The purpose of a network intrusion detection (NID) is to detect intrusions in the network, which plays critical role ensuring security Internet Things (IoT). Recently, deep learning (DL) has achieved great success field detection. However, limited computing capabilities and storage IoT devices hinder actual deployment DL-based high-complexity models. In this article, we propose novel NID method for based on lightweight neural (LNN). data preprocessing stage, avoid high-dimensional raw...

The proliferation of wireless sensor networks (WSNs) and their applications has attracted remarkable growth in unsolicited intrusions security threats, which disrupt the normal operations WSNs. Deep learning (DL)-based network intrusion detection (NID) methods have been widely investigated developed. However, high computational complexity DL seriously hinders actual deployment DL-based model, particularly devices WSNs that do not powerful processing performance due to power limitation. In...

Federated learning (FL) has become an increasingly popular solution for intrusion detection to avoid data privacy leakage in Internet of Things (IoT) edge devices. Existing FL-based methods, however, suffer from three limitations: 1) model parameters transmitted each round may be used recover private data, which leads security risks; 2) not independent and identically distributed (non-IID) seriously adversely affect the training FL (especially distillation-based FL); 3) high communication...

Traffic classification is a critical task in network security and management. Recent research has demonstrated the effectiveness of deep learning-based traffic method. However, following limitations remain: (1) representation simply generated from raw packet bytes, resulting absence important information; (2) model structure directly applying learning algorithms does not take characteristics into account; (3) scenario-specific classifier training usually requires labor-intensive...

With the development of Industrial Internet Things (IIoT), complex traffic generated by large-scale IIoT devices presents challenges for analysis. Most existing deep learning-based analysis methods use a single flow classification, resulting in being misled irrelevant flow. Thus, it is necessary to sequences However, models fail effectively distinguish unimportant flows sequence, which affects classification performance. To address aforementioned challenges, we propose novel classifier...

With the increasing demand for protection of personal network meta-data, encrypted networks have grown in popularity, so do challenge monitoring and analyzing traffic. Currently, some deep learning-based methods been proposed to leverage statistical features traffic classification, which are barely affected by encryption techniques. However, these works still suffer from two main intrinsic limitations: (1) feature extraction process lacks a mechanism take into account correlations between...

The pervasive deployment of the Internet Things (IoT) has significantly facilitated manufacturing and living. diversity continual updates IoT systems make their security a crucial challenge, among which detection malicious network traffic turns out to be most common yet destructive threat. Despite efforts on feature engineering classification backend designing, established intrusion sometimes lack robustness are inflexible against shift distribution. To deal with these disadvantages, we...

With the increased need for clean water and decreased supply of fresh water, purification contaminated can provide that supplements from nature. Physical adsorption, photo-degradation solar-driven evaporation processes are promising ways to generate with minimum environmental impact. However, different process requires dispersion state treating agent maximize its performance. Herein, we demonstrate Fe3O4-reduced graphene oxide (MrGO) particles switchable be used as multifunctional agents...

Backdoor attacks have emerged as an urgent threat to Deep Neural Networks (DNNs), where victim DNNs are furtively implanted with malicious neurons that could be triggered by the adversary. To defend against backdoor attacks, many works establish a staged pipeline remove backdoors from DNNs: inspecting, locating, and erasing. However, in scenario few clean data can accessible, such is fragile cannot erase completely without sacrificing model accuracy. address this issue, paper, we propose...

With the ubiquitous network applications and continuous development of attack technology, all social circles have paid close attention to cyberspace security. Intrusion detection systems (IDS) plays a very important role in ensuring computer communication Recently, deep learning has achieved great success field intrusion detection. However, high computational complexity poses major hurdle for practical deployment DL-based models. In this paper, we propose novel approach based on lightweight...

Identifying anonymity services from network traffic is a crucial task for management and security. Currently, some works based on deep learning have achieved excellent performance analysis, especially those flow sequence (FS), which utilizes information features of the flow. However, these models still face serious challenge because lacking mechanism to take into account relationships between flows, resulting in mistakenly recognizing irrelevant flows FS as clues identifying traffic. In this...

When a liquid droplet is placed on sufficiently hot surface, it will be levitated by the vapor cushion between and surface due to Leidenfrost effect. Such Leidenfrost-based levitation can greatly reduce friction thus offers promising approach for low-friction devices. In this work, we demonstrated self-propelled rotational rotor made of wet paper with asymmetric mass distribution. The has shown capability reaching angular velocities more than 30 rad/s prolonged rotation duration through...

Container technology has become a popular development that can conveniently accelerate building, running, and sharing applications. However, container image packaging collection of software usually lurks various defects threatening consumer safety, such as embedded malware, vulnerability, privacy leakage, etc. Moreover, developers users share images through centralized, public, massive repository (e.g., Docker Hub), which magnify the impact these security in fast-spreading way....

Traffic classification is a critical task in network security and management. Recent research has demonstrated the effectiveness of deep learning-based traffic method. However, following limitations remain: (1) representation simply generated from raw packet bytes, resulting absence important information; (2) model structure directly applying learning algorithms does not take characteristics into account; (3) scenario-specific classifier training usually requires labor-intensive...

Network intrusion detection (NID) is an important cyber security scheme to identify attacks in network traffic. Recent years, a large amount of studies try improve the accuracy NID by kinds deep learning approaches. However, these models always require lot calculation and space, which constitutes major hurdle practical implementation DL-based models. Thus, lightweight model imperative, but there are very few applications algorithms In this paper, we propose knowledge distillation (LKD) for...

Although text-based captcha, which is used to differentiate between human users and bots, has faced many attack methods, it remains a widely security mechanism employed by some websites. Some deep learning-based text captcha solvers have shown excellent results, but the labor-intensive time-consuming labeling process severely limits their viability. Previous works attempted create easy-to-use using limited collection of labeled data. However, they are hampered inefficient preprocessing...

Electrocorticogram (ECoG) is an effective way for Epilepsy research, as well automatic seizure detection. This study proposes a method feature extraction and classification of pre-ictal ictal ECoGs, based upon mutual information (MI) support vector machine (SVM) which has not only high accuracy but also fast speed. First, the among 76 channels computed converted into 76×76 matrix, then statistical significance splicing between ECoGs tested, coefficients variation fluctuation indexes MI...

Supervising anonymity network is a critical issue in the field of security, and traditional traffic analysis methods cannot cope with complex traffic. In recent years, method based on deep learning has achieved good performance. However, most existing studies do not consider temporal-spatial correlation traffic, only use single flow for classification. A few works take continuous flows as sequence classification, but they distinguish different importance each flow. To tackle this issue, we...

To detect the start time and end of each action in an untrimmed video Track 3 AI City Challenge, this paper proposes a powerful network architecture, Multi-Attention Transformer. The previous methods extract features by setting fixed sliding window whitch means interval, predict times action. We believe that adopting series windows will corrupt feature containing contextual information. So we present transformer module which combines local attention global to fix problem. method equipped...

Federated learning (FL) has become an increasingly popular solution for intrusion detection to avoid data privacy leakage in Internet of Things (IoT) edge devices. However, most the current FL-based methods still suffer from three limitations: (1) model parameters transmitted each round may be used recover private which leads security risks, (2) not independent and identically distributed (non-IID) seriously adversely affects training FL (especially distillation-based FL), (3) high...

With the development of Industrial Internet Things (IIoT), more frequent attacks occur to intrude IIoT devices. A reasonably designed intrusion detection method can effectively guarantee security IIoT. Over past decade, different methods based on deep learning (DL) have been proposed, which helps keep evolving and become robust. However, these previous researches usually require participation a large number experts, gradually invalid with continuous methods. The limited compute capability...