ISCAP Proceedings: Abstract Presentation
From Technical Capabilities to Organizational Outcomes: A Systematic Review of Generative AI in Cybersecurity
Kamal Wada Ringim Sabo
Morgan State University
Dr. Abirami Radhakrishnan
Morgan State University
Jigish Zaveri
Morgan State University
Dr. Dessa David
Morgan State University
Dr. Farzad Moazzami
Morgan State University
Abstract
Despite rapid investment in GenAI for cybersecurity, the organizational conditions that determine whether technical capabilities translate into sustained security improvements remain largely unexplored in the existing literature. This study addresses that gap through a systematic review of 71 peer-reviewed studies published between 2020-2026. Covering GenAI applications across nine cybersecurity domains: intrusion detection, malware analysis, cyber threat intelligence, vulnerability detection, phishing and social engineering defense, synthetic data augmentation, cross-domain AI frameworks, security operations, and emerging security areas.
The review yields two core findings. First, GenAI demonstrates measurable technical gains across these domains. GAN-based models mitigate class imbalance in intrusion detection, producing botnet detection improvements of up to 29 percentage points in controlled evaluations. LLM-based systems achieve vulnerability-detection accuracy as high as 94%, expand the speed and breadth of threat-intelligence synthesis, and enhance security operations center (SOC) workflows. Second, despite this strong body of technical evidence, literature evaluates algorithmic performance while providing little explanation of the organizational mechanisms through which these capabilities translate into sustained security outcomes. Most evaluations occur in laboratory or simulated environments, and reported benefits rely primarily on machine-learning metrics rather than organizational measures such as incident-response performance or sustained risk reduction.
Taken together, these findings show that the field has established a strong technical case for what GenAI can do but has not yet explained why organizations with similar GenAI capabilities may experience different security outcomes. This gap suggests the need for research that moves beyond assessing GenAI’s technical performance to examining the organizational processes that shape how AI generated knowledge is translated into sustained security and privacy improvements. We propose absorptive capacity as a theoretically grounded mechanism for explaining this translation.