Deep neural networks (DNNs) face a growing threat from backdoor attacks, which embed hidden malicious functionalities triggered by specific inputs. This survey examines the escalating arms race between backdoor defenses and increasingly evolving adaptive attacks. We explore how attackers exploit vulnerabilities in existing defenses, highlighting a concerning asymmetry: current defenses often lag behind the adaptability of attack strategies. Through systematic categorization and analysis of both defense and attack methods, we illuminate this complex interplay and emphasize the urgent need for a paradigm shift towards proactive defense mechanisms that anticipate and counteract the evolving nature of backdoor attacks.