Augmented Intelligence Framework for Human–Artificial Intelligence Teaming in Cybersecurity

Augmented Intelligence Framework for Human–Artificial Intelligence Teaming in Cybersecurity

연구 분야: Safety

학회: Human-Centric Intelligent Systems

As cyberattacks grow more complex and frequent, organisations and nations face critical challenges in safeguarding their information systems and sensitive data. Recognising the limitations of traditional, solely human-centric defences, there is increasing agreement among practitioners and researchers on the need for a collaborative approach that integrates human intelligence with artificial intelligence (AI). This paper introduces the cybersecurity Augmented Intelligence Framework (cAIF), a conceptual framework designed to optimise human-AI teaming (HAIT) in cybersecurity. Augmented intelligence is about the role of AI enhancing rather than replacing human intelligence through a more harmonious working relationship. The methods followed consist of three phases. First, reviewing existing literature to identify foundational human–machine interaction (HMI) paradigms. A systematic review of papers from three databases led to a final selection of 20 analysis units. Second, the strengths and weaknesses of the identified paradigms for HAIT were evaluated. Lastly, outlining the core architectural components of the cAIF from the strengths of each paradigm. Six key HMI paradigms were identified: Human-in-the-loop (HITL), Human-out-of-the-loop (HOOTL), Human-on-the-loop (HOTL), Human-alongside-the-loop, Human-in-command, and Coactive Systems. Each paradigm offers unique strengths: for instance, HITL emphasises active and direct human intervention, while HOOTL supports full autonomy of AI operations. On the other hand, HOTL balances AI autonomy with human oversight. The analysed data suggests that strategically leveraging the strength of each paradigm allows for a hybrid intelligent framework comprising five core components: the Decision-Making Matrix, Paradigm Allocation Engine, Task-Specific Modules, Feedback and Learning System, and Interoperability Framework. The cAIF shows promise in enhancing human-AI collaboration, integrating human insights with AI capabilities to improve resilience and adaptability against evolving cyber threats. Future research should focus on empirically validating the cAIF in various cybersecurity domains, including healthcare and finance.