A CYBER-RESILIENT COMPUTING MODEL FOR INTRUSION DETECTION, RISK MITIGATION, AND SECURE DATA COMMUNICATION IN NETWORKED SYSTEMS

Abstract

The exponential growth of networked systems, encompassing cloud infrastructure, Internet of Things (IoT) deployments, and edge computing architectures, has created an expansive and highly heterogeneous attack surface that adversaries increasingly exploit through sophisticated, multi-vector intrusion campaigns. Traditional intrusion detection systems (IDS), which rely on static rule sets and signature-based detection, have proven fundamentally inadequate in adapting to the dynamic threat landscape characteristic of modern networked environments. The convergence of high-speed data communication with distributed computing paradigms has further complicated the enforcement of consistent security policies, creating critical gaps in threat visibility and response capability that demand novel, adaptive security architectures. This research addresses four principal challenges confronting contemporary cyber-security practitioners: (1) the high false positive rates that undermine operational efficiency of deployed IDS solutions; (2) the latency constraints that preclude real-time threat response in high-throughput network environments; (3) the inability of static risk mitigation strategies to respond adaptively to evolving attack vectors; and (4) the substantial computational overhead imposed by encryption and integrity verification mechanisms on secure data communication channels. We propose a Cyber-Resilient Computing Model (CRCM) that integrates a deep transfer learning-based intrusion detection engine, a dynamic risk scoring and mitigation framework, and an optimised secure communication layer. The CRCM employs three novel algorithms: the Adaptive Deep Threat Classification Algorithm (ADTCA), the Dynamic Risk Scoring and Mitigation Algorithm (DRSMA), and the Lightweight Authenticated Encryption Protocol (LAEP). The architecture is implemented across five functional layers and evaluated on the NSL-KDD, CICIDS-2017, and UNSW-NB15 benchmark datasets. Experimental evaluation demonstrated that CRCM achieves 98.7% intrusion detection accuracy, reduces false positive rates to 0.9% at the optimal threshold, maintains detection latency below 11.4 ms at 5,000 kpps packet rates, and sustains system throughput of 8.0 Gbps under 10,000 concurrent connections. Risk mitigation scores exceeded 94.3% across all eight evaluated attack categories, and secure communication overhead was reduced to 11.2% even at 5,000 KB message sizes. The proposed CRCM framework delivers demonstrably superior performance across all evaluated security dimensions compared to existing methods, establishing a robust, scalable, and computationally efficient foundation for next-generation cyber-resilient networked system protection.