Chained Dual-Generative Adversarial Network:A Generalized Defense Against Adversarial Attacks

Neural networks play a significant role in the field of image classification.When an input image is modified by adversarial attacks,the changes are imperceptible to the human eye,but it still leads to misclassification of the images.Researchers have demonstrated these attacks to make production self-driving cars misclassify StopRoad signs as 45 Miles Per Hour(MPH)road signs and a turtle being misclassified as AK47.Three primary types of defense approaches exist which can safeguard against such attacks i.e.,Gradient Masking,Robust Optimization,and Adversarial Example Detection.Very few approaches use Generative Adversarial Networks(GAN)for Defense against Adversarial Attacks.In this paper,we create a new approach to defend against adversarial attacks,dubbed Chained Dual-Generative Adversarial Network(CD-GAN)that tackles the defense against adversarial attacks by minimizing the perturbations of the adversarial image using iterative oversampling and undersampling using GANs.CD-GAN is created using two GANs,i.e.,CDGAN’s Sub-ResolutionGANandCDGAN’s Super-ResolutionGAN.The first is CDGAN’s Sub-Resolution GAN which takes the original resolution input image and oversamples it to generate a lower resolution neutralized image.The second is CDGAN’s Super-Resolution GAN which takes the output of the CDGAN’s Sub-Resolution and undersamples,it to generate the higher resolution image which removes any remaining perturbations.Chained Dual GAN is formed by chaining these two GANs together.Both of these GANs are trained independently.CDGAN’s Sub-Resolution GAN is trained using higher resolution adversarial images as inputs and lower resolution neutralized images as output image examples.Hence,this GAN downscales the image while removing adversarial attack noise.CDGAN’s Super-Resolution GAN is trained using lower resolution adversarial images as inputs and higher resolution neutralized images as output images.Because of this,it acts as an Upscaling GAN while removing the adversarial attak noise.Furthermore,CD-GAN has a modular design such that it can be prefixed to any existing classifier without any retraining or extra effort,and 2542 CMC,2023,vol.74,no.2 can defend any classifier model against adversarial attack.In this way,it is a Generalized Defense against adversarial attacks,capable of defending any classifier model against any attacks.This enables the user to directly integrate CD-GANwith an existing production deployed classifier smoothly.CD-GAN iteratively removes the adversarial noise using a multi-step approach in a modular approach.It performs comparably to the state of the arts with mean accuracy of 33.67 while using minimal compute resources in training.

Exploiting Human Pose and Scene Information for Interaction Detection

Identifying human actions and interactions finds its use in manyareas, such as security, surveillance, assisted living, patient monitoring, rehabilitation,sports, and e-learning. This wide range of applications has attractedmany researchers to this field. Inspired by the existing recognition systems,this paper proposes a new and efficient human-object interaction recognition(HOIR) model which is based on modeling human pose and scene featureinformation. There are different aspects involved in an interaction, includingthe humans, the objects, the various body parts of the human, and the backgroundscene. Themain objectives of this research include critically examiningthe importance of all these elements in determining the interaction, estimatinghuman pose through image foresting transform (IFT), and detecting the performedinteractions based on an optimizedmulti-feature vector. The proposedmethodology has six main phases. The first phase involves preprocessing theimages. During preprocessing stages, the videos are converted into imageframes. Then their contrast is adjusted, and noise is removed. In the secondphase, the human-object pair is detected and extracted from each image frame.The third phase involves the identification of key body parts of the detectedhumans using IFT. The fourth phase relates to three different kinds of featureextraction techniques. Then these features are combined and optimized duringthe fifth phase. The optimized vector is used to classify the interactions in thelast phase. TheMSRDaily Activity 3D dataset has been used to test this modeland to prove its efficiency. The proposed system obtains an average accuracyof 91.7% on this dataset.

Real Objects Understanding Using 3D Haptic Virtual Reality for E-Learning Education

In the past two decades,there has been a lot of work on computer vision technology that incorporates many tasks which implement basic filtering to image classification.Themajor research areas of this field include object detection and object recognition.Moreover,wireless communication technologies are presently adopted and they have impacted the way of education that has been changed.There are different phases of changes in the traditional system.Perception of three-dimensional(3D)from two-dimensional(2D)image is one of the demanding tasks.Because human can easily perceive but making 3D using software will take time manually.Firstly,the blackboard has been replaced by projectors and other digital screens so such that people can understand the concept better through visualization.Secondly,the computer labs in schools are now more common than ever.Thirdly,online classes have become a reality.However,transferring to online education or e-learning is not without challenges.Therefore,we propose a method for improving the efficiency of e-learning.Our proposed system consists of twoand-a-half dimensional(2.5D)features extraction using machine learning and image processing.Then,these features are utilized to generate 3D mesh using ellipsoidal deformation method.After that,3D bounding box estimation is applied.Our results show that there is a need to move to 3D virtual reality(VR)with haptic sensors in the field of e-learning for a better understanding of real-world objects.Thus,people will have more information as compared to the traditional or simple online education tools.We compare our result with the ShapeNet dataset to check the accuracy of our proposed method.Our proposed system achieved an accuracy of 90.77%on plane class,85.72%on chair class,and car class have 72.14%.Mean accuracy of our method is 70.89%.

A Novel Capability of Object Identification and Recognition Based on Integrated mWMM

In the last decade,there has been remarkable progress in the areas of object detection and recognition due to high-quality color images along with their depth maps provided by RGB-D cameras.They enable artificially intelligent machines to easily detect and recognize objects and make real-time decisions according to the given scenarios.Depth cues can improve the quality of object detection and recognition.The main purpose of this research study to find an optimized way of object detection and identification we propose techniques of object detection using two RGB-D datasets.The proposed methodology extracts image normally from depth maps and then performs clustering using the Modified Watson Mixture Model(mWMM).mWMM is challenging to handle when the quality of the image is not good.Hence,the proposed RGB-D-based system uses depth cues for segmentation with the help of mWMM.Then it extracts multiple features from the segmented images.The selected features are fed to the Artificial Neural Network(ANN)and Convolutional Neural Network(CNN)for detecting objects.We achieved 92.13%of mean accuracy over NYUv1 dataset and 90.00%of mean accuracy for the Redweb_v1 dataset.Finally,their results are compared and the proposed model with CNN outperforms other state-of-the-art methods.The proposed architecture can be used in autonomous cars,traffic monitoring,and sports scenes.

Tracking and Analysis of Pedestrian’s Behavior in Public Places

Crowd management becomes a global concern due to increased population in urban areas.Better management of pedestrians leads to improved use of public places.Behavior of pedestrian’s is a major factor of crowd management in public places.There are multiple applications available in this area but the challenge is open due to complexity of crowd and depends on the environment.In this paper,we have proposed a new method for pedestrian’s behavior detection.Kalman filter has been used to detect pedestrian’s usingmovement based approach.Next,we have performed occlusion detection and removal using region shrinking method to isolate occluded humans.Human verification is performed on each human silhouette and wavelet analysis and particle gradient motion are extracted for each silhouettes.Gray Wolf Optimizer(GWO)has been utilized to optimize feature set and then behavior classification has been performed using the Extreme Gradient(XG)Boost classifier.Performance has been evaluated using pedestrian’s data from avenue and UBI-Fight datasets,where both have different environment.The mean achieved accuracies are 91.3%and 85.14%over the Avenue and UBI-Fight datasets,respectively.These results are more accurate as compared to other existing methods.

A Pattern Classification Model for Vowel Data Using Fuzzy Nearest Neighbor

Classification of the patterns is a crucial structure of research and applications. Using fuzzy set theory, classifying the patterns has become of great interest because of its ability to understand the parameters. One of the problemsobserved in the fuzzification of an unknown pattern is that importance is givenonly to the known patterns but not to their features. In contrast, features of thepatterns play an essential role when their respective patterns overlap. In this paper,an optimal fuzzy nearest neighbor model has been introduced in which a fuzzifi-cation process has been carried out for the unknown pattern using k nearest neighbor. With the help of the fuzzification process, the membership matrix has beenformed. In this membership matrix, fuzzification has been carried out of the features of the unknown pattern. Classification results are verified on a completelyllabelled Telugu vowel data set, and the accuracy is compared with the differentmodels and the fuzzy k nearest neighbor algorithm. The proposed model gives84.86% accuracy on 50% training data set and 89.35% accuracy on 80% trainingdata set. The proposed classifier learns well enough with a small amount of training data, resulting in an efficient and faster approach.

Automatic Clustering of User Behaviour Profiles for Web Recommendation System

Web usage mining,content mining,and structure mining comprise the web mining process.Web-Page Recommendation(WPR)development by incor-porating Data Mining Techniques(DMT)did not include end-users with improved performance in the obtainedfiltering results.The cluster user profile-based clustering process is delayed when it has a low precision rate.Markov Chain Monte Carlo-Dynamic Clustering(MC2-DC)is based on the User Behavior Profile(UBP)model group’s similar user behavior on a dynamic update of UBP.The Reversible-Jump Concept(RJC)reviews the history with updated UBP and moves to appropriate clusters.Hamilton’s Filtering Framework(HFF)is designed tofilter user data based on personalised information on automatically updated UBP through the Search Engine(SE).The Hamilton Filtered Regime Switching User Query Probability(HFRSUQP)works forward the updated UBP for easy and accuratefiltering of users’interests and improves WPR.A Probabilistic User Result Feature Ranking based on Gaussian Distribution(PURFR-GD)has been developed to user rank results in a web mining process.PURFR-GD decreases the delay time in the end-to-end workflow for SE personalization in various meth-ods by using the Gaussian Distribution Function(GDF).The theoretical analysis and experiment results of the proposed MC2-DC method automatically increase the updated UBP accuracy by 18.78%.HFRSUQP enabled extensive Maximize Log-Likelihood(ML-L)increases to 15.28%of User Personalized Information Search Retrieval Rate(UPISRT).For feature ranking,the PURFR-GD model defines higher Classification Accuracy(CA)and Precision Ratio(PR)while uti-lising minimum Execution Time(ET).Furthermore,UPISRT's ranking perfor-mance has improved by 20%.

Alpha Fusion Adversarial Attack Analysis Using Deep Learning

The deep learning model encompasses a powerful learning ability that integrates the feature extraction,and classification method to improve accuracy.Convolutional Neural Networks(CNN)perform well in machine learning and image processing tasks like segmentation,classification,detection,identification,etc.The CNN models are still sensitive to noise and attack.The smallest change in training images as in an adversarial attack can greatly decrease the accuracy of the CNN model.This paper presents an alpha fusion attack analysis and generates defense against adversarial attacks.The proposed work is divided into three phases:firstly,an MLSTM-based CNN classification model is developed for classifying COVID-CT images.Secondly,an alpha fusion attack is generated to fool the classification model.The alpha fusion attack is tested in the last phase on a modified LSTM-based CNN(CNN-MLSTM)model and other pre-trained models.The results of CNN models show that the accuracy of these models dropped greatly after the alpha-fusion attack.The highest F1 score before the attack was achieved is 97.45 And after the attack lowest F1 score recorded is 22%.Results elucidate the performance in terms of accuracy,precision,F1 score and Recall.

On the Detection of COVID-19 from Chest X-Ray Images Using CNN-Based Transfer Learning

Coronavirus disease(COVID-19)is an extremely infectious disease and possibly causes acute respiratory distress or in severe cases may lead to death.There has already been some research in dealing with coronavirus using machine learning algorithms,but few have presented a truly comprehensive view.In this research,we show how convolutional neural network(CNN)can be useful to detect COVID-19 using chest X-ray images.We leverage the CNN-based pre-trained models as feature extractors to substantiate transfer learning and add our own classifier in detecting COVID-19.In this regard,we evaluate performance of five different pre-trained models with fine-tuning the weights from some of the top layers.We also develop an ensemble model where the predictions from all chosen pre-trained models are combined to generate a single output.The models are evaluated through 5-fold cross validation using two publicly available data repositories containing healthy and infected(both COVID-19 and other pneumonia)chest X-ray images.We also leverage two different visualization techniques to observe how efficiently the models extract important features related to the detection of COVID-19 patients.The models show high degree of accuracy,precision,and sensitivity.We believe that the models will aid medical professionals with improved and faster patient screening and pave a way to further COVID-19 research.

Customer Prioritization for Medical Supply Chain During COVID-19 Pandemic

During COVID-19,the escalated demand for various pharmaceutical products with the existing production capacity of pharmaceutical companies has stirred the need to prioritize its customers in order to fulfill their demand.This study considers a two-echelon pharmaceutical supply chain considering various pharma-distributors as its suppliers and hospitals,pharmacies,and retail stores as its customers.Previous studies have generally considered a balanced situation in terms of supply and demand whereas this study considers a special situation of COVID-19 pandemic where demand exceeds supply Various criteria have been identified from the literature that influences the selection of customers.A questionnaire has been developed to collect primary data from pharmaceutical suppliers pertaining to customerselection criteria.These criteria have been prioritized with respect to eigenvalues obtained from Principal Component Analysis and also validated with the experts’domain-related knowledge using Analytical Hierarchy Process.Profit potential appeared to be the most important criteria of customer selection followed by trust and service convenience brand loyalty,commitment,brand awareness,brand image,sustainable behavior,and risk.Subsequently,Multi Criteria Decision Analysis has been performed to prioritize the customerselection criteria and customers with respect to selection criteria.Three experts with seven and three and ten years of experience have participated in the study.Findings of the study suggest large hospitals,large pharmacies,and small retail stores are the highly preferred customers.Moreover,findings of prioritization of customer-selection criteria fromboth Principal Component Analysis and Analytical Hierarchy Process are consistent.Furthermore,this study considers the experience of three experts to calculate an aggregate score of priorities to reach an effective decision.Unlike traditional supply chain problems of supplier selection,this study considers a selection of customers and is useful for procurement and supply chain managers to prioritize customers while considering multiple selection criteria.

A New Handover Management Model for Two-Tier 5G Mobile Networks

There has been an exponential rise in mobile data traffic in recent times due to the increasing popularity of portable devices like tablets,smartphones,and laptops.The rapid rise in the use of these portable devices has put extreme stress on the network service providers while forcing telecommunication engineers to look for innovative solutions to meet the increased demand.One solution to the problem is the emergence of fifth-generation(5G)wireless communication,which can address the challenges by offering very broad wireless area capacity and potential cut-power consumption.The application of small cells is the fundamental mechanism for the 5Gtechnology.The use of small cells can enhance the facility for higher capacity and reuse.However,it must be noted that small cells deployment will lead to frequent handovers of mobile nodes.Considering the importance of small cells in 5G,this paper aims to examine a new resource management scheme that can work to minimize the rate of handovers formobile phones through careful resources allocation in a two-tier network.Therefore,the resource management problem has been formulated as an optimization issue thatwe aim to overcome through an optimal solution.To find a solution to the existing problem of frequent handovers,a heuristic approach has been used.This solution is then evaluated and validated through simulation and testing,during which the performance was noted to improve by 12%in the context of handover costs.Therefore,this model has been observed to be more efficient as compared to the existing model.

A Machine Learning Based Algorithm to Process Partial Shading Effects in PV Arrays

Solar energy is a widely used type of renewable energy.Photovoltaic arrays are used to harvest solar energy.The major goal,in harvesting the maximum possible power,is to operate the system at its maximum power point(MPP).If the irradiation conditions are uniform,the P-V curve of the PV array has only one peak that is called its MPP.But when the irradiation conditions are non-uniform,the P-V curve has multiple peaks.Each peak represents an MPP for a specific irradiation condition.The highest of all the peaks is called Global Maximum Power Point(GMPP).Under uniform irradiation conditions,there is zero or no partial shading.But the changing irradiance causes a shading effect which is called Partial Shading.Many conventional and soft computing techniques have been in use to harvest solar energy.These techniques perform well under uniform and weak shading conditions but fail when shading conditions are strong.In this paper,a new method is proposed which uses Machine Learning based algorithm called Opposition-Based-Learning(OBL)to deal with partial shading conditions.Simulation studies on different cases of partial shading have proven this technique effective in attaining MPP.

Enhancing Septic Shock Detection through Interpretable Machine Learning

This article presents an innovative approach that leverages interpretable machine learning models and cloud computing to accelerate the detection of septic shock by analyzing electronic health data.Unlike traditional methods,which often lack transparency in decision-making,our approach focuses on early detection,offering a proactive strategy to mitigate the risks of sepsis.By integrating advanced machine learning algorithms with interpretability techniques,our method not only provides accurate predictions but also offers clear insights into the factors influencing the model’s decisions.Moreover,we introduce a preference-based matching algorithm to evaluate disease severity,enabling timely interventions guided by the analysis outcomes.This innovative integration significantly enhances the effectiveness of our approach.We leverage a clinical health dataset comprising 1,552,210 Electronic Health Records(EHR)to train our interpretable machine learning models within a cloud computing framework.Through techniques like feature importance analysis and model-agnostic interpretability tools,we aim to clarify the crucial indicators contributing to septic shock prediction.This transparency not only assists healthcare professionals in comprehending the model’s predictions but also facilitates the integration of our system into existing clinical workflows.We validate the effectiveness of our interpretable models using the same dataset,achieving an impressive accuracy rate exceeding 98%through the application of oversampling techniques.The findings of this study hold significant implications for the advancement of more effective and transparent diagnostic tools in the critical domain of sepsis management.