The rapid advancement of technologies such as artificial intelligence has led to increasing complexity in human computer fusion systems, which poses significant challenges to the security of these systems. Recent developments in hardware, software, and algorithms have exacerbated the security landscape of human computer fusion complex systems across multiple dimensions, underscoring the need for a comprehensive review of attack and defense technologies in this domain. In this paper, we systematically review security issues in human computer fusion complex systems from various perspectives, with the aim of summarizing the current state of the field and encouraging further exploration by researchers. Specifically, our review is organized into the following key areas based on the security protection targets: (1) Human-machine interaction safety. We explore methods for measuring human-machine trust and discuss strategies for its repair and recalibration in dynamic contexts. (2) Human-computer collaborative safety. We discuss potential issues arising during human-device interactions and communication, alongside security challenges of intelligent algorithms and data privacy. Following a review of representative works, we discuss the experimental findings. Finally, we summarize the challenges in each area and point out some promising directions.