Exploring High Performance AI Architectures for Cyber Attack Detection
Guardado en:
| Publicado en: | ProQuest Dissertations and Theses (2025) |
|---|---|
| Autor principal: | |
| Publicado: |
ProQuest Dissertations & Theses
|
| Materias: | |
| Acceso en línea: | Citation/Abstract Full Text - PDF |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Resumen: | Static, non-adaptive cyber resilience algorithms are ineffective against the rapidly evolving risks posed by modern cyberattacks in the defense sector. This disparity highlights the critical need for sophisticated, real-time adaptive systems and predictive solutions capable of recognizing and responding to both known and unknown cyber-attacks. Recent advancements in Network Optimized Architecture (NOA) and Artificial Intelligence (AI) present significant potential for tackling complex cybersecurity challenges. Reinforcement Learning (RL) has demonstrated efficacy in the progressive development and refinement of cyber-attack detection neural network classifiers, enabling systems to adapt to new attack-related information, vulnerabilities, and evolving threats. This study seeks to reduce the necessity for frequent retraining by employing reinforcement learning (RL), facilitating a continuous learning approach for cyber resilience. This approach integrates reinforcement learning into neural network augmentation for automatic tuning and introspection in a novel manner. The proposed approach emphasizes the continuous adaptation of reinforcement learning, facilitating real-time and robust defense against evolving and complex cyber threats. A framework for learning cyber resilience policies and integrating abstract attack problems is presented. This improves defense companies' capacity to counter unexpected and novel cyber-attacks by enabling the transfer of learned policies to address previously unencountered threats. To address the evolving landscape of cyber threats and vulnerabilities, this praxis incorporates the dynamic development of diverse network architectures to manage the rapidly changing attack surface. The growing volume of information regarding global cyberattacks, vulnerabilities, and emerging trends underscores the necessity of developing solutions that can adapt incrementally to the dynamic cyber threat landscape, particularly within the intricate realm of the defense industry. In many cases, traditional algorithms fail to address this level of complexity. This study examines a reinforcement learning-based policy aimed at developing an efficient neural network architecture for cyber-attack detection, specifically tailored for classification and regression tasks. This study utilizes the findings of Vadhera (2023) as a foundation for analyzing cyber-attack data and vulnerabilities pertinent to defense. This praxis outlines a cyber-resilience framework designed to establish an optimal attack detection network architecture for addressing unexpected attacks. This approach facilitates enhanced cyber defense through the continuous improvement of neural network models. This will ultimately enable the management of both existing and emerging threats within a constantly evolving threat landscape. |
|---|---|
| ISBN: | 9798290939247 |
| Fuente: | ProQuest Dissertations & Theses Global |