Voice synthesis technologies have advanced rapidly, raising serious concerns about content security and trust. While many fake voice detectors achieve strong performance in controlled settings, they often overfit to speaker-irrelative features (SiFs), exhibit poor robustness, and fail in multi-speaker scenarios. To address these limitations, we propose MultiSiFer, a novel fake voice detector grounded in a new design philosophy: rather than merely distinguishing synthetic from human voices, it explicitly prioritizes learning essential human voice characteristics. MultiSiFer leverages a pre-trained speech representation model to enhance this learning and is the first detector trained on a newly curated multi-speaker fake voice dataset, enabling effective generalization across speakers. Experiments show that MultiSiFer outperforms existing methods in both standard and multi-speaker settings, achieving 10.84% average equal error rate (EER).

Event-driven neuromorphic computing involves sparse and asynchronous signal activity, which leads to irregular computation patterns and fine-grained concurrency. As a result, processing architectures need to support both high parallelism and energy efficiency. Among existing architectural solutions, superscalar designs exhibit significant potential for addressing high parallelism demands. However, conventional superscalar processors, which rely on synchronous circuits, maintain high-frequency clocking at all times, leading to substantial power inefficiency in sparse computation scenarios. To address this issue, we propose an asynchronous superscalar architecture that replaces global clocking with fully local handshake-based control, implemented using a bundled-data asynchronous protocol. The design supports decoding of up to 64 scalar instructions per cycle and implements the RISC-V RV32IMC instruction set. A prototype was fabricated using a 110 nm complementary metal oxide semiconductor (CMOS) process and was evaluated through post-layout simulation. Operating at 1.2 V, the processor delivers a peak INT8 throughput of 669.4 GOPS, with a static power consumption of 421 mW.

Constructing online High-Definition (HD) maps is crucial for the static environment perception of autonomous driving systems (ADS). Existing solutions typically attempt to detect vectorized HD map elements with unified models; however, these methods often overlook the distinct characteristics of different non-cubic map elements, making accurate distinction challenging. To address these issues, we introduce an expert-based online HD map method, termed MapExpert. MapExpert utilizes sparse experts, distributed by our routers, to describe various non-cubic map elements accurately. Additionally, we propose an auxiliary balance loss function to distribute the load evenly across experts. Furthermore, we theoretically analyze the limitations of prevalent bird's-eye view (BEV) feature temporal fusion methods and introduce an efficient temporal fusion module called Learnable Weighted Moving Descentage. This module effectively integrates relevant historical information into the final BEV features. Combined with an enhanced slice head branch, the proposed MapExpert achieves state-of-the-art performance and maintains good efficiency on both nuScenes and Argoverse2 datasets.

The longest match strategy in LZ77, a major bottleneck in the compression process, is accelerated in enhanced algorithms such as LZ4 and ZSTD by using a hash table. However, it may results in numerous random memory accesses, which modern CPUs handle inefficiently, thus reducing the compression speed. In this paper, we introduce the LZHV algorithm, which significantly reduces unnecessary memory accesses by over 99% through the implementation of hash verification within the hash table. By integrating LZHV into LZ4 at its default compression level and ZSTD at levels 3 and 4, we achieve a compression speed improvement of over 10% across various platforms.

Open-source software (OSS) is widely reused in Internet of Things (IoT) devices, leading to widespread N-Day vulnerabilities when outdated components remain unpatched. Existing methods typically encode features of different Common Vulnerabilities and Exposures (CVEs) within a shared representation space. However, the model’s limited capacity, combined with the new vulnerability features, can disrupt previously learned patterns. Minimal code modifications in tiny-patch vulnerabilities are often overshadowed by variations introduced by different compilation settings, making it more difficult to distinguish vulnerable functions from their patched counterparts. This paper introduces ISGraphVD, a novel graph-based and function-level vulnerability detection approach that supports cross-compilation settings and enhances detection accuracy. By modeling each CVE independently through a one-model-per-CVE strategy, ISGraphVD reduces feature interference and improves detection accuracy across diverse CVEs. To better detect tinypatch vulnerability, we propose ISGraph, a fine-grained graph representation that models variable dependencies within and across basic blocks by integrating control flow analysis. Then, ISGraphVD utilizes a Graph Matching Network (GMN) with a cross-graph attention mechanism to identify critical vulnerability patterns. Experiments on IoT OSS projects show that ISGraphVD outperforms state-of-the-art methods, achieving a 6.3 percentage-point (pp) accuracy improvement over the strongest baseline, and real-world tests further validate its effectiveness in IoT supply chains.

The application of deep learning in voice cloning has significantly enhanced the quality of cloned voices. While advanced voice cloning technologies are widely applied across various domains, they also pose serious security challenges such as producing natural Deepfakes. In response, numerous studies have focused on detecting fake voices, with many reporting outstanding performance. However, is the issue truly resolved? This paper introduces Adversarial Neural Mimicry Attack (ANMA) which leverages a specialized model to predict the behavior of other similar models, transforming black-box attacks into white-box scenarios indirectly. Based on ANMA and Speaker-irrelative Features (SiFs), we propose a novel black-box attack framework called SiFMimicEvader, designed to evade fake voice detectors with high success rates and minimal query requirements. The framework utilizes speech representation models as the breakthrough to predict the behaviors of fake voice detectors and employs a series of SiFs editing operations as perturbations to deceive these detectors. Experimental results demonstrate the effectiveness of SiFMimicEvader, achieving an average attack success rate exceeding 50% across various detectors, significantly outperforming other attack methods, while also showing great performance in audio quality and query scale, indicating its high availability in real-world scenarios.

Autonomous driving represents a prominent application of artificial intelligence. Recent approaches have shifted from focusing solely on common scenarios to addressing complex, long-tail situations such as subtle human behaviors, traffic accidents, and non-compliant driving patterns. Given the demonstrated capabilities of large language models (LLMs) in understanding visual and natural language inputs and following instructions, recent methods have integrated LLMs into autonomous driving systems to enhance reasoning, interpretability, and performance across diverse scenarios. However, existing methods typically rely either on real-world data, which is suitable for industrial deployment, or on simulation data tailored to rare or hard case scenarios. Few approaches effectively integrate the complementary advantages of both data sources. To address this limitation, we propose a novel VLM-guided, end-to-end adversarial transfer framework for autonomous driving that transfers long-tail handling capabilities from simulation to real-world deployment, named CoC-VLA. The framework comprises a teacher VLM model, a student VLM model, and a discriminator. Both the teacher and student VLM models utilize a shared base architecture, termed the Chain-of-Causality Visual–Language Model (CoC VLM), which integrates temporal information via an end-to-end text adapter. This architecture supports chain-of-thought reasoning to infer complex driving logic. The teacher and student VLM models are pre-trained separately on simulated and real-world datasets. The discriminator is trained adversarially to facilitate the transfer of long-tail handling capabilities from simulated to real-world environments by the student VLM model, using a novel backpropagation strategy. Experimental results show that our method effectively bridges the gap between simulation and real-world autonomous driving, indicating a promising direction for future research.

Robotic navigation in unknown environments is challenging due to the lack of high-definition maps. Building maps in real time requires significant computational resources. Nevertheless, sensor data can provide sufficient environmental context for robots’ navigation. This paper presents an interpretable and mapless navigation method using only two-dimensional (2D) light detection and ranging (LiDAR), mimicking human strategies to escape from dead ends. Unlike traditional planners, which depend on global paths or vision-based and learning-based methods, requiring heavy data and hardware, our approach is lightweight and robust, and it requires no prior map. It effectively suppresses oscillations and enables autonomous recovery from local minimum traps. Experiments across diverse environments and routes, including ablation studies and comparisons with existing frameworks, show that the proposed method achieves map-like performance without a map—reducing the average path length by 50.51% when compared to the classical mapless Bug2 algorithm and increasing it by only 17.57% when compared to map-based navigation.

The normalized difference vegetation index (NDVI) serves as an essential measure for vegetation assessment and plays a pivotal role in environmental monitoring, ecosystem conservation, and the advancement of sustainable practices. However, NDVI forecasting typically presents a spatio-temporal challenge. While traditional techniques, such as convolutional long short-term memory (LSTM) and graph neural networks (GNNs) are frequently utilized to address this issue, the explicit spatial attributes inherent in the geographic coordinates of observation sites are often neglected in existing studies. To fill this research gap, we introduce an innovative time-aware model, fuzzy convolutional neural network long short-term memory (CNN-LSTM), designed specifically for NDVI prediction within the Chinese context. This model leverages the adaptive neuro-fuzzy inference system (ANFIS) to encapsulate the spatial nuances presented by the latitude, longitude, and elevation of the observation points. It also harnesses the power of 1-D convolution and LSTM to delineate temporal patterns. We incorporate a gate control mechanism to effectively blend the spatial intelligence rendered by ANFIS with the temporal insights captured by CNN-LSTM. We also combine deep neural fuzzy systems with traditional temporal neural networks. A comparative analysis spanning several temporal intervals highlights the superior performance of our proposed model in spatio-temporal forecasting in relation to conventional methods. Subsequent empirical evaluations confirm that the model has strong generalizability across diverse provinces.

As an important renewable energy source, photovoltaic (PV) power is widely used in power systems because of its clean, sustainable and increasingly mature technology, and ultra-short-term PV power forecasting is very important for the real-time dispatchment of power grids. However, PV power has strong periodicity and randomness, changes in weather factors such as illumination and temperature will significantly affect its power generation efficiency, which also brings challenges for PV power forecasting. In this paper, we propose a novel deep learning model called SL-ANFIS-LSTM for ultra-short-term PV power forecasting. SL-ANFIS-LSTM utilizes a structure learnable ANFIS (SL-ANFIS) module based on the adaptive neuro-fuzzy inference system (ANFIS) and Kolmogorov-Arnold networks (KAN) to characterize uncertain fuzzy factors such as weather and temperature, and a long short-term memory (LSTM) module to extract the temporal dependency. Besides, physical information is also introduced to correct illegal errors caused by changes in sunlight intensity. Experimental results show that the proposed model is superior to related models for ultra-short-term PV power forecasting.

SQL injection is a significant and persistent threat to web services. Most existing protections against SQL injections rely on traffic-level anomaly detection, which often results in high false-positive rates and can be easily bypassed by attackers. This paper introduces SQLStateGuard, the world's first middleware-driven statement-level SQL injection defense approach, to address these issues. The SQLStateGuard uses a custom SQL middleware based on the idea of Runtime Application Self-Protection to capture raw SQL statements. These statements are then analyzed by SQLSG-Net, a database-oriented detection network based on gated linear units. If SQLSG-Net detects malicious SQL statements, the SQL middleware will block them. Experiments show that the detection accuracy of SQLStateGuard exceeds 99%, outperforming existing approaches, and it can identify the type of a specific SQL injection. Additionally, SQLStateGuard has no fingerprint and does not respond to SQL syntax errors, making it more challenging for attackers to gather information. This paper also presents a novel dataset generation process for SQLStateGuard and shares two statement-level SQL injection datasets with the research community, including over 145,000 malicious SQL statements categorized by the type of SQL injection.

Recent speech synthesis technology can generate high-quality speech indistinguishable from human speech, thus introducing various security and privacy risks. Numerous recent studies have focused on fake voice detection to address these risks, with many claiming to achieve ideal performance. However, is this really the case? A recent research work introduced Speaker-Irrelative-Features (SiFs), unrelated to the information in speech files but capable of influencing fake detectors. This means that existing detectors may rely on SiFs to a certain extent to distinguish real and fake speech. In this paper, we introduce an evaluation framework to evaluate the influence of SiFs in existing fake voice detectors in depth. We evaluate three SiFs which include background noise, the mute parts before and after voice, and the sampling rate on ASVspoof2019 and FoR. Our results confirm the substantial influence of SiFs on fake voice detection performance, and we delve into the analysis of the underlying mechanisms.

Low-end field programmable gate arrays (FPGAs) are difficult to deploy typical convolutional neural networks (C- NNs) owing to the limited hardware resources and the increasing model computational complexity. Fast Fourier transform (FFT) is a promising solution for saving both computation and memory footprint by convolving in the frequency domain. However, few FPGA accelerators can take full advantage at the computation level, because of the distinct element-wise complex calculation in the frequency domain. In this work, we present an FPGA-based 8-bit inference accelerator (called FAF) that packs frequency-domain calculations into digital signal processing (DSP) blocks to fully utilize DSPs for performance boost. We then provide a mapping dataflow to maximize the reduction of redundant packing operations by frequency-domain data reuse. Evaluations based on representative CNN benchmarks show that our work can achieve 1.5-6.9× better power efficiency compared with representative FPGA baselines.

Open-source software has emerged as a pivotal force in the advancement of information technology. Robust open-source compliance governance is essential for the sustainable and healthy growth of both open-source software and its communities. License incompatibility analysis, in particular, represents a critical challenge hindering the progress of open-source software. Traditional methods of incompatibility analysis often fail to account for diverse usage scenarios or are tailored to a limited subset of scenarios. This limitation obstructing their ability to handle the intricate compatibility arising from varied programming language interactions, leading to a high false positives. Our study embarks from an examination of license exceptions, delving into the incompatibility analysis challenges through extensive empirical research on these exceptions. We discovered that the majority of exceptions are, in fact, detectable. Leveraging this empirical insight, our research further develops the license compatibility analysis model by introducing a new, refined legal terminology representation alongside a novel method for license compatibility reasoning. This approach begins with modeling different scenarios to represent license compatibility variably. Furthermore, based on these modeling outcomes, we have designed and implemented LiScopeLens, a tool capable of discerning dependency behaviors for granular compatibility assessment, starting with binary dependencies. Our experimental findings affirm that LiScopeLens proficiently determines the license compatibility status of open-source software across various usage scenarios, demonstrating its significant practical utility.

The canonical video action recognition methods usually label categories with numbers or one-hot vectors and train neural networks to classify a fixed set of predefined categories, thereby constraining their ability to recognise complex actions and transferable ability to unseen concepts.In contrast, cross-modal learning can improve the performance of individual modalities.Based on the facts that a better action recogniser can be built by reading the statements used to describe actions, we exploited the recent multimodal foundation model CLIP for action recognition.In this study, an effective Vision-Language action recognition adaptation was implemented based on few-shot examples spanning different modalities.We added semantic information to action categories by treating textual and visual label as training examples for action classifier construction rather than simply labelling them with numbers.Due to the different importance of words in text and video frames, simply averaging all sequential features may result in ignoring keywords or key video frames.To capture sequential and hierarchical representation, a weighted token-wise interaction mechanism was employed to exploit the pair-wise correlations adaptively.Extensive experiments with public datasets show that cross-modal action recognition learning helps for downstream action images classification, in other words, the proposed method can train better action classifiers by reading the sentences describing action itself.The method proposed in this study not only reaches good generalisation and zero-shot/few-shot transfer ability on Out of Distribution (OOD) test sets, but also performs lower computational complexity due to the lightweight interaction mechanism with 84.15% Top-1 accuracy on the Kinetics-400.

Tibetan language processing is crucial for preserving its rich cultural heritage and reducing communication barriers between different languages. However, as a low-resource language, the development of Tibetan natural language processing has lagged behind. To address the unique and complex structural information of Tibetan, this article improves the embedding model based on fundamental Tibetan Component-and-Character-and-Word-based Embedding (TCCWE) to enhance the effectiveness of word vector representation. We incorporate position information into the training of Tibetan word vectors, developing models based on components, characters, and their integration. Furthermore, to evaluate the effectiveness of these word vectors, we propose an intrinsic evaluation set, wordsimT, based on k -means clustering. Experimental results demonstrate that the character-based positional vector integration model achieves a Spearman's rank correlation coefficient of 79.99% on the wordsimT benchmark, outperforming the baseline TCCWE model by 1.51%. Additionally, we validate the proposed models in downstream text classification tasks. These findings underscore the importance of incorporating positional information in Tibetan word vectors.

Random number generation is essential for applications in simulation, numerical analysis, and data encryption. The ubiquitous presence of system-on-chip (SoC) field-programmable gate array (FPGA) embedded devices in critical sectors necessitates robust random number generators (RNGs) that operate within these specialized environments. Traditional RNGs in GNU/Linux systems derive entropy from peripheral hardware events, which are scarce in SoC FPGA platforms lacking standard PC peripherals. Addressing this challenge, this paper proposes a novel random number generator named NDSTRNG that leverages the unique hardware structure of the SoC FPGA and the inherent randomness of GNU/Linux. The proposed generator employs a non-deterministic sampling model to circumvent reliance on various peripherals while ensuring unbiased output via a linear feedback shift register (LFSR)-based post-processing method. We implement this random number generator in SoC FPGA GNU/Linux using minimal FPGA resources and only one Linux task for sampling. NDSTRNG achieved a throughput exceeding 700 Kbps. Moreover, the entropy source of the generator is evaluated using NIST SP 800-90B, while the quality of the generated random numbers is assessed through ENT, NIST SP 800-22, and DIEHARDER. The results confirm that NDSTRNG meets the stringent criteria for both high-quality and high-speed random number generation, making it suitable for deployment in communication, defense, and medical domains where reliable RNGs are indispensable.

Pulmonary arterial hypertension (PAH) is considered the third most common cardiovascular disease after coronary heart disease and hypertension. The diagnosis of PAH is mainly based on the comprehensive judgment of computed tomography and other medical image examinations. Medical image processing based on deep learning has achieved significant success. However, the data belongs to the patient's privacy; therefore, the medical institutions as data custodians have the responsibility to protect the security of their data privacy. This situation makes medical institutions face a dilemma when building data-driven deep learning-assisted medical diagnosis methods. On the one hand, they need to pursue more high-quality data based on Big Data architecture for deep learning; on the other hand, they need to protect patient privacy to avoid data leakage. In response to the above challenges, we propose a hierarchical hybrid automatic segmentation model for pulmonary blood vessels based on local learning and federated learning approaches for segmenting the pulmonary blood vessels. The experiments prove the proposal could automatically segment the vessels from the original CT. It also indicates that the model based on a federated learning approach can achieve impressive performance under the premise of protecting data privacy for Big Data.

Pre-copy-based Virtual Machine (VM) live migration seamlessly migrates the running VM to the target physical server by pre-copying memory pages and realizing updates through loop iterations. This method, which has high reliability and robustness, can effectively achieve load balancing and reduce energy consumption. It is widely used in the industry to manage server cluster resources. However, it also involves many problems, such as many dirty memory pages resulting from repeated transmission and convergence failure of iterative transmission. Hence, pre-copy live migration cannot efficiently allocate server cluster resources. To resolve these problems, a VM pre-copy live migration technology based on the similarity of dirty memory pages is proposed in this paper. The access priority of historical dirty memory pages was determined by calculating the similarity weight based on the Hamming distance. A priority-based delay transmission scheme for high dirty pages and low dirty pages was used to decrease the frequent transmission of high dirty memory pages, increase the convergence speed of the live-migration iterative copy process, and reduce the overall migration time of VMs. A comparative analysis of experimental results based on six dimensions showed that the proposed method achieved better migration efficiency than the conventional live migration strategy.

Social networks contributed massive text data generated by users in it, which were crucial in information explosion. These unstructured and ambiguous expressed data may result in difficulty in obtaining available contextual information from it, which can help us gain more accurate insights into user-generated content, user preferences, and topic dynamics within social networks. By training on a large-scale unsupervised corpus and fine-tuning parameters using a limited amount of supervised data, the pretrained language model can effectively capture rich contextual information and achieve excellent performance in numerous downstream tasks of natural language processing (NLP). For the low-resource language such as Tibetan, the distributed representation results of dynamic changes obtained from pretrained language models can effectively alleviate the problem of insufficient labeled data. In order to achieve more effective contextual information and word-level semantic information in Tibetan social media, we collected a large amount of Tibetan language corpus and trained a Tibetan pretrained language model, named as Tibetan-BERT-wwm, by using the whole word masking strategy. Additionally, we apply the model to analyze textual data from social networks to assess its efficacy in capturing user sentiments and news topic in Tibetan social media. In this study, accuracy, precision, recall, and F1 score were used to evaluate its performance. The results showed that the macro-F1 of Tibetan-BERT-wwm in the public dataset TNCC document and title are 75.55% and 64.17%, the self-built sentiment analysis dataset is 70.98%. Compared with other pretrained language models, the Tibetan-BERT-wwm model can capture the semantic information of Tibetan well and improve the Tibetan classification effect.

Photovoltaic power prediction plays an extremely important role in the construction of smart power grid and power grid security protection. In order to solve the problem of unstable power generation and even damages to the power grid caused by the ever changing irradiance and meteorological conditions, this paper leverages the traditional time series prediction modeling method to the machine learning approaches, such as recurrent neural network (RNN), convolutional neural network (CNN) and decision tree (DT), and uses the ensemble machine learning method to improve the final prediction accuracy, by training and testing on the radiation and meteorological data collected from a photovoltaic power station in Gansu Province of China, which enjoys the best solar resources in the country. The experimental results show that the ensemble model achieves the highest prediction accuracy, and its root mean square error(RMSE) is 0.4477. This is of great significance to the power generation evaluation and dispatching of photovoltaic power station.

Voice is an essential medium for human communication and collaboration, and its trustworthiness is of great importance to humans. Synthesizing fake voices and detecting synthesized voices are two sides of a coin. Both sides have made great strides with the recently prospering deep learning techniques. Attackers started using AI techniques to synthesize, even clone, human voices. Researchers also proposed a series of AI-synthesized voice detection approaches and achieved promising results in laboratory environments.In this paper, we introduced the concept of speaker-irrelative features (SiFs) and a novel detection-bypass idea to camouflage AI-synthesized voices: replacing SiFs of AI-synthesized voices with crafted ones. We implemented a proof-of-concept framework named SiF-DeepVC based on our detection-bypass idea. Experiments show that the existing detection systems would consider the voices output by SiF-DeepVC more human-like than human voices, proving our detection-bypass idea is effective and SiFs are noteworthy in camouflaging AI-synthesized voices.

Voice is a vital medium for transmitting information. The advancement of speech synthesis technology has resulted in high-quality synthesized voices indistinguishable from human ears. These fake voices have been widely used in natural Deepfake production and other malicious activities, raising serious concerns regarding security and privacy. To deal with this situation, there have been many studies working on detecting fake voices and reporting excellent performance. However, is the story really over? In this paper, we propose SiFDetectCracker, a black-box adversarial attack framework based on Speaker-Irrelative Features (SiFs) against fake voice detection. We select background noise and mute parts before and after the speaker's voice as the primary attack features. By modifying these features in synthesized speech, the fake speech detector will make a misjudgment. Experiments show that SiFDetectCracker achieved a success rate of more than 80% in bypassing existing state-of-the-art fake voice detection systems. We also conducted several experiments to evaluate our attack approach's transferability and activation factor.

Safety-critical systems play a significant role in industrial domains, and their complexity is increasing with advanced technologies such as Artificial Intelligence (AI). To provide efficient services, safety-critical systems that integrate AI applications are always built based on Linux, where Linux offers massive amounts of features and an incredibly perfect software ecosystem for AI applications. Since Linux is a pre-existing complex software system, different research programmes aim to pave the way for developing Linux-based safety-critical systems. Still, only some focus on the system calls for file operations. However, the execution path of a system call is effectively non-deterministic in Linux kernel space, which challenges the test coverage-based verification recommended by the functional safety standards. This research analyzes the influence of system state on Linux kernel path variability from two perspectives: file system type and system load. Therefore, an online data collection system for system call execution paths was constructed based on Ftrace, network file system (NFS), and MD5 hash function, uniquely identifying the system call execution path. The collected data were processed and analysed in this study. Evaluations show that the number of function execution paths of the system calls relevant to file systems increased with the increase in system load but would eventually be stable. Additionally, the function execution paths of the system call varied in different file systems. Based on the evaluations, the results of this work can provide advice for analyzing Linux-based safety-critical systems. In addition, the method introduced in this research can also provide support for the verification of Linux-based safety-critical systems.

The Internet of Things (IoT) has become a focus of information infrastructure development in recent years. The smart blockchain can provide various solutions for trust, security, and privacy (TSP) challenges to protect IoT data, and smart contracts are the foundation of blockchain intelligence, and greatly enhance the ability of smart blockchain to solve TSP problems. So, the security of smart contracts must be addressed. We propose an efficient smart contract vulnerability detector to improve the safety of smart contracts. It comprises a graph extraction method and a complete vulnerability detection process. The graph extraction method consists of vulnerability pattern extraction and a graph generation process. The vulnerability detection process first uses the approximate graph matching algorithm to select representative SCGraphs from the data set to build vulnerability SCGraph libraries. Second, determine whether the contract contains vulnerabilities by calculating the similarity between the SCGraphs generated from the contracts to be detected and the SCGraphs in the vulnerability library. Experiments show that our approach achieves an inspiring high detection rate and is the fastest among existing vulnerability detection tools, which indicates that it can provide good vulnerability detection for smart contracts.

Convolutional neural networks (CNNs) (including 2D and 3D convolutions) are popular in video analysis tasks such as action recognition and activity understanding. Fast algorithms such as fast Fourier transforms (FFTs) are promising in significantly reducing computation complexity by transforming convolution into frequency domain. In frequency space, conventional spatial convolutions are replaced with simpler element-wise complex multiplications. Conventional application-specific-integrated-circuit (ASIC) based frequency-domain accelerators can achieve effective performance boost but come at the cost of significant energy consumption, owing to the hierarchical memory organization. We propose a frequency-domain resistive random access memory (ReRAM) based inference accelerator called FDA that can process element-wise complex multiplication in memory for both 2D and 3D CNNs. Each ReRAM-based frequency-domain process element (PE) with two ReRAM cells can perform an element-wise complex multiplication in two continuous execution cycles. We then provide a flexible dataflow to alleviate the redundant data movements by frequency-domain data reuse and inherent symmetrical characteristic for both 2D and 3D convolutions. Evaluation results based on representative both 2D and 3D CNN benchmarks demonstrate that FDA outperforms state-of-the-art baselines with better performance and energy efficiency.

As an essential intermediate courier station on the expressway, the service area plays an important role in regulating expressway traffic, ensuring expressway safety, and alleviating driver fatigue. With the rapid development of information technology, Expressway Service Areas (ESA) are developing in digitization and intelligence. This paper conducts exploratory research on Expressway Intelligent Service Area (EISA), sum-marizes the definition and characteristics of EISA, analyzes the current situation and existing problems of EISA construction, and elaborates the key technologies that the construction of EISA relies on. The research of this paper is in an effort to provide ideas for the subsequent reconstruction and expansion planning and design of the EISA.

Background: With the boosting development of IoT technology, the supply chains of IoT devices become more powerful and sophisticated, and the security issues introduced by code reuse are becoming more prominent. Therefore, the detection and management of vulnerabilities through code similarity detection technology is of great significance for protecting the security of IoT devices. Aim: We aim to propose a more accurate, parallel-friendly, and realistic software supply chain vulnerability detection solution for IoT devices. Method: This paper presents PG-VulNet, standing for Vulnerability-detection Network based on Pseudo-code Graphs. It is a ”multi-model” cross-architecture vulnerability detection solution based on pseudo-code and Graph Matching Network (GMN). PG-VulNet extracts both behavioral and structural features of pseudo-code to build customized feature graphs and then uses GMN to detect supply chain vulnerabilities based on these graphs. Results: The experiments show that PG-VulNet achieves an average detection accuracy of 99.14%, significantly higher than existing approaches like Gemini, VulSeeker, FIT, and Asteria. In addition to this, PG-VulNet also excels in detection overhead and false alarms. In the real-world evaluation, PG-VulNet detected 690 known vulnerabilities in 1,611 firmwares. Conclusions: PG-VulNet can effectively detect the vulnerabilities introduced by software supply chain in IoT firmwares and is well suited for large-scale detection. Compared with existing approaches, PG-VulNet has significant advantages.

With the development of IoT technology, more and more IoT devices are connected to the network. Due to the hardware constraints of IoT devices themselves, it is difficult for developers to embed security software into them. Therefore, it is better to protect IoT devices at the traffic level. The effect of malicious traffic detection based on neural networks is promising. Still, the slow computation brings some difficulties to deploying AI-based detection systems on edge servers. Time Convolutional Network (TCN) is a high-speed neural network suitable for massively parallel computation. In this paper, we propose Multi-class S-TCN, an improved network supporting multiple classifications based on TCN for the practical needs of IoT scenarios. Besides, we implement a complete IoT traffic security detection procedure based on deep packet inspection and protocol analysis. The proposed Multi-class S-TCN significantly improves the detection speed without degrading the detection effect. Experiments show that this work has better detection performance and faster detection speed compared to existing approaches, proving the effectiveness of the proposed detection flow and Multi-class S-TCN in IoT scenarios.

Tibetan is a low-resource language with few existing electronic reference materials. The goal of Tibetan sentence boundary disambiguation (SBD) is to segment long text into sentences, and it is the foundation for downstream tasks corpora building. This study implemented the Tibetan SBD at the syllable level to avoid word segmentation (WS) errors affecting the accuracy of SBD. Specifically, the attention mechanism is introduced based on a recurrent neural network (RNN) to study Tibetan SBD. The primary objective is to determine, using a trained model, whether the shad contained in Tibetan text is the ending of the sentence, and implement experiments on syllable embedding and component embedding to measure the model's performance. The highest accuracy for Tibetan syllable embedding and component embedding is 96.23% and 95.40 %, respectively, and the F1 score reaches 96.23% and 95.37%, respectively. The experimental results demonstrate that the proposed method can achieve better results than the established rule-based and statistical methods without considering various syntactic and part-of-speech (POS) tagging rules. German and English data from the Europarl corpus and Thai data from the IWSLT2015 corpus are validated to prove the models’ reliability and generalizability. The results demonstrate that this method is efficient not only for low-resource languages but also for high-resource languages. More importantly, we can formally apply the experimental results of this study to the research of downstream tasks, such as machine translation and automatic summarization.

Edge computing (EC) is an emerging paradigm that extends cloud computing by pushing computing resources onto edge servers that are attached to base stations or access points at the edge of the cloud in close proximity with end-users. Due to edge servers’ geographic distribution, the EC paradigm is challenged by many new security threats, including the notorious distributed Denial-of-Service (DDoS) attack. In the EC environment, edge servers usually have constrained processing capacities due to their limited sizes. Thus, they are particularly vulnerable to DDoS attacks. DDoS attacks in the EC environment render existing DDoS mitigation approaches obsolete with its new characteristics. In this article, we make the first attempt to tackle the edge DDoS mitigation (EDM) problem. We model it as a constraint optimization problem and prove its$\mathcal {NP}$-hardness. To solve this problem, we propose an optimal approach named EDMOpti and a novel game-theoretical approach named EDMGame for mitigating edge DDoS attacks. EDMGame formulates the EDM problem as a potential EDM Game that admits a Nash equilibrium and employs a decentralized algorithm to find the Nash equilibrium as the solution to the EDM problem. Through theoretical analysis and experimental evaluation, we demonstrate that our approaches can solve the EDM problem effectively and efficiently.

Different from the existing train delay studies that had strived to explore sophisticated algorithms, this paper focuses on finding the bound of improvements on predicting multi-scenario train delays with different machine learning methods. Motivated by the observation of deep learning methods failing to improve the prediction performance if the delay occurs rarely, we present a novel augmented machine learning approach to improve the overall prediction accuracy further. Our solution proposes a rule-driven automation (RDA) method, including a delay status labeling (DSL) algorithm, and the resilience of section (RSE) and resilience of station (RST) indicators to generate the forecast for train delays. The experiment results demonstrate that the Random Forest based implementation of our RDA method (RF-RDA) can significantly improve the generalization ability of multivariate multi-step forecast models for multi-scenario train delay prediction. The proposed solution surpasses state-of-art baselines based on real-world traffic datasets, which treat various real-time delays differently. Even when the predictability of conventional deep learning methods decreases, the performance of our method is still acceptable for practical use to provide accurate forecasts.

In this paper, we propose a new traffic control method based on multiagent reinforcement learning and communication flow for autonomous vehicles and traffic lights. With the aim to ease traffic overload flow, traffic lights smartly tune the time of green light according to a crossroad situation. Beyond that, crossroad situation information can be transferred between traffic lights and autonomous vehicles. Due to the communication dispatch algorithm, autonomous vehicles can dynamically design new routes for avoiding traffic jams and traffic lights dynamically adjust to real-time traffic more efficiently. We demonstrate that our method outperforms the traditional traffic control method and provides high practicability in the future for autonomous vehicles.

Clothing is one of the symbols of human civilisation. Clothing design is an art form that combines practicality and artistry. The Dunhuang clothes culture has a long history which represents ancient Chinese aesthetics. Artificial intelligence (AI) technology has been recently applied to multiple areas, which is also drawing increasing attention in fashion. However, little research has been done on the usage of AI for the creation of clothing, especially in traditional culture. It is challenging that the exploration of computer science and Dunhuang clothing design, which is a cross-history interaction between AI and Chinese classical culture. In this paper, we propose ClothGAN, which is an innovative framework for “designing” new patterns and styles of clothes based on generative adversarial network (GAN) and style transfer algorithm. Besides, we built the Dunhuang clothes dataset and conducted experiments to generate new patterns and styles of clothes with Dunhuang elements. We evaluated these clothing works generated from different models by computing inception score (IS), human prefer score (HPS) and generated score (IS and HPS). The results show that our framework outperformed others in these designing works.

In video coding, although inter-channel redundancy has been somewhat decorrelated through the conversion of RGB-to-YUV, there are still some correlations among Y, Cb, and Cr components [1]. As a result, chroma components can be predicted from the luma component based on a linear model. The parameters of the linear model are estimated through a Least Square Solution (LSE) by using reconstructed neighboring chroma and the causal luma samples as training data at both the encoder and the decoder sides. However, it is difficult to derive the optimal parameters especially when there are insufficient training samples for prediction blocks with smaller sizes. To address this issue, we propose new method on deriving linear model and multi-model linear model based on up-sampling chroma samples (LM_UP and MMLM_UP) for chroma intra prediction. This proposed method performs consistently better than the state-of-the-art Joint Exploration Model version 7 (JEM-7.0) with an average of 0.06%, 0.31%, and 0.25% BD-rate reduction for Y, Cb, and Cr components, respectively, while the complexity increases at both encoder and decoder sides are negligible.

Learning path in online learning systems refers to a sequence of learning objects which are designated to help the students in improving their knowledge or skill in particular subjects or degree courses. In this paper, we review the recent research on learning path adaptation to pursue two goals, first is to organize and analyze the parameter of adaptation in learning path; the second is to discuss the challenges in implementing learning path adaptation. The survey covers the state of the art and aims at providing a comprehensive introduction to the learning path adaptation for researchers and practitioners.

Proview is probably the first Open Source system based on Linux for process control. It is modern, powerful and general. It contains all functions normally required for successful sequential control, adjustment, data acquisition, supervision, communication, etc. But it is not used widely yet. This paper proposes a new Soft PLC (Programmable Logic Controller) solution of automatic burdening system based on Proview. The system adopts advanced DCS (Distributed Control System) structure and PID control to get high precision. The focus is on the principle of process control and system structure in Proview, providing I/O configuration, hardware design circuit, and part of ladder program. Finally, verification is carried out with the performance test of the overall machine.

Fault-tolerance has huge impact on embedded safety-critical systems. As technology that assists to the development of such improvement, Safe Node Sequence Protocol (SNSP) is designed to make part of such impact. In this paper, we present a mechanism for fault-tolerance and recovery based on the Safe Node Sequence Protocol (SNSP) to strengthen the system robustness, from which the correctness of a fault-tolerant prototype system is analyzed and verified. In order to verify the correctness of more than thirty failure modes, we have partitioned the complete protocol state machine into several subsystems, followed to the injection of corresponding fault classes into dedicated independent models. Experiments demonstrate that this method effectively reduces the size of overall state space, and verification results indicate that the protocol is able to recover from the fault model in a fault-tolerant system and continue to operate as errors occur.

Service-oriented architecture provides a scalable and flexible framework to implement loosely-coupled, standards-based, and protocol-independent distribute computing. One of its goals is to make use of the distributed services with different functions to build powerful composite services. Service composition is an active research area in service computing. Existing research is endeavoring to achieve desirable quality levels of composite services and improve customer satisfaction. The service-oriented approach using Web services is also of great interest for the implementation of data-intensive processes such as data mining, image processing and so on. The applications based on data-intensive services have become the most challenging type of applications in service-oriented architecture. The service-based strategy provides maximal flexibility when designing data-intensive applications. Huge data sets that may each be replicated in different data centers have to be exchanged between several services. The movement of mass data influences the performance of the whole application process. Especially, the price of services will be different when considering the data center's locations and the amount of data transferred. It is desirable to find the cost minimized service composition solution in service computing. Therefore, how to select appropriate data centers for accessing data replicas and how to select services with lowest associated costs are emerging problems when deploying and executing data-intensive service applications. In this paper, a cost minimizing service composition model for data-intensive applications is proposed. Furthermore, how bio-inspired algorithms offer advantages to solve such problems will be presented.

The emergence of Cloud Services and Mobile Internet has influenced the society a lot, including the tourism industry. This paper proposes a design of backpacking service, not only aimed at the travelling routines, but also focus on the low-cost airlines. This kind of service aids backpackers with an effective travelling and satisfy the price requirement. With this low-cost airlines system, the backpackers can experience real e-Tourism and enjoy a better travel aided by the real-time information. According to the basic principles of low-cost airlines, this paper provides the backpackers with a definitely efficient way to check out a suitable flight itinerary under an acceptable price.

In most cases of safety-related systems, the network is an indispensable part. At this point, the system reliability is as important as the system communication quality. With the emergence of multi-core architectures, the first generation usually aims to provide reliable and deterministic computing resources. Therefore, with the boost requirement of reliability and throughput that cannot be satisfied by general single-core processors, the deployment of safety-related systems is transferred and processed multi-core environments. In this paper, we propose Reliable Communication Server on SPU (RCSoS), which is a server model for reliable communication utilizing SPU (Synergistic Processor Unit) in Cell/B.E (Cell Broadband Engine). It simulates SPU as a communication server and guarantees the reliability and determinacy by the isolation mode of SPU and contract model. We have implemented RCSoS in PlayStation 3, which dynamically adjust parameters, and inform applications on contract violations. Experiments show the performance of this model.

Abstract Mental health care is a major cost to all EU nations, and in many case, it results in additional costs to a country's economy due to the lost of productivity and concerns at the societal level. Cognitive behaviour therapy (CBT) is by far the treatment of choice for mental disorders such as depression, anxiety disorders, eating disorders, chronic insomnia, etc. Although effective, CBT suffers from a lack of accessibility and personalisation. In this paper, we proposed an ontology‐enhanced ubiquitous monitoring and treatment model for the purpose of assisting people to overcome the challenges of mental disorders. We first presented a context ontology for mental disorders as the basis upon which semantics‐enhanced methods are developed for gathering, formalising and manipulating patients' data. We implemented the proposed framework to facilitate online CBT, for treating depression at the current stage, which combines together talk/chat/messaging services, helps in retrieving neurofeedback, and supports collaborative diagnosis when necessary. An online statistic report was also integrated into our system. Finally, the paper discussed the proposed framework by comparing it against relevant research in the field and elaborated on possible further research directions. Copyright © 2009 John Wiley & Sons, Ltd.

Abstract In this paper, a novel multimedia QoS monitor, p3m (Portable MultiMedia Monitor), is presented. The main function of p3m is data package analysis and statistics. Breaking through the conventional method, this new QoS monitor moves jobs into kernel space, by which to meet the real‐time requirements of multimedia services. With the development of broadband Internet, more and more heterogeneous multimedia services and protocols have been widely deployed. As known, there are two traditional QoS monitors. One is server‐based, depending on the specific multimedia server, such as Helix. Another is tcpdump‐based, e.g. mmdump, which can measure the nature of multimedia traffic merely. However, it is still difficult to monitor the QoS of multimedia networks by these two traditional strategies. To resolve the problems, our new system adopts MMPM (MultiMedia Packages Management) mechanism which can make monitor tool become server‐independent and monitor the multimedia packets by preference, thus greatly reducing resource requirements with offering high performance. This paper describes the design and implementation of p3m and demonstrates one instance of its utility in monitoring the QoS of RTSP with Helix Server. The evaluation of p3m is also presented in this paper. Copyright © 2009 John Wiley & Sons, Ltd.

A novel network communication model based on Cell, which is a multi-core processor system, is presentment. And the I/O data transmission performance of the novel system will be showed in the paper too. In order to meet the high computation capacity request, the multicore platform has been adopted in the software radio system. But the research result presents that the data I/O transmission becomes the bottleneck for the software radio system. To resolve the problem, the paper redesign the network model based on Cell system, moving the protocol stack and net device driver components to SPU from PPU. The novel model cannot only release the PPU resource, but also has efficient I/O transmission performance.

This letter presents a study of the correlation between the eigenvalue spectra of synaptic matrices and the dynamical properties of asymmetric neural networks with associative memories. For this type of neural network, it was found that there are essentially two different dynamical phases: the chaos phase, with almost all trajectories converging to a single chaotic attractor, and the memory phase, with almost all trajectories being attracted toward fixed-point attractors acting as memories. We found that if a neural network is designed in the chaos phase, the eigenvalue spectrum of its synaptic matrix behaves like that of a random matrix (i.e., all eigenvalues lie uniformly distributed within a circle in the complex plan), and if it is designed in the memory phase, the eigenvalue spectrum will split into two parts: one part corresponds to a random background, the other part equal in number to the memory attractors. The mechanism for these phenomena is discussed in this letter.