Software-Defined Networking (SDN) has manifested both its bright and dark sides so far. On the one hand, it has been advocated by research communities and industry for its open nature and programmability. Every stakeholder, such as researcher, practitioner, and developer, can design an innovative networking service with a rich set of APIs and a global network view by escaping from the vendor-dependent control plane. On the other hand, its new architecture has introduced many security challenges that did not exist in the legacy environment. However, while new attacks and vulnerabilities within SDN have been steadily discovered, fewer efforts have been made to systematize the vulnerabilities from security aspects. In this paper, we aim to scrutinize prior literature that disclosed attack cases in SDN from an architectural perspective through identifying their root causes, penetration routes, and outcomes. Then, we conduct an in-depth yet comprehensive discussion of their underlying problems and introduce countermeasures proposed by researchers to mitigate those attacks. We believe that this study can contribute to revisiting various security problems around the current SDN architecture and envisioning a guideline for security research for SDN in the future.