Navigating Autonomy: Unveiling Security Experts' Perspectives on Augmented Intelligence in Cybersecurity

The rapidly evolving cybersecurity threat landscape and shortage of skilled professionals are amplifying the need for technical support. AI tools offer great opportunities to support security experts by augmenting their intelligence and allowing them to focus on their unique human skills and expertise. For the successful design of AI tools and expert-AI interfaces, however, it is essential to understand the specialised security-critical context and the experts' requirements. To this end, 27 in-depth interviews with security experts, mostly in high-level managerial roles, were conducted and analysed using a grounded theory approach. The interviews showed that experts assigned tasks to AI, humans, or the human-AI team according to the skills they attributed to them. However, deciding how autonomously an AI tool should be able to perform tasks is a challenge that requires experts to weigh up factors such as trust, type of task, benefits, and risks. The resulting decision framework enhances understanding of the interplay between trust in AI, especially influenced by its transparency, and different levels of autonomy. As these factors affect the adoption of AI and the success of expert-AI collaboration in cybersecurity, it is important to further investigate them in the context of experts' AI-related decision-making processes.