Artificial intelligence (AI) technologies offer new opportunities to detect and address unknown cybersecurity threats. This study aims to examine how AI augments human expertise in safeguarding information systems using an interactive AI conversation.
The authors use the retrieval augmented generation approach for enhancing cyber intelligence driven by its ability to provide accurate, timely and contextually relevant responses. The authors have built a working application based on real-life security data sources and tested it through constructed expert-AI conversations. It enables users to ask “How?” about the data sources informing each AI-generated response.
The authors have simulated full advisory sessions where the AI guided the cybersecurity human expert step by step in assessing the threat landscape, generating coherent and accurate outputs validated by external experts. The authors recommend the optimal AI model type and its fine-tuned parameters.
This study lays the groundwork for further research, allowing for refinement by integrating additional resources, such as near-real-time alerts from external or internal sources and other AI models.
The AI-powered advisor significantly streamlines threat detection and mitigation to help organizations’ experts respond more efficiently to attacks.
Using cybersecurity, an AI advisor can lead to a safer digital environment for both organizations and society, given the pervasive nature of AI in everyday life.
Unlike previous studies, the authors used the Rumsfeld matrix to frame questions within cybersecurity uncertainty domains, simulating AI responses and reducing barriers. This approach reveals crucial model characteristics and fine-tuning parameters that significantly enhance AI outcomes.
