DYNAMIC RANSOMWARE DETECTION USING TIME-BASED API CALLING ANALYSIS

Abstract

Numerous malware varieties, such as ransomware, adware, keyloggers, rootkits, and botnets, have emerged as a result of the quick evolution of cyberthreats and present serious risks to digital systems. The machine learning-based multiclass malware detection system presented in this research is intended to correctly categorise various kinds of harmful software. Before being fed into many machine learning models, such as Random Forest, XGBoost, and Long ShortTerm Memory (LSTM), the system uses a feature-based dataset that has been preprocessed and normalised. To assess these models' performance in terms of classification accuracy and dependability, a comparison analysis is carried out. Because XGBoost can efficiently capture complicated feature interactions and minimise overfitting, it was the model with the best accuracy. A Flask-based web application is used to implement the suggested system, allowing users to enter feature values and obtain real-time forecasts for each form of malware along with thorough explanations and countermeasures. The findings show that ensemble learning techniques, especially XGBoost, are very successful in classifying multi-class malware and can be used in practical cybersecurity systems