Deep neural networks (DNNs) have been widely adopted but they are vulnerable to intentionally crafted adversarial examples. Various attack methods against DNNs proposed, yet there still lacks theoretical explanation of In this paper, we aim understand examples from the attacking process and assume adding perturbations key/sensitive regions image could fool classification DNNs. We propose gradient shielding verify assumption which ignores insensitive information during generating...

The COVID-19 pandemic has caused serious consequences in the last few months and trying to control it been most important objective. With effective prevention methods, epidemic gradually under some countries is essential ensure safe work resumption future. Although approaches are proposed measure people's healthy conditions, such as filling health information forms or evaluating travel records, they cannot provide a fine-grained assessment of risk. In this paper, we propose novel risk method...

Speaker recognition is a task that identifies the speaker from multiple audios. Recently, advances in deep learning have considerably boosted development of speech signal processing techniques. or has been widely adopted such applications as smart locks, vehicle-mounted systems, and financial services. However, neural network-based systems (SRSs) are susceptible to adversarial attacks, which fool system make wrong decisions by small perturbations, this drawn attention researchers security...

With the rapid development of information technologies, security cyberspace has become increasingly serious. Network intrusion detection is a practical scheme to protect network systems from cyber attacks. However, as new vulnerabilities and unknown attack types are constantly emerging, only few samples such attacks can be captured for analysis, which cannot handled by existing methods deployed in real systems. To handle this problem, we propose few-shot class-incremental learning method...

In order to cope with ever-evolving and increasing cyber threats, intrusion detection systems have become a crucial component of security. Compared signature-based methods, anomaly-based methods typically employ machine learning techniques train models possess the capability discover unknown attacks. However, face challenge low rates for minority class attacks due imbalanced data distributions. Traditional algorithms address this issue by resampling or generating synthetic data....

The recent advancement of the Internet Things (IoT) and information technology has led to rapid expansion interconnectivity among a billion devices across various applications. advent massive data resulted in greater computational dependence, posing obstacles applying security policies energy-sensitive devices. However, public-key-based encryption algorithms are impractical or impossible execute on these resource-limited terminals. In this paper, we propose lightweight framework called...

A plethora of healthcare data is produced every day due to the proliferation prominent technologies such as Internet Medical Things (IoMT). Digital-driven smart devices like wearable watches, wristbands and bracelets are utilized extensively in modern applications. Mining valuable information from distributed at owners' level useful, but it challenging preserve privacy. Federated learning (FL) has swiftly surged popularity its efficacy dealing privacy vulnerabilities. Recent studies have...

This paper addresses the challenges of data sparsity and personalization limitations inherent in current recommendation systems when processing extensive academic datasets. To overcome these issues, present work introduces an innovative model that integrates wealth structured information from knowledge graphs refines amalgamation temporal relational data. By applying attention mechanisms neural network technologies, thoroughly explores text characteristics papers evolving patterns user...

A digital twin (DT) is an electronic replica of a real-world item. It built on top asset-specific data items and often enhanced using semantic technologies simulation environments. The DT lays the way for anything from routine monitoring to hands-off administration physical entity. With development metaverse concept gains importance. As it helps manage entity in metaverse. Therefore, beneficial use detection mitigation different types cyber attacks. In this context, we identification DDoS...

With the rapid development of information technologies, more and cyberattacks are emerging to cause serious consequences critical infrastructures in industrial cyber-physical systems. As becoming complicated, which might be composed by multiple steps, obtaining attack strategies can help understand better defend these attacks. However, there many unknown every day, while attackers will not reveal steps tools normally, it is a persistent challenging problem obtain strategies. Cyber range...

The license plate recognition system (LPRS) has been widely adopted in daily life due to its efficiency and high accuracy. Deep neural networks are commonly used the LPRS improve However, researchers have found that deep their own security problems may lead unexpected results. Specifically, they can be easily attacked by adversarial examples generated adding small perturbations original images, resulting incorrect recognition. There some classic methods generate examples, but cannot on...

A lattice basis reduction multisequence synthesis (LBRMS) algorithm was presented with a new, intuitive and vector-form model. In this paper, refined version of the LBRMS is deduced, from which Massey's conjectured can be derived. Moreover, if we modify used in algorithm, new equivalent to generalization Berlekamp-Massey proposed by Feng Tzeng (1989). Therefore, provides unified approach shift-register problem.

Neural network models have proved capable of achieving remarkable performance in sentence and document modeling. Convolutional neural networks (CNNs) recurrent (RNNs), two mainstream architectures for such modeling tasks, adopt totally different ways understanding natural languages. The classical CNN, despite its wide application image classification, is rarely used text classification. RNN processing texts variable lengths can hence facilitate In this study, we aim to analyze the both on...

The experimental results of this paper showed that the final grades most students would be improved in class with low English proficiency case textbooks same difficulty. However, when multimodal instruction was used, failure rate second semester decreased by 13 percentage points compared first semester, while high-achieving increased 6 points. Therefore, using teaching method, learning effect from mid-term to end term has been significantly improved, which method more superior than...

Abstract We compare several income distributions in urban China the late 1980s and mid-1990s using tests for stochastic dominance order to decompose gender differentials. Examination of entire distribution gives insight into uniformity such differentials across distribution. Moreover, based on allow robust welfare comparisons. Our analysis reveals: (i) large increasing predicted earnings lower tail distribution, but few differences upper tail, (ii) discrimination explains one-third one-half...

The development of ICT following Moore's Law has resulted in a situation where mobile users are able to make use wealth versatile services. Yet, battery and electronics technology not followed this as well, which user's can only enable few hours active use. Therefore, we need focus increasingly on energy efficient wireless communication reduce consumption, but also cut down greenhouse emissions improve business competitiveness. Due significant consumption transmitting data over networks,...