RoGRUT: A Hybrid Deep Learning Model for Detecting Power Trapping in Smart Grids

Electricity theft is a widespread non-technical issue that has a negative impact on both power grids and electricity users.It hinders the economic growth of utility companies,poses electrical risks,and impacts the high energy costs borne by consumers.The development of smart grids is crucial for the identification of power theft since these systems create enormous amounts of data,including information on client consumption,which may be used to identify electricity theft using machine learning and deep learning techniques.Moreover,there also exist different solutions such as hardware-based solutions to detect electricity theft that may require human resources and expensive hardware.Computer-based solutions are presented in the literature to identify electricity theft but due to the dimensionality curse,class imbalance issue and improper hyper-parameter tuning of such models lead to poor performance.In this research,a hybrid deep learning model abbreviated as RoGRUT is proposed to detect electricity theft as amalicious and non-malicious activity.The key steps of the RoGRUT are data preprocessing that covers the problem of class imbalance,feature extraction and final theft detection.Different advanced-level models like RoBERTa is used to address the curse of dimensionality issue,the near miss for class imbalance,and transfer learning for classification.The effectiveness of the RoGRUTis evaluated using the dataset fromactual smartmeters.A significant number of simulations demonstrate that,when compared to its competitors,the RoGRUT achieves the best classification results.The performance evaluation of the proposed model revealed exemplary results across variousmetrics.The accuracy achieved was 88%,with precision at an impressive 86%and recall reaching 84%.The F1-Score,a measure of overall performance,stood at 85%.Furthermore,themodel exhibited a noteworthyMatthew correlation coefficient of 78%and excelled with an area under the curve of 91%.

Image Splicing Forgery Detection Using Feature-Based of Sonine Functions and Deep Features

The growing prevalence of fake images on the Internet and social media makes image integrity verification a crucial research topic.One of the most popular methods for manipulating digital images is image splicing,which involves copying a specific area from one image and pasting it into another.Attempts were made to mitigate the effects of image splicing,which continues to be a significant research challenge.This study proposes a new splicing detectionmodel,combining Sonine functions-derived convex-based features and deep features.Two stages make up the proposed method.The first step entails feature extraction,then classification using the“support vector machine”(SVM)to differentiate authentic and spliced images.The proposed Sonine functions-based feature extraction model reveals the spliced texture details by extracting some clues about the probability of image pixels.The proposed model achieved an accuracy of 98.93% when tested with the CASIA V2.0 dataset“Chinese Academy of Sciences,Institute of Automation”which is a publicly available dataset for forgery classification.The experimental results show that,for image splicing forgery detection,the proposed Sonine functions-derived convex-based features and deep features outperform state-of-the-art techniques in terms of accuracy,precision,and recall.Overall,the obtained detection accuracy attests to the benefit of using the Sonine functions alongside deep feature representations.Finding the regions or locations where image tampering has taken place is limited by the study.Future research will need to look into advanced image analysis techniques that can offer a higher degree of accuracy in identifying and localizing tampering regions.

Deep Learning Based Cyber Event Detection from Open-Source Re-Emerging Social Data

Social media forums have emerged as the most popular form of communication in the modern technology era,allowing people to discuss and express their opinions.This increases the amount of material being shared on social media sites.There is a wealth of information about the threat that may be found in such open data sources.The security of already-deployed software and systems relies heavily on the timely detection of newly-emerging threats to their safety that can be gleaned from such information.Despite the fact that several models for detecting cybersecurity events have been presented,it remains challenging to extract security events from the vast amounts of unstructured text present in public data sources.The majority of the currently available methods concentrate on detecting events that have a high number of dimensions.This is because the unstructured text in open data sources typically contains a large number of dimensions.However,to react to attacks quicker than they can be launched,security analysts and information technology operators need to be aware of critical security events as soon as possible,regardless of how often they are reported.This research provides a unique event detection method that can swiftly identify significant security events from open forums such as Twitter.The proposed work identified new threats and the revival of an attack or related event,independent of the volume of mentions relating to those events on Twitter.In this research work,deep learning has been used to extract predictive features from open-source text.The proposed model is composed of data collection,data transformation,feature extraction using deep learning,Latent Dirichlet Allocation(LDA)based medium-level cyber-event detection and final Google Trends-based high-level cyber-event detection.The proposed technique has been evaluated on numerous datasets.Experiment results show that the proposed method outperforms existing methods in detecting cyber events by giving 95.96% accuracy.

Latent Space Representational Learning of Deep Features for Acute Lymphoblastic Leukemia Diagnosis

Acute Lymphoblastic Leukemia(ALL)is a fatal malignancy that is featured by the abnormal increase of immature lymphocytes in blood or bone marrow.Early prognosis of ALL is indispensable for the effectual remediation of this disease.Initial screening of ALL is conducted through manual examination of stained blood smear microscopic images,a process which is time-consuming and prone to errors.Therefore,many deep learning-based computer-aided diagnosis(CAD)systems have been established to automatically diagnose ALL.This paper proposes a novel hybrid deep learning system for ALL diagnosis in blood smear images.The introduced system integrates the proficiency of autoencoder networks in feature representational learning in latent space with the superior feature extraction capability of standard pretrained convolutional neural networks(CNNs)to identify the existence of ALL in blood smears.An augmented set of deep image features are formed from the features extracted by GoogleNet and Inception-v3 CNNs from a hybrid dataset of microscopic blood smear images.A sparse autoencoder network is designed to create an abstract set of significant latent features from the enlarged image feature set.The latent features are used to perform image classification using Support Vector Machine(SVM)classifier.The obtained results show that the latent features improve the classification performance of the proposed ALL diagnosis system over the original image features.Moreover,the classification performance of the system with various sizes of the latent feature set is evaluated.The retrieved results reveal that the introduced ALL diagnosis system superiorly compete the state of the art.

IoT-Based Women Safety Gadgets (WSG): Vision, Architecture, and Design Trends

In recent years,the growth of female employees in the commercial market and industries has increased.As a result,some people think travelling to distant and isolated locations during odd hours generates new threats to women’s safety.The exponential increase in assaults and attacks on women,on the other hand,is posing a threat to women’s growth,development,and security.At the time of the attack,it appears the women were immobilized and needed immediate support.Only self-defense isn’t sufficient against abuse;a new technological solution is desired and can be used as quickly as hitting a switch or button.The proposed Women Safety Gadget(WSG)aims to design a wearable safety device model based on Internet-of-Things(IoT)and Cloud Technology.It is designed in three layers,namely layer-1,having an android app;layer-2,with messaging and location tracking system;and layer-3,which updates information in the cloud database.WSG can detect an unsafe condition by the pressure sensor of the finger on the artificial nail,consequently diffuses a pepper spray,and automatically notifies the saved closest contacts and police station through messaging and location settings.WSG has a response time of 1000 ms once the nail is pressed;the average time for pulse rate measure is 0.475 s,and diffusing the pepper spray is 0.2–0.5 s.The average activation time is 2.079 s.

Type 2 Diabetes Risk Prediction Using Deep Convolutional Neural Network Based-Bayesian Optimization

Diabetes mellitus is a long-term condition characterized by hyperglycemia.It could lead to plenty of difficulties.According to rising morbidity in recent years,the world’s diabetic patients will exceed 642 million by 2040,implying that one out of every ten persons will be diabetic.There is no doubt that this startling figure requires immediate attention from industry and academia to promote innovation and growth in diabetes risk prediction to save individuals’lives.Due to its rapid development,deep learning(DL)was used to predict numerous diseases.However,DLmethods still suffer from their limited prediction performance due to the hyperparameters selection and parameters optimization.Therefore,the selection of hyper-parameters is critical in improving classification performance.This study presents Convolutional Neural Network(CNN)that has achieved remarkable results in many medical domains where the Bayesian optimization algorithm(BOA)has been employed for hyperparameters selection and parameters optimization.Two issues have been investigated and solved during the experiment to enhance the results.The first is the dataset class imbalance,which is solved using Synthetic Minority Oversampling Technique(SMOTE)technique.The second issue is the model’s poor performance,which has been solved using the Bayesian optimization algorithm.The findings indicate that the Bayesian based-CNN model superbases all the state-of-the-art models in the literature with an accuracy of 89.36%,F1-score of 0.88.6,andMatthews Correlation Coefficient(MCC)of 0.88.6.

Risk Prevention and Control for Agricultural Non-Point Source Pollution Based on the Process of Pressure-Transformation-Absorption in Chongqing, China

According to China’s second national survey of pollution sources, the contribution of agricultural non-point sources(ANS) to water pollution is still high. Risk prevention and control are the main means to control costs and improve the efficiency of ANS, but most studies directly take pollution load as the risk standard, leading to a considerable misjudgment of the actual pollution risk. To objectively reflect the risk of agricultural non-point source pollution(ANSP) in Chongqing, China, we investigated the influences of initial source input, intermediate transformation, and terminal absorption of pollutants via literature research and the Delphi method and built a PTA(pressure kinetic energy, transformation kinetic energy, and absorption kinetic energy) model that covers 12 factors, with the support of geographical information system(GIS) technology. The terrain factor calculation results and the calculation results of other factors were optimized by Python tools to reduce human error and workload. Via centroid migration analysis and Kernel density analysis, the risk level, spatial aggregation degree, and key prevention and control regions could be accurately determined. There was a positive correlation between the water quality of the rivers in Chongqing and the risk assessment results of different periods, indirectly reflecting the reliability of the assessment results by the proposed model. There was an obvious tendency for the low-risk regions transforming into high-risk regions. The proportion of high-risk regions and extremely high-risk regions increased from 17.82% and 16.63%in 2000 to 18.10% and 16.76% in 2015, respectively. And the risk level in the main urban areas was significantly higher than that in the southeastern and northeastern areas of Chongqing. The centroids of all grades of risky areas presented a successive distribution from west to east, and the centroids of high-risk and extremely high-risk regions shifted eastward. From 2000 to 2015, the centroids of highrisk and extremely high-risk regions moved 4.63 km(1.68°) and 4.48 km(12.08°) east by north, respectively. The kernel density analysis results showed that the high-risk regions were mainly concentrated in the main urban areas and that the distribution of agglomeration areas overall displayed a transition trend from contiguous distribution to decentralized concentration. The risk levels of the regions with a high proportion of cultivated land and artificial surface were significantly increased, and the occupation of cultivated land in the process of urbanization promoted the movement of the centroids of high-risk and extremely high-risk regions. The identification of key areas for risk prevention and control provides data scientific basis for the development of prevention and control strategies.

Root-Of-Trust for Continuous Integration and Continuous Deployment Pipeline in Cloud Computing

Cloud computing has gained significant use over the last decade due to its several benefits,including cost savings associated with setup,deployments,delivery,physical resource sharing across virtual machines,and availability of on-demand cloud services.However,in addition to usual threats in almost every computing environment,cloud computing has also introduced a set of new threats as consumers share physical resources due to the physical co-location paradigm.Furthermore,since there are a growing number of attacks directed at cloud environments(including dictionary attacks,replay code attacks,denial of service attacks,rootkit attacks,code injection attacks,etc.),customers require additional assurances before adopting cloud services.Moreover,the continuous integration and continuous deployment of the code fragments have made cloud services more prone to security breaches.In this study,the model based on the root of trust for continuous integration and continuous deployment is proposed,instead of only relying on a single signon authentication method that typically uses only id and password.The underlying study opted hardware security module by utilizing the Trusted Platform Module(TPM),which is commonly available as a cryptoprocessor on the motherboards of the personal computers and data center servers.The preliminary proof of concept demonstrated that the TPM features can be utilized through RESTful services to establish the root of trust for continuous integration and continuous deployment pipeline and can additionally be integrated as a secure microservice feature in the cloud computing environment.

Secure Rotation Invariant Face Detection System for Authentication

Biometric applications widely use the face as a component for recognition and automatic detection.Face rotation is a variable component and makes face detection a complex and challenging task with varied angles and rotation.This problem has been investigated,and a novice algorithm,namely RIFDS(Rotation Invariant Face Detection System),has been devised.The objective of the paper is to implement a robust method for face detection taken at various angle.Further to achieve better results than known algorithms for face detection.In RIFDS Polar Harmonic Transforms(PHT)technique is combined with Multi-Block Local Binary Pattern(MBLBP)in a hybrid manner.The MBLBP is used to extract texture patterns from the digital image,and the PHT is used to manage invariant rotation characteristics.In this manner,RIFDS can detect human faces at different rotations and with different facial expressions.The RIFDS performance is validated on different face databases like LFW,ORL,CMU,MIT-CBCL,JAFFF Face Databases,and Lena images.The results show that the RIFDS algorithm can detect faces at varying angles and at different image resolutions and with an accuracy of 99.9%.The RIFDS algorithm outperforms previous methods like Viola-Jones,Multi-blockLocal Binary Pattern(MBLBP),and Polar HarmonicTransforms(PHTs).The RIFDS approach has a further scope with a genetic algorithm to detect faces(approximation)even from shadows.

Use of artificial neural networks to identify and analyze polymerized actin-based cytoskeletal structures in 3D confocal images

Background:Living cells need to undergo subtle shape adaptations in response to the topography of their substrates.These shape changes are mainly determined by reorganization of their internal cytoskeleton,with a major contribution from filamentous(F)actin.Bundles of F-actin play a major role in determining cell shape and their interaction with substrates,either as“stress fibers,”or as our newly discovered“Concave Actin Bundles”(CABs),which mainly occur while endothelial cells wrap micro-fibers in culture.Methods:To better understand the morphology and functions of these CABs,it is necessary to recognize and analyze as many of them as possible in complex cellular ensembles,which is a demanding and time-consuming task.In this study,we present a novel algorithm to automatically recognize CABs without further human intervention.We developed and employed a multilayer perceptron artificial neural network(“the recognizer”),which was trained to identify CABs.Results:The recognizer demonstrated high overall recognition rate and reliability in both randomized training,and in subsequent testing experiments.Conclusion:It would be an effective replacement for validation by visual detection which is both tedious and inherently prone to errors.

Systematic analysis of artificial intelligence in the era of industry 4.0

Artificial Intelligence has been playing a profound role in the global economy,social progress,and people’s daily life.With the increasing capabilities and accuracy of AI,the application of AI will have more impacts on manufacturing and service areas in the era of industry 4.0.This study conducts a systematic literature review to study the state-of-the-art on AI in industry 4.0.This paper describes the development of industries and the evolution of AI.This paper also identifies that the development and application of AI will bring not only opportunities but also challenges to industry 4.0.The findings provide a valuable reference for researchers and practitioners through a multi-angle systematic analysis of AI.In the era of industry 4.0,AI system will become an innovative and revolutionary assistance to the whole industry.

Evaluation of Spatiotemporal Dynamics of Simulated Land Use/Cover in China Using a Probabilistic Cellular Automata-Markov Model

Using the fuzzy rule-based classification method, normalized difference vegetation index (NDVI) images acquired from 1982 to 1998 were classified into seventeen phases. Based on these classification images, a probabilistic cellular automata-Markov Chain model was developed and used to simulate a land cover scenario of China for the year 2014. Spatiotemporal dynamics of land use/cover in China from 1982 to 2014 were then analyzed and evaluated. The results showed that the change trends of land cover type from 1998 to 2014 would be contrary to those from 1982 to 1998. In particular, forestland and grassland areas decreased by 1.56% and 1.46%, respectively, from 1982 to 1998, and should increase by 1.5% and 2.3% from 1998 to 2014, respectively.

An Integrated Incentive Framework for Mobile Crowdsourced Sensing

Currently, mobile devices （e.g., smartphones） are equipped with multiple wireless interfaces and rich builtin functional sensors that possess powerful computation and communication capabilities, and enable numerous Mobile Crowdsourced Sensing （MCS） applications. Generally, an MCS system is composed of three components： a publisher of sensing tasks, crowd participants who complete the crowdsourced tasks for some kinds of rewards, and the crowdsourcing platform that facilitates the interaction between publishers and crowd participants. Incentives are a fundamental issue in MCS. This paper proposes an integrated incentive framework for MCS, which appropriately utilizes three widely used incentive methods： reverse auction, gamification, and reputation updating. Firstly, a reverse-auction-based two-round participant selection mechanism is proposed to incentivize crowds to actively participate and provide high-quality sensing data. Secondly, in order to avoid untruthful publisher feedback about sensing-data quality, a gamification-based verification mechanism is designed to evaluate the truthfulness of the publisher＇s feedback. Finally, the platform updates the reputation of both participants and publishers based on their corresponding behaviors. This integrated incentive mechanism can motivate participants to provide high-quality sensed contents, stimulate publishers to give truthful feedback, and make the platform profitable.

THS-GWNN:a deep learning framework for temporal network link prediction

1 Introduction and main contributions Link prediction for temporal networks aims to evaluate the likelihood of the future linkage among nodes,which has significant applications in social networks,biological networks and traffic analysis[1],etc.Network embedding[2]is an important analytical tool for temporal network link prediction,which helps us better understand network evolution[3].How to encode high-dimensional and non-Euclidean network information is a crucial problem for node embedding in temporal networks.One of the challenges is to reveal the spatial structure at each timestamp and the temporal property over time[4].Some existing work[5]shows that extracting the spatial relation of each node can be used as a valid feature representation for each node.Moreover,the emergence of deep learning techniques[4,5]brings new insights for learning temporal properties,but most models using deep learning still fail to achieve satisfying prediction accuracy.

Probabilistic synthesis against GR(1) winning condition

Reactive synthesis is a technique for automatic generation of a reactive system from a high level specification.The system is reactive in the sense that it reacts to the inputs from the environment.The specification is in general given as a linear temporal logic(LTL)formula.The behaviour of the system interacting with the environment can be represented as a game in which the system plays against the environment.Thus,a problem of reactive synthesis is commonly treated as solving such a game with the specification as the winning condition.Reactive synthesis has been thoroughly investigated for more two decades.A well-known challenge is to deal with the complex uncertainty of the environment.We understand that a major issue is due to the lack of a sufficient treatment of probabilistic properties in the traditional models.For example,a two-player game defined by a standard Kriple structure does not consider probabilistic transitions in reaction to the uncertain physical environment;and a Markov Decision Process(MDP)in general does not explicitly separate the system from its environment and it does not describe the interaction between system and the environment.In this paper,we propose a new and more general model which combines the two-player game and the MDP.Furthermore,we study probabilistic reactive synthesis for the games of General Reactivity of Rank 1(i.e.,GR(1))defined in this model.More specifically,we present an algorithm,which for given model,a location and a GR(1)specification,determines the strategy for each player how to maximize/minimize the probabilities of the satisfaction of at location.We use an example to describe the model of probabilistic games and demonstrate our algorithm.