Résumé:
Ransomware is malicious software that encrypts victims' data and demands a ransom
to decrypt them. This type of malware attacks are becoming more sophisticated,
posing a significant threat to individuals and organizations. This research focuses on
developing a powerful ransomware detection model that integrates behavioral
analysis, deep learning, and bootstrapping techniques. The model uses behavioral
analysis to identify ransomware samples, while deep learning techniques train
multiple specialized models to detect zero-day ransomware attacks and minimize
false positives. The proposed model outperforms machine learning algorithms in
terms of accuracy, precision, and recall. This work should serve as the first step for
further research and exploration of additional features, behavioral indicators, static
analysis techniques, and hybrid approaches to enhance detection capabilities and
combat ransomware threats, and finally to deployment in production.
Keywords: Ransomware Detection, Deep Learning, Feedforward Neural Network, Machine Learning, Ensemble Learning.
