DOI: https://doi.org/10.36347/sjet.2026.v14i02.001

Pages: 76-86

The increasing complexity of cyber threats and the limitations of traditional cybersecurity systems have spurred the development of more adaptive and intelligent approaches. Decision-centric cybersecurity, which integrates Human-in-the-Loop (HITL) systems with machine learning (ML), has emerged as a promising solution. This review explores the role of HITL in machine learning models for cybersecurity, emphasizing the importance of combining the speed and scalability of automation with the contextual judgment and ethical considerations provided by human experts. The review covers the types of machine learning techniques commonly used in cybersecurity, such as supervised, unsupervised, and reinforcement learning, and discusses their strengths and weaknesses in addressing modern cyber threats. We also examine the challenges of integrating HITL into cybersecurity systems, including human error, scalability issues, and ethical concerns. The future of decision-centric cybersecurity lies in enhancing machine learning algorithms, improving explainability, and developing more autonomous systems, while still maintaining the crucial role of human oversight. Ultimately, this review highlights the collaborative potential of human expertise and machine learning in creating more effective, adaptable, and ethical cybersecurity defences in the face of evolving digital threats.