Email-Based Privilege Escalation Attack Detection in Cloud Environments Using Machine Learning

Cloud computing offers scalable data access and operational efficiency but introduces serious security challenges—most notably, privilege escalation attacks where insiders gain unauthorized administrative control. This project presents a machine learning-based system to detect such attacks through email analysis. Email content is transformed into numerical features using vectorization techniques. Several classifiers including Random Forest, KNN, SVM, XGBoost, and LightGBM are trained on a labeled dataset to differentiate between safe and malicious emails. An ensemble voting classifier ensures higher accuracy and reduces false positives. The system performs real-time detection and generates alerts for suspicious activities, thereby enhancing cloud security against internal threats.