Anomaly Prediction in Cybersecurity: A Machine Learning Model from the Perspective of Data Engineering and Fingerprinting

This project utilizes artificial intelligence (AI) and machine learning through the development of a mathematical-predictive model to reliably detect cyber anomalies. Using the BETH dataset and the CRISP-DM methodology, this research has addressed the problem of combining kernel and network traffic data, achieving a 37.27% increase in the detection of malicious activities compared to the initial data. Additionally, an innovative dataset was formulated, in which the complete trace of a new botnet attack pattern was discovered, previously unknown to BETH, involving the entire monitored network in illicit cryptocurrency mining. Finally, several models were successfully built and trained using Random Forest and Decision Trees algorithms, with accuracies of 100% and 99%, respectively.