• Version
• Download 10
• File Size 451.29 KB
• File Count 1
• Create Date 20 March 2026
• Last Updated 20 March 2026

Identification and Prevention of Cyber Attacks with Authentications

1.R. Satya Teja, 2.S. Sri Ram Bharadwaj, 3.Bharat, 4.G. Veerendra Nath, 5.G. Hariharan

Dept. of Computer Science and Engineering

Jyothishmathi Institute of Technology and Science, Karimnagar, Telangana, India

satyateja@jits.ac.in, srirambharadwajsammeta@gmail.com, bharat@gmail.com, veerendranath@gmail.com, hariharan@gmail.com

Abstract—As web application attacks evolve in sophistication, reliance on static firewalls or standalone Machine Learning (ML) classifiers is no longer sufficient. This paper presents the Adaptive Hybrid Intrusion Prevention System (AH-IPS), a novel architecture combining Stacked Ensemble Learning with a deterministic rule-based engine. We utilized a high-fidelity synthetic dataset of 10,000 events to train and evaluate three classifiers: Random Forest, XGBoost, and Logistic Regression. Results demonstrate that XGBoost yields the highest standalone performance with an F1-Score of 0.8802. Crucially, we introduce a Hybrid Decision Engine that mitigates the inherent false negatives of the ML model by enforcing heuristic rules for high- risk vectors like SQL Injection and XSS. The proposed system effectively balances probabilistic detection with deterministic prevention, offering a robust defense strategy for modern web applications.

Index Terms—Intrusion Prevention, Machine Learning, Ensemble Learning, SQL Injection, XSS, Cybersecurity.