Machine Learning Driven Trust and Reliability In Human-AI Collaboration For Cybersecurity

As cyber threats become more sophisticated and deceptive, traditional cybersecurity measures using human models are no longer sufficient. Hence, a combination of human and AI-driven automated strategies is indispensable for effective cybersecurity. This research explores the challenges of HumanAI collaborations in cybersecurity by focusing on trust and reliability. It proposes a comprehensive methodology to analyze a synthetic dataset of 500 observations for determining trust level, reliability score, and success rate of Human-AI collaboration. The analysis employed several Machine-Learning models, including Logistic Regression and Random-Forest; and Natural Language Processing (NLP) techniques, along with Latent Dirichlet Allocation (LDA). The encouraging findings indicate that a medium trust level of 36 % can result in a 70 % success rate of Human-AI collaboration. Furthermore, feature importance analysis highlights that reliability scores, explainable AI (XAI), continuous model validation, and robust feedback loops will build a more effective and trustworthy Human-AI collaboration in future cybersecurity.