Deep convolutional forest: a dynamic deep ensemble approach for spam detection in text.

Area of Science:

• Computer Science
• Information Security

Background:

• Mobile messaging services facilitate social engineering attacks like phishing, leading to sensitive data theft.
• Misinformation and rumors, especially during health crises like COVID-19, spread rapidly on social media, causing public fear.
• Current spam classification methods, including machine learning and deep learning, face limitations such as manual feature engineering and high computational costs.

Purpose of the Study:

• To develop an advanced spam detection model that overcomes the limitations of existing approaches.
• To introduce a dynamic deep ensemble model capable of automatic feature extraction and adaptive complexity for enhanced spam classification.

Main Methods:

• The proposed model integrates convolutional and pooling layers for automatic feature extraction.
• It utilizes base classifiers like random forests and extremely randomized trees.
• Ensemble learning techniques, including boosting and bagging, are employed to enhance classification performance.

Main Results:

• The dynamic deep ensemble model achieved a high precision, recall, f1-score, and accuracy of 98.38%.
• The model demonstrates superior performance in classifying texts as spam or legitimate.

Conclusions:

• The developed dynamic deep ensemble model offers an efficient and accurate solution for spam detection.
• This approach effectively mitigates risks associated with phishing and misinformation spread through digital communication channels.