PhishStack: Evaluation of Stacked Generalization in Phishing URLs Detection

Rahman, Sheikh Shah Mohammad Motiur; Islam, Takia; Jabiullah, Md. Ismail

doi:10.1016/j.procs.2020.03.294

Cited by 17 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the stacking method [ 27 ], the primary learner is trained first, and then the prediction result of the primary learner is used as the new input to train the secondary learner. In the training phase, the secondary learner is generated by the primary learner.…”

Section: Ensemble Modelmentioning

confidence: 99%

Application of Bagging, Boosting and Stacking Ensemble and EasyEnsemble Methods for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area of China

Wang

2023

IJERPH

View full text Add to dashboard Cite

Since the impoundment of the Three Gorges Reservoir area in 2003, the potential risks of geological disasters in the reservoir area have increased significantly, among which the hidden dangers of landslides are particularly prominent. To reduce casualties and damage, efficient and precise landslide susceptibility evaluation methods are important. Multiple ensemble models have been used to evaluate the susceptibility of the upper part of Badong County to landslides. In this study, EasyEnsemble technology was used to solve the imbalance between landslide and nonlandslide sample data. The extracted evaluation factors were input into three bagging, boosting, and stacking ensemble models for training, and landslide susceptibility mapping (LSM) was drawn. According to the importance analysis, the important factors affecting the occurrence of landslides are altitude, terrain surface texture (TST), distance to residences, distance to rivers and land use. The influences of different grid sizes on the susceptibility results were compared, and a larger grid was found to lead to the overfitting of the prediction results. Therefore, a 30 m grid was selected as the evaluation unit. The accuracy, area under the curve (AUC), recall rate, test set precision, and kappa coefficient of a multi-grained cascade forest (gcForest) model with the stacking method were 0.958, 0.991, 0.965, 0.946, and 0.91, respectively, which a significantly better than the values produced by the other models.

show abstract

Section: Ensemble Modelmentioning

confidence: 99%

Application of Bagging, Boosting and Stacking Ensemble and EasyEnsemble Methods for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area of China

Wang

2023

IJERPH

View full text Add to dashboard Cite

show abstract

“…They summarize the data concisely, and the placement of the box and whisker markers makes it simple to compare the classification accuracy. More information on the performance indicators can be found in [27], [30].…”

Section: Performance Evaluationmentioning

confidence: 99%

Machine Learning Aided Classification of Noise Distribution in Scanning Electron Microscopy Images

Rahman,

Salomon,

Dembélé

2023

2023 3rd International Conference on Computer, Control and Robotics (ICCCR)

View full text Add to dashboard Cite

Noise estimation is a crucial part of any modern supervised denoiser. Various statistical approaches are studied to estimate the noise, but generally these depend on a manual analysis of the images. To remove the manual dependency, it is important to automate the noise estimation process. In this paper, the initial phase of noise estimation, which is to identify the types of noise distribution, is performed using machine learning (ML) techniques. To make the images workable with ML techniques, a feature extraction process was performed. Hu's moment invariants, Haralick's texture, and color histogram are extracted from the images and stacked horizontally by scaling with MinMax scale the three features into one. Label encoder is used for normalizing the labels. Multiple ML techniques are trained and validated, and then tested with unknown images. The result is that stacking multiple ML techniques can produce better results with an accuracy above 90%. Stacking with the test set produces the following scores for Precision, Recall, and F1-score: 0.98, 0.88, 0.93, and 0.89, 0.98, 0.93 for Gaussian and Poisson respectively, with an average precision of 88%. These promising results prove the capability of ML techniques for image noise classification tasks where noise is artificially added. However, in real case, i.e. when the images come from a Zeiss Auriga FE SEM, which is the initial target, the classification is not as efficient. Thus, it is not always possible to work in a real denoising scenario if the model is trained with synthetic data.

show abstract

“…This illicit activity exploits social networking technologies and platforms to amass a target's identity data and account information. In the effective counteraction of spam emails, diverse techniques and principles have been devised, where email filtering assumes a pivotal position [3] [4]. Email filtering involves an array of approaches, incorporating mechanisms like blacklisting, visual similarity assessments, heuristics, and machine learning algorithms.…”

Section: Introductionmentioning

confidence: 99%

Spam Email Detection Using Deep Learning Model

Narayan,

Kumar

2024

International Journal of Innovative Research in Computer Science and Technology (IJIRCST)

View full text Add to dashboard Cite

The persistent menace of spam emails poses a formidable cybersecurity threat, demanding the implementation of robust classification models for prompt detection and prevention. This research addresses the intricate challenge of classifying spam emails by evaluating diverse machine learning and deep learning models. The focal point of the identified problem lies in the inherent complexity of distinguishing spam emails from legitimate ones, leading to potential security vulnerabilities. Existing models' limitations, particularly in managing false positives and false negatives, underscore the imperative for enhanced techniques. The chosen methodology incorporates conventional machine learning models, such as Random Forest, Multinomial Naive Bayes (MNB), and Support Vector Machine (SVM), in conjunction with the Netcraft dataset. Additionally, deep learning models, including Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Bidirectional LSTM (BiLSTM), are deployed with meticulous hyperparameter tuning. An attention mechanism is introduced to dynamically allocate distinct weights to email headers and bodies based on their significance. The outcomes unveil the consistently superior performance of the Random Forest model, attaining a flawless F1-score of 1.0 for spam email classification. The SVM model also exhibits excellence with impeccable precision and recall across all folds. Nevertheless, the MNB model encounters difficulties in classifying legitimate emails, influencing overall metrics. The deep learning models, especially LSTM, CNN, BiL STM, and RNN, demonstrate exceptional efficacy with remarkably minimal test losses and elevated accuracies, underscoring their prowess in spam email classification tasks.

show abstract

PhishStack: Evaluation of Stacked Generalization in Phishing URLs Detection

Cited by 17 publications

References 16 publications

Application of Bagging, Boosting and Stacking Ensemble and EasyEnsemble Methods for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area of China

Application of Bagging, Boosting and Stacking Ensemble and EasyEnsemble Methods for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area of China

Machine Learning Aided Classification of Noise Distribution in Scanning Electron Microscopy Images

Spam Email Detection Using Deep Learning Model

Contact Info

Product

Resources

About