With the establishment of the internet to the widespread application industries such as e-commerce, online banking, healthcare system, and other day-to-day essential activities include gaming, shopping, chatting, incite systems above, and activities exposed to various threats and attacks, which are increasing tremendously. A zero-day attack is the specific threat of an unknown security vulnerability in the system, and detecting such attacks is complicated due to the unknown historical attack signatures.
   In cyber security, intrusion detection systems are utilized to identify the zero-day cyber attack. Conventional machine learning techniques used for intrusion detection are failed to recognize new, zero-day attacks and own the limitation of high false-negative rates, thus restricting them to employ in real-world applications. Deep learning models are introduced in intrusion detection for detecting zero-day cyber-attacks with high detection accuracy.
   Deep neural network-based zero-day attack detection overcomes issues faced by conventional methods for the intrusion detection system. Recent advancement in zero-day attack detection system is incorporating federated deep learning in order to facilitate data privacy issues.