An Agent-Orchestrated Framework for Adaptive Cybersecurity Intelligence

Grade Level at Time of Presentation

Senior

Major

Computer Science, Computer Information Technology

Minor

Mathematics, Computer Science

Institution 25-26

Northern Kentucky University

KY House District #

Campbell County

KY Senate District #

Campbell

Department

School of Computing and Analytics

Abstract

Human behavior remains a critical factor in cybersecurity vulnerabilities, requiring security awareness solutions that are adaptive, personalized, and continuously evolving. This work presents a multi-agentic framework for enhancing organizational security awareness through distributed and decentralized decision-making. The framework is organized as an interactive, multi-stage input–process–output (IPO) workflow that orchestrates logic across specialized agents. These agents are responsible for data processing, machine learning and Large Language Model (LLM)–based analysis, output generation, and feedback processing. Decomposing security awareness into coordinated agent-level functions enables scalable and context-aware adaptation to individual user behavior. Rather than relying on static training models, the system employs a continuous feedback loop that dynamically adapts awareness strategies through diverse delivery mechanisms, enabling intelligent decision-making. A proof-of-concept prototype is being used to evaluate the framework design towards feasibility and performance of the design, demonstrating the practicality for deployment within organizational environments.

This document is currently not available here.

Share

COinS
 

An Agent-Orchestrated Framework for Adaptive Cybersecurity Intelligence

Human behavior remains a critical factor in cybersecurity vulnerabilities, requiring security awareness solutions that are adaptive, personalized, and continuously evolving. This work presents a multi-agentic framework for enhancing organizational security awareness through distributed and decentralized decision-making. The framework is organized as an interactive, multi-stage input–process–output (IPO) workflow that orchestrates logic across specialized agents. These agents are responsible for data processing, machine learning and Large Language Model (LLM)–based analysis, output generation, and feedback processing. Decomposing security awareness into coordinated agent-level functions enables scalable and context-aware adaptation to individual user behavior. Rather than relying on static training models, the system employs a continuous feedback loop that dynamically adapts awareness strategies through diverse delivery mechanisms, enabling intelligent decision-making. A proof-of-concept prototype is being used to evaluate the framework design towards feasibility and performance of the design, demonstrating the practicality for deployment within organizational environments.