primitives of SKINNY. With our algorithm, one or better algorithm components can be obtained more quickly. The results in the experiments show that our search results are better than those of the genetic algorithm and the simulated annealing algorithm. The search efficiency is significantly improved. The algorithm we proposed can be generalized to the design of Symmetric-Key primitives of other lightweight block ciphers with clear evaluation indicators, where the corresponding indicators can be used as the objective functions.

It remains a challenge to provision resources in the cloud such that performance is maximized and financial cost is minimized. A fixed budget can be used to rent a wide variety of resource configurations for varying durations. The two steps - resource acquisition and scheduling/allocation - are dependent on each other and are particularly difficult when considering complex resource usage such as workflows, where task precedence need to be preserved and the budget constraint is assigned for the whole cloud application instead of every single job. The ability to acquire resources dynamically and trivially in the cloud - while being incredibly powerful and useful - exacerbates this particular resource acquisition and scheduling problem. In this paper, we design, implement and evaluate two auto-scaling solutions to minimize job turnaround time within budget constraints for cloud workflows. The scheduling-first algorithm distributes the application-wide budget to each individual job, determines the fastest execution plan and then acquires the cloud resources, while the scaling-first algorithm determines the size and the type of the cloud resources first and then schedules the workflow jobs on the acquired instances. The scaling-first algorithm shows better performance when the budget is low while the scheduling-first algorithm performs better when the budget is high. The two algorithms can reduce the job turnaround time by 9.6% - 45.2% compared to choosing a fixed general machine type. Moreover, they show good tolerance (between -10.2% and 16.7%) to inaccurate parameters (±20% estimation error).

One of many advantages of the cloud is the elasticity, the ability to dynamically acquire or release computing resources in response to demand. However, this elasticity is only meaningful to the cloud users when the acquired Virtual Machines (VMs) can be provisioned in time and be ready to use within the user expectation. The long unexpected VM startup time could result in resource under-provisioning, which will inevitably hurt the application performance. A better understanding of the VM startup time is therefore needed to help cloud users to plan ahead and make in-time resource provisioning decisions. In this paper, we study the startup time of cloud VMs across three real-world cloud providers -- Amazon EC2, Windows Azure and Rackspace. We analyze the relationship between the VM startup time and different factors, such as time of the day, OS image size, instance type, data center location and the number of instances acquired at the same time. We also study the VM startup time of spot instances in EC2, which show a longer waiting time and greater variance compared to on-demand instances.

A goal in cloud computing is to allocate (and thus pay for) only those cloud resources that are truly needed. To date, cloud practitioners have pursued schedule-based (e.g., time-of-day) and rule-based mechanisms to attempt to automate this matching between computing requirements and computing resources. However, most of these "auto-scaling" mechanisms only support simple resource utilization indicators and do not specifically consider both user performance requirements and budget concerns. In this paper, we present an approach whereby the basic computing elements are virtual machines (VMs) of various sizes/costs, jobs are specified as workflows, users specify performance requirements by assigning (soft) deadlines to jobs, and the goal is to ensure all jobs are finished within their deadlines at minimum financial cost. We accomplish our goal by dynamically allocating/deallocating VMs and scheduling tasks on the most cost-efficient instances. We evaluate our approach in four representative cloud workload patterns and show cost savings from 9.8% to 40.4% compared to other approaches.

Abstract Ontology mapping (OM) seeks to find semantic correspondences between similar elements of different ontologies. OM is critical to achieve semantic interoperability in the World Wide Web. To solve the OM problem, this article proposes a non‐instance learning‐based approach that transforms the OM problem into a binary classification problem and utilizes machine learning techniques as a solution. Same as other machine learning‐based approaches, a number of features (i.e. linguistic, structural, and web features) are generated for each mapping candidate. However, in contrast to other learning‐based mapping approaches, the features proposed in our approach are generic and do not rely on the existence and sufficiency of instances. Therefore, our approach can be generalized to different domains without extra training efforts. To evaluate our approach, two experiments (i.e. within‐task vs cross‐task) are implemented and the SVM (support vector machine) algorithm is applied. Experimental results show that our non‐instance learning‐based OM approach performs well on most of OAEI benchmark tests when training and testing on the same mapping task; and the results of approach vary according to the likelihood of training data and testing data when training and testing on different mapping tasks. Copyright © 2010 John Wiley & Sons, Ltd.

A significant open issue in cloud computing is performance. Few, if any, cloud providers or technologies offer quantitative performance guarantees. Regardless of the potential advantages of the cloud in comparison to enterprise-deployed applications, cloud infrastructures may ultimately fail if deployed applications cannot predictably meet behavioral requirements. In this paper, we present the results of comprehensive performance experiments we conducted on Windows Azure from October 2009 to February 2010. In general, we have observed good performance of the Windows Azure mechanisms, although the average 10 minute VM startup time and the worst-case 2x slowdown for SQL Azure in certain situations -relative to commodity hardware within the enterprise- must be accounted for in application design. In addition to a detailed performance evaluation of Windows Azure, we provide recommendations for potential users of Windows Azure based on these early observations. Although the discussion and analysis is tailored to scientific applications, the results are broadly applicable to the range of existing and future applications running in Windows Azure.

Ontology mapping seeks to find semantic correspondences between similar elements of different ontologies. This paper proposes a neural network based approach to search for a global optimal solution that best satisfies ontology constraints. Experiments on OAEI benchmark tests show it dramatically improves the performance of preliminary mapping results.

Many application partitioning methods have been proposed based on different functional modules in distributed environment to gain efficient use of resources, improved performance and high scalability. This paper introduces popular application partitioning architectures, and presents a different partitioning method and an on-line workload balance algorithm from a new perspective: runtime data, called data based application partitioning. This architecture benefits applications with higher performance, scalability and dynamic workload balancing. A financial trading system reengineered from J2EE standalone environment into data partitioning distributed environment gives a nice proof.

No abstract available.

Knowledge rich approach of processing documents has been viewed as a method to improve over simple bag-of-word representation. Extracting location information from documents and link them to some ontology such as world gazetteer through a disambiguation process becomes an interesting and important topic. Lacking of training data is a problem in disambiguation method. In this paper we described a method to automatically extract training data from large collection of documents based on local context disambiguation, and then sense profiles are generated automatically for disambiguation use. Another topic of this paper is to describe a linear combination method to combine different types of evidences of disambiguation. We explored three different evidences including location sense context in training documents, local neighbor context, and the popularity of individual location sense. Our results show that combining the three evidences generates reasonable results