In recent years, the massive proliferation of Unmanned Aerial Systems (UAS), also known as Remotely Piloted Aircraft Systems (RPAS) or Drones in both military and civilian fields, has been fueled by significant advancements in electronics, energy, communication, artificial intelligence, materials science, and regulatory frameworks. While these innovations have enhanced capabilities such as autonomy, efficiency, and accessibility of drones, they have also raised security and privacy concerns that need to be identified, assessed, prioritized, and mitigated. UAS, which are classified as Cyber-Physical Systems (CPS), present unique security challenges due to their complex integration of cyber and physical processes. Thus, defining a well-structured architecture of the system that represents all the processes within it, along with the interactions between them, is the first step toward effective security analysis and threat modeling. Through a detailed examination of existing CPS and UAS architectures, we suggest that these representations cannot fully capture the distinguishing features and specifics of UAS from a CPS perspective. In this work, we propose a novel multi-layer architecture for UAS from a CPS perspective. We then explore the multifaceted security challenges associated with these systems, emphasizing the necessity for robust solutions to ensure their safe and secure deployment in diverse environments. Our discussion includes how this architecture can enhance the security and reliability of UAS by facilitating better threat identification, assessment, and mitigation.