AN ENHANCED SYSTEM FOR DETECTION OF DENIAL OF SERVICE ATTACKS IN DISTRIBUTED SYSTEMS USING DEEP LEARNING

Abstract

The proliferation of distributed systems has fundamentally transformed how organizations manage their computational infrastructure, yet this advancement has simultaneously exposed critical vulnerabilities to Denial of Service (DoS) attacks. Traditional detection mechanisms struggle to identify sophisticated attack patterns in real-time, particularly within cloud-based and edge computing environments. This research introduces an enhanced detection framework leveraging deep learning architectures, specifically combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to analyze network traffic patterns. Through experimental validation on a dataset comprising 2.3 million network packets collected from enterprise distributed systems, our proposed model achieved a detection accuracy of 98.7%, significantly outperforming conventional machine learning approaches. The system demonstrates remarkable capability in identifying zero-day attack variants while maintaining minimal false positive rates below 1.2%. Implementation across three distinct cloud environments revealed average detection latency of 47 milliseconds, making it viable for real-time deployment. This research contributes to cybersecurity literature by establishing a scalable, adaptive framework that addresses the evolving threat landscape facing distributed computing infrastructure, offering practical implications for system administrators and security professionals managing large-scale networked environments.