Statistical Distribution Of Data Research Articles

Local fuzzy rough attribute reduction for large-scale mixed data with limited missing labels based on local fuzzy self information

The advent of the era of big data is accompanied by the generation of large-scale data of various types. Extracting the potential value and rules from such data has always been a challenge. Due to various external and internal factors, it is commonplace for large-scale data to exhibit the phenomenon of missing limited labels. In addressing a large-scale mixed information system with limited label missing (LSMDISLML), local neighborhood rough set model (LNRS-model) is typically employed. However, the identical neighborhood radius is often used by such model when confronted with numerical attributes, which could potentially attenuate the classification capability of the data. Local fuzzy rough set model (LFRS-model) can overcome this point. This paper studies local fuzzy rough attribute reduction for large-scale mixed data with limited missing labels based on LFRS-model via local fuzzy self information and overlap degree function. First, leveraging the statistical distribution of data as a foundation, fuzzy relations on the entire sample set are established, which has the advantage of being able to use different fuzzy similarity radii to calculate similarity, thereby adapting to different data distributions. Subsequently, the samples with missing labels are discarded as they constitute a small proportion of the entire sample set and have little impact on overall performance of dataset. The limited computing resources and storage space are focused on the sample set with complete labels (denoted as target set). Thereafter, based on the target set, local fuzzy λ-upper and lower approximations are defined, and LFRS-model is constructed. This model not only reduces processing time and sources of error in large-scale data but also improves data quality and enhances the reliability of the experimental results. Then, local fuzzy λ-self information is introduced and used to design a local fuzzy rough attribute reduction algorithm in a LSMDISLML. Furthermore, a overlap degree function is introduced to evaluate and reorder the attributes based on their importance, prioritizing the elimination of redundant attributes with high overlap and low importance from the preordered attribute set. This strategy effectively improves the efficiency of obtaining the optimal subset. Finally, a series of experiments are carried out. The experiment results demonstrate that the designed algorithm exhibits excellent performance in classification tasks and outlier detection tasks, surpassing existing four algorithms.

The Representative Points of Generalized Alpha Skew-t Distribution and Applications

Assuming the underlying statistical distribution of data is critical in information theory, as it impacts the accuracy and efficiency of communication and the definition of entropy. The real-world data are widely assumed to follow the normal distribution. To better comprehend the skewness of the data, many models more flexible than the normal distribution have been proposed, such as the generalized alpha skew-t (GAST) distribution. This paper studies some properties of the GAST distribution, including the calculation of the moments, and the relationship between the number of peaks and the GAST parameters with some proofs. For complex probability distributions, representative points (RPs) are useful due to the convenience of manipulation, computation and analysis. The relative entropy of two probability distributions could have been a good criterion for the purpose of generating RPs of a specific distribution but is not popularly used due to computational complexity. Hence, this paper only provides three ways to obtain RPs of the GAST distribution, Monte Carlo (MC), quasi-Monte Carlo (QMC), and mean square error (MSE). The three types of RPs are utilized in estimating moments and densities of the GAST distribution with known and unknown parameters. The MSE representative points perform the best among all case studies. For unknown parameter cases, a revised maximum likelihood estimation (MLE) method of parameter estimation is compared with the plain MLE method. It indicates that the revised MLE method is suitable for the GAST distribution having a unimodal or unobvious bimodal pattern. This paper includes two real-data applications in which the GAST model appears adaptable to various types of data.

The Forecasting of the Spread of Infectious Diseases Based on Conditional Generative Adversarial Networks

New epidemics encourage the development of new mathematical models of the spread and forecasting of infectious diseases. Statistical epidemiology data are characterized by incomplete and inexact time series, which leads to an unstable and non-unique forecasting of infectious diseases. In this paper, a model of a conditional generative adversarial neural network (CGAN) for modeling and forecasting COVID-19 in St. Petersburg is constructed. It takes 20 processed historical statistics as a condition and is based on the solution of the minimax problem. The CGAN builds a short-term forecast of the number of newly diagnosed COVID-19 cases in the region for 5 days ahead. The CGAN approach allows modeling the distribution of statistical data, which allows obtaining the required amount of training data from the resulting distribution. When comparing the forecasting results with the classical differential SEIR-HCD model and a recurrent neural network with the same input parameters, it was shown that the forecast errors of all three models are in the same range. It is shown that the prediction error of the bagging model based on three models is lower than the results of each model separately.

Bayesian model of tilling wheat confronting climatic and sustainability challenges.

Conventional farming poses threats to sustainable agriculture in growing food demands and increasing flooding risks. This research introduces a Bayesian Belief Network (BBN) to address these concerns. The model explores tillage adaptation for flood management in soils with varying organic carbon (OC) contents for winter wheat production. Three real soils, emphasizing texture and soil water properties, were sourced from the NETMAP soilscape of the Pang catchment area in Berkshire, United Kingdom. Modified with OC content at four levels (1, 3, 5, 7%), they were modeled alongside relevant variables in a BBN. The Decision Support System for Agrotechnology Transfer (DSSAT) simulated datasets across 48 cropping seasons to parameterize the BBN. The study compared tillage effects on wheat yield, surface runoff, and GHG-CO2 emissions, categorizing model parameters (from lower to higher bands) based on statistical data distribution. Results revealed that NT outperformed CT in the highest parametric category, comparing probabilistic estimates with reduced GHG-CO2 emissions from "7.34 to 7.31%" and cumulative runoff from "8.52 to 8.50%," while yield increased from "7.46 to 7.56%." Conversely, CT exhibited increased emissions from "7.34 to 7.36%" and cumulative runoff from "8.52 to 8.55%," along with reduced yield from "7.46 to 7.35%." The BBN model effectively captured uncertainties, offering posterior probability distributions reflecting conditional relationships across variables and offered decision choice for NT favoring soil carbon stocks in winter wheat (highest among soils "NT.OC-7%PDPG8," e.g., 286,634 kg/ha) over CT (lowest in "CT.OC-3.9%PDPG8," e.g., 5,894 kg/ha). On average, NT released minimum GHG- CO2 emissions to "3,985 kgCO2eqv/ha," while CT emitted "7,415 kgCO2eqv/ha." Conversely, NT emitted "8,747 kgCO2eqv/ha" for maximum emissions, while CT emitted "15,356 kgCO2eqv/ha." NT resulted in lower surface runoff against CT in all soils and limits runoff generations naturally for flood alleviation with the potential for customized improvement. The study recommends the model for extensive assessments of various spatiotemporal conditions. The research findings align with sustainable development goals, e.g., SDG12 and SDG13 for responsible production and climate actions, respectively, as defined by the Agriculture and Food Organization of the United Nations.

Outcome of Patients with Early Glottic Laryngeal Carcinoma Treated with Radical Radiotherapy

Introduction: Glottic cancer is the most common laryngeal cancer and it has the most favorable prognosis, mainly if diagnosed early. Ideally, early glottic cancer (T1–T2 N0) should be managed with a single modality either microsurgery or definitive Radiotherapy. The local control rate with definitive radiotherapy is approximately 83–95% for T1 and 70–80% for T2 lesions. This study intends to find out the treatment outcome in early glottic cancer treated with radical radiation treatment from 1st January 2014 to 31st December 2019 at our center. Methods: This was a retrospective analysis of medical records and radiation charts of all patients of early stage glottis cancer who underwent radiation treatment from 1st January 2014 to 31st December 2019 at our centre. The statistical analysis was done using SPSS version 20.0.Continuous variables such as age was summarized in terms of either mean ± Standard Deviation or median (inter quartile range) depending on the statistical distribution of data. Categorical variables were e pressed in terms of frequency and percentage. Survival analysis was done using Kaplan Meier curves. Results: A total of 91 patients underwent radical radiation for early glottic carcinoma at our centre. The median age was 63 years (42-85 years). Out of them 94% were males and 6% were females. 78% of patients had stage 1 disease and 22% had stage 11 disease. 52% of patients had co morbidities and 48% did not have any comorbidity All cases had Squamous cell carcinoma histology. 59% of patients had anterior commissure involvement. 12 out of 91 (13%) patients developed local recurrence on follow up. The five year disease free survival was 84.3%. The five year overall survival was 87.7%. There was a statistically significant difference in survival between stage I and stage II. Conclusion: For T1–T2 glottic carcinomas, definitive RT can provide a good term local control and survival with the preservation of the normal structure and function of larynx.

Feature selection for label distribution learning based on the statistical distribution of data and fuzzy mutual information

Label distribution learning (LDL) is an emerging framework in machine learning. Fuzzy mutual information is mutual information under a fuzzy environment and plays an important role in handling uncertainty. This paper explores feature selection for LDL data based on the statistical distribution of data and fuzzy mutual information. The similarity between the feature values in the feature space is first defined by means of the statistical distribution of the data, and a threshold is introduced to control the similarity. Then, the fuzzy similarity relation for each feature subset is established via the similarity. This method utilizes adjustable fuzzy similarity radii to establish a fuzzy similarity relation and improve the classification ability of the data. The decision relation in the label space is then presented, and the decision class of each sample is constructed. Subsequently, two feature selection algorithms based on fuzzy mutual information are designed to remove the irrelevant features by employing a strategy that considers the correlation between the features and labels as well as the redundancy between the features in the LDL data. Finally, the experimental results show that the designed algorithms can effectively measure the uncertainty of LDL data and outperform four state-of-the-art feature selection algorithms. Specifically, our algorithms, LDFM and LDFMR, demonstrate their superiority by achieving overall average ranking improvements of 63.64% and 58.52%, respectively, across six evaluation metrics compared to the other four algorithms.

FedSBS: Federated-Learning participant-selection method for Intrusion Detection Systems

Federated Learning (FL) is a decentralized machine learning approach in which multiple participants collaboratively train a model. Participants keep data locally, train their local models, and aggregate them in a single global model in a federated server. Collaborative FL-based Intrusion Detection Systems face challenges on an uneven statistical distribution of data and malicious participants trying to subvert the learning process. The statistical hurdles associated with imbalanced data and malicious participants pose a risk of skewing the training with biased or random data. The inability to effectively manage these statistical inconsistencies may degrade system performance, leading to false intrusion detection or opening avenues for cybersecurity breaches. To overcome these challenges, we propose a training method that employs score-based participant selection and utilizes global momentum for model aggregation. Our method improves the global model performance while mitigating the risks posed by malicious participants. The proposal incorporates a scoring system based on an information gain variant to evaluate each participant’s contribution. The scoring system and an epsilon greedy selection method ensure robust participant selection in each aggregation round. Furthermore, incorporating a global momentum term helps preserve previous knowledge at each aggregation round, contributing to model stability and overall learning. The proposed solution has demonstrated superior performance, delivering 80% F1-Score and 90% accuracy on experiments even in the presence of malicious participants, revealing the robustness and effectiveness of the proposal in mitigating statistical challenges. Consequently, the proposed method significantly enhances the performance of federated learning models, leading to more secure and efficient collaborative intrusion detection systems.

An adaptive spatio-temporal neural network for PM2.5 concentration forecasting

Accurate PM2.5 concentration prediction is essential for environmental control management, therefore numerous air quality monitoring stations have been established, which generate multiple time series with spatio-temporal correlation. However, the statistical distribution of data from different monitoring stations varies widely, which needs to provide higher flexibility in the feature extraction stage. Moreover, the spatio-temporal correlation and mutation characteristics of the time series are difficult to capture. To this end, an adaptive spatio-temporal prediction network (ASTP-NET) is proposed, in which the encoder adaptively extracts the input data features, then captures the spatio-temporal dependencies and dynamic changes of the time series, the decoder part maps the encoded features into a predicted future time series representation, while an objective function is proposed to measure the overall fluctuations of the model’s multi-step prediction. In this paper, ASTP-NET is evaluated based on the Xi'an air quality dataset, and the results show that the model outperforms other baseline methods.

RACBase: A cloud-based database of recycled aggregate concrete durability

This paper presents the first universal cloud-based database of recycled aggregate concrete (RAC) durability, named RACBase. The database collects experimental data on the durability of RAC from published literature all over the globe. Each set of data consists of the literature source, materials information, and the unified test result. The resistance to chloride ion attack, carbonation, sulfate attack, and freeze-thaw for RAC is quantified by the passed electric charge, carbonation depth, compressive strength loss, and durability factor, respectively. Based on the web interface, RACBase provides the functions of searching, adding, acquiring, and storing data, and gives the statistical distribution of data for each durability test. Users can freely access RACBase at https://www.racbase.com and browse and upload their data after online registration. One illustration of building a machine learning model for predicting the chloride resistance of RAC is presented to show the potential application of the database. The extreme gradient boosting model shows superiority to the artificial neural network and random forest models with the R2 of 0.975, RMSE of 246.719, and MAPE of 0.081. RACBase enriches the way for users to conduct in-depth research on the durability of RAC and construct performance prediction models.

Improving the robustness of adversarial attacks using an affine-invariant gradient estimator

As designers of artificial intelligence try to outwit hackers, both sides continue to hone in on AI’s inherent vulnerabilities. Designed and trained from certain statistical distributions of data, deep neural networks (DNNs) remain vulnerable to deceptive inputs that violate a DNN’s statistical, predictive assumptions. Before being fed into a neural network, however, most existing adversarial examples cannot maintain malicious functionality when applied to an affine transformation. For practical purposes, maintaining that malicious functionality serves as an important measure of the robustness of adversarial attacks. To help DNNs learn to defend themselves more thoroughly against attacks, we propose an affine-invariant adversarial attack, which can consistently produce more robust adversarial examples over affine transformations. For efficiency, we propose to disentangle current affine-transformation strategies from the Euclidean geometry coordinate plane with its geometric translations, rotations and dilations; we reformulate the latter two in polar coordinates. Afterwards, we construct an affine-invariant gradient estimator by convolving the gradient at the original image with derived kernels, which can be integrated with any gradient-based attack methods. Extensive experiments on ImageNet, including some experiments under physical condition, demonstrate that our method can significantly improve the affine invariance of adversarial examples and, as a byproduct, improve the transferability of adversarial examples, compared with alternative state-of-the-art methods.

Investigating the relationship between talent management and knowledge sharing among physical education teachers in Fardis city

The purpose of this research was to investigate the relationship between talent management and knowledge sharing among physical education teachers in Fardis city. Descriptive-correlation research method that has been done by field. The statistical population of physical education teachers in Fardis city was determined to be 130 people and the statistical sample was determined as all 130 people and then the statistical sample was selected by simple random and after collecting the questionnaires it was observed that 117 questionnaires could be used. Measurement tools include Talent Management Questionnaire (Sarmad et al., 2011) and Knowledge Sharing Questionnaire (Dixon, 2001). Content and face validity of the questionnaires were confirmed by management professors and then their reliability was 0.83 for the talent management questionnaire and 0.74 for the knowledge sharing questionnaire, respectively. Descriptive and inferential statistics (Pearson correlation coefficient) and Kolmograph-Smirnov test were used for analysis to determine the normality of statistical data distribution. Findings showed that there is a positive and significant relationship between talent management and knowledge sharing. Therefore, it can be concluded that if the staff talent management organization is properly managed and supervised, we will see the sharing of knowledge among employees as well as the responsibility of Physical education teachers.

Influence of Maintenance and Repair of Axle Boxes on the Risks of Their Failure

Purpose. The main purpose of the study is to identify the interconnection between the risks of failure of axle boxes of freight cars and the success of performing a complex task of maintenance and repair. To do this, one should identify the sequence of calculating the values of successful completion of the complex task of maintenance and repair of freight cars, make practical calculations using the example of car repair companies of Prydniprovska Railway regional branch and identify the dependencies of failure risks of the freight cars and the process of successful completion of the task. Methodology. The paper constructs the interconnections of the probabilities of performing and not performing a complex task of maintenance and repair of freight cars. In addition, the calculation sequence of static probability estimates of successful performing and not performing of the set task, static estimates of the average deviation time and the main time of successful performing, as well as its intensity is presented. Findings. The obtained statistical data of the effective implementation of the complex task of maintenance and repair of freight cars at Prydniprovska Railway regional branch and the results of their empirical distribution throughout a period of 90 days, showed that the sample average is 24.1078 cars, the sample variance is 3.28 cars, the value of the standard deviation is 1.81 cars, and the coefficient of variation is 0.075 cars. The influence of diagnostics on the failure risks of axle boxes of freight cars during operation after fulfilling the task of maintenance and repair is also established. Originality. Based on statistical processing of experimental data, for the first time an empirical distribution of statistical data of successful completion of a complex task of maintenance and repair of freight cars was performed, which allows calculating the reliability indicators of axle boxes after maintenance and repair of freight cars. For the first time, the dependence of axle boxes diagnostics on the risk of failure of performing the task of maintenance and repair of freight cars is received, which will allow reducing the risk of failure if the task is successfully completed. Practical value. The use of diagnostics of axle boxes of freight cars during maintenance and repair can reduce the risk of failure by 2… 4.5 times.

Genomic selection for resistance to mammalian bark stripping and associated chemical compounds in radiata pine.

The integration of genomic data into genetic evaluations can facilitate the rapid selection of superior genotypes and accelerate the breeding cycle in trees. In this study, 390 trees from 74 control-pollinated families were genotyped using a 36K Axiom SNP array. A total of 15,624 high-quality SNPs were used to develop genomic prediction models for mammalian bark stripping, tree height, and selected primary and secondary chemical compounds in the bark. Genetic parameters from different genomic prediction methods-single-trait best linear unbiased prediction based on a marker-based relationship matrix (genomic best linear unbiased prediction), multitrait single-step genomic best linear unbiased prediction, which integrated the marker-based and pedigree-based relationship matrices (single-step genomic best linear unbiased prediction) and the single-trait generalized ridge regression-were compared to equivalent single- or multitrait pedigree-based approaches (ABLUP). The influence of the statistical distribution of data on the genetic parameters was assessed. Results indicated that the heritability estimates were increased nearly 2-fold with genomic models compared to the equivalent pedigree-based models. Predictive accuracy of the single-step genomic best linear unbiased prediction was higher than the ABLUP for most traits. Allowing for heterogeneity in marker effects through the use of generalized ridge regression did not markedly improve predictive ability over genomic best linear unbiased prediction, arguing that most of the chemical traits are modulated by many genes with small effects. Overall, the traits with low pedigree-based heritability benefited more from genomic models compared to the traits with high pedigree-based heritability. There was no evidence that data skewness or the presence of outliers affected the genomic or pedigree-based genetic estimates.

Data-Driven Analysis of Cycle-to-Cycle Variations, Scatter, and Furcation in the UH-60A Wind Tunnel Rotor Airloads Measurements

A data-driven clustering algorithm based on proper orthogonal decomposition was applied to assess the scatter found in the UH-60A wind tunnel airloads measurements. Upon verifying the capability of the algorithm, pushrod loads, blade surface pressures, sectional loads, and torsional moments were analyzed. Spatial eigenmodes resulting from the decomposition provided the optimal basis; projection of the individual cycles onto the high singular value modes allowed visualizing the statistical distribution of data over the entire azimuth. While not all cases showed furcation in the data, the bimodal distribution was found in the high-thrust cases, where statistically normal distribution is generally assumed. Consequent clustering of the measured cycles produced an excellent correlation among clusters found in the pushrod loads, blade surface pressure, and torsional moment that suggests a common source for furcation in the data. The cycles assigned to one group repeatedly showed distinguishable variations from the other group in terms of the presence/absence of a dynamic stall vortex, azimuthal occurrence of the stall, chordwise location of separation, reattachment, and so on. When one of the clusters is smaller in size compared to the other, the conventional phase average obscured all the intricate features even when the loads are substantially higher than the larger cluster. In general, clustering the dataset when warranted showed not only higher peak loads but also lower variance for both the clusters across the entire azimuth compared to the conventional simple phase-average results. Computational simulations were conducted using CREATETM -AV Helios towards understanding the underlying flow field. Misinterpreted earlier as under-/overpredictive when compared with the simple phase-average data, Helios results consistently showed significantly improved correlation with one of the two clusters. Combining the clustered results and the flow visualization provided by Helios, aperiodicity in the spatial location and the strength of both the trim tab vortices and tip vortices have been hypothesized as potential sources of furcation.

Evaluation of the Probabilistic Distribution of Statistical Data Used in the Process of Developing Fragility Curves

Evaluation of the Probabilistic Distribution of Statistical Data Used in the Process of Developing Fragility Curves

Validation of artificial intelligence prediction models for skin cancer diagnosis using dermoscopy images: the 2019 International Skin Imaging Collaboration Grand Challenge

BackgroundPrevious studies of artificial intelligence (AI) applied to dermatology have shown AI to have higher diagnostic classification accuracy than expert dermatologists; however, these studies did not adequately assess clinically realistic scenarios, such as how AI systems behave when presented with images of disease categories that are not included in the training dataset or images drawn from statistical distributions with significant shifts from training distributions. We aimed to simulate these real-world scenarios and evaluate the effects of image source institution, diagnoses outside of the training set, and other image artifacts on classification accuracy, with the goal of informing clinicians and regulatory agencies about safety and real-world accuracy.MethodsWe designed a large dermoscopic image classification challenge to quantify the performance of machine learning algorithms for the task of skin cancer classification from dermoscopic images, and how this performance is affected by shifts in statistical distributions of data, disease categories not represented in training datasets, and imaging or lesion artifacts. Factors that might be beneficial to performance, such as clinical metadata and external training data collected by challenge participants, were also evaluated. 25 331 training images collected from two datasets (in Vienna [HAM10000] and Barcelona [BCN20000]) between Jan 1, 2000, and Dec 31, 2018, across eight skin diseases, were provided to challenge participants to design appropriate algorithms. The trained algorithms were then tested for balanced accuracy against the HAM10000 and BCN20000 test datasets and data from countries not included in the training dataset (Turkey, New Zealand, Sweden, and Argentina). Test datasets contained images of all diagnostic categories available in training plus other diagnoses not included in training data (not trained category). We compared the performance of the algorithms against that of 18 dermatologists in a simulated setting that reflected intended clinical use.Findings64 teams submitted 129 state-of-the-art algorithm predictions on a test set of 8238 images. The best performing algorithm achieved 58·8% balanced accuracy on the BCN20000 data, which was designed to better reflect realistic clinical scenarios, compared with 82·0% balanced accuracy on HAM10000, which was used in a previously published benchmark. Shifted statistical distributions and disease categories not included in training data contributed to decreases in accuracy. Image artifacts, including hair, pen markings, ulceration, and imaging source institution, decreased accuracy in a complex manner that varied based on the underlying diagnosis. When comparing algorithms to expert dermatologists (2460 ratings on 1269 images), algorithms performed better than experts in most categories, except for actinic keratoses (similar accuracy on average) and images from categories not included in training data (26% correct for experts vs 6% correct for algorithms, p<0·0001). For the top 25 submitted algorithms, 47·1% of the images from categories not included in training data were misclassified as malignant diagnoses, which would lead to a substantial number of unnecessary biopsies if current state-of-the-art AI technologies were clinically deployed.InterpretationWe have identified specific deficiencies and safety issues in AI diagnostic systems for skin cancer that should be addressed in future diagnostic evaluation protocols to improve safety and reliability in clinical practice.FundingMelanoma Research Alliance and La Marató de TV3.

PENGARUH METODE LATIHAN FARTLEK TERHADAP PENINGKATAN DAYA TAHAN AEROBIC PESERTA EKTRAKURIKULER FUTSAL PUTRA

AbstrakThe problem in this study is the ability of aerobic endurance for futsal players at SMA Muhammadiyah 1 Pontianak, who take part in futsal extracurriculars are still categorized as lacking and the Fartlek method has not been implemented. The purpose of this study was to find the effect of the Fartlek method on aerobic endurance ability. The population in this study were students who took extracurricular activities at SMA Muhammadiyah 1 Pontianak, amounting to 24 people. Sampling in this study using saturated sampling technique, namely the technique of determining the sample if all the population is used as a sample. The form of research used in this study is a Pre Experimental Design with a One-Group Pretest-Posttest Design model, namely a research design that is given a pretest before being given treatment, so that the results of the treatment can be known more accurately. Based on the calculated chi square for pretest = 10,806 and posttest = 8.834, it is smaller than the chi square table (11,070), then the distribution of statistical data for the 24 athletes can be declared normal, F count is smaller than F table 1.179 &lt;2.01 for F table 0 0.05, it can be concluded that the variance of the data to be analyzed is homogeneous. The calculation of the T test obtained the value of t count = 8,306 &gt; t table = 1,714. The percentage increase in aerobic endurance using the fartlek method is 6.75%. Thus, the conclusion of this study is that increasing aerobic endurance by using the fartlek method has been shown to have an effect on increasing the aerobic endurance of students who take part in futsal extracurricular at SMA Muhammadiyah 1 Pontianak.Keywords: Fartlek, endurance

Learning From Network Data Changes for Unsupervised Botnet Detection

The networks of infected devices (a.k.a., botnets) threaten network security due to their dynamic nature and support to different attacks (e.g., Distributed Denial of Services and personal data theft). Detecting botnets is a challenging task because the infected devices (bots) are numerous, widely and geographically spread. Significant attention has been given to improve the efficiency, robustness and adaptability of network security approaches. However, in the literature, botnet detection techniques usually ignore fast changes in statistical data distribution, performing over static windows, i.e., fixed intervals of time or fixed quantity of flows. Changes in statistical data distribution are known as concept drifts and they make the classification models obsolete. Furthermore, those works employing approaches aware of concept drift use supervised machine learning, which is slow, costly, and prone to error. Therefore, this article presents TRUSTED, a system for online and unsupervised botnet detection aware of concept drifts. Unlike other works, the TRUSTED system improves the learning process for botnet detection, applying concept drift in an online and unsupervised classification. Evaluations comprise offline and online scenarios. Results show that the TRUSTED system detects botnets using concept drift identification, reaching 87% to 95% accuracy, precision, recall, and F1-scores.

Improving the Robustness of Adversarial Attacks Using an Affine-Invariant Gradient Estimator

As designers of artificial intelligence try to outwit hackers, both sides continue to hone in on AI's inherent vulnerabilities. Designed and trained from certain statistical distributions of data, AI's deep neural networks (DNNs) remain vulnerable to deceptive inputs that violate a DNN's statistical, predictive assumptions. Before being fed into a neural network, however, most existing adversarial examples cannot maintain malicious functionality when applied to an affine transformation. For practical purposes, maintaining that malicious functionality serves as an important measure of the robustness of adversarial attacks. To help DNNs learn to defend themselves more thoroughly against attacks, we propose an affine-invariant adversarial attack, which can consistently produce more robust adversarial examples over affine transformations. For efficiency, we propose to disentangle current affine-transformation strategies from the Euclidean geometry coordinate plane with its geometric translations, rotations and dilations; we reformulate the latter two in polar coordinates. Afterwards, we construct an affine-invariant gradient estimator by convolving the gradient at the original image with derived kernels, which can be integrated with any gradient-based attack methods. Extensive experiments on ImageNet, including some experiments under physical condition, demonstrate that our method can significantly improve the affine invariance of adversarial examples and, as a byproduct, improve the transferability of adversarial examples, compared with alternative state-of-the-art methods.

Exploring the Impact of Government Interventions on COVID-19 Pandemic Spread in Kuwait

To model the trajectory of the pandemic in Kuwait from February 24, 2020 to February 28, 2021, we used two modeling procedures: Auto Regressive Integrated Moving Average (ARIMA) with structural breaks and Multivariate Adaptive Regression Splines (MARS), and then mapped the key breakpoints of the models to the set of government-enforced interventions. The MARS model, as opposed to the ARIMA model, provides a more precise interpretation of the intervention's effects. It demonstrates that partial and total lockdown interventions were highly effective in reducing the number of confirmed cases. When some interventions, such as enforcing regional curfews, closing workplaces, and imposing travel restrictions, were combined, their impact became significant. MARS method is recommended to be applied when exploring the impact of interventions on the spread of a disease. It does not require any prior assumptions about the statistical distribution of data, does not affect data collinearity, has simple and transparent functions, and allows for a more accurate analysis of intervention results.