In the modern day where the world in evolving and getting more interconnected bringing to the forefront the importance of network security against cybercrimes. Human beings have always lived trying to seek protection to unusual behavior, which is why Intrusion Detection Systems (IDS) are of utmost importance. Defensive measures like these, however, are mostly effective when they utilize attack signatures that are significant in defining most of the attacks, making them obsolete to new threats. The paper's goal is to demonstrate how it is possible to use deep learning (DL) and machine learning (ML) to improve establishments’ laws, particularly intrusion detection. We'll find such works and analyze the applications of these methods in IDS, determining pros and cons. At the same time, we will evaluate such important aspects as the selection of data, how much training is needed, and whether it is possible to use the developed system in real time. In general, this paper gives a promising outlook on using deep learning and machine learning technologies for increasing the efficiency and accuracy of intrusion detection systems under zero-day attack circumstances.