DeepCyber-IDS: A Deep Learning Based Intrusion Detection System

Due to the rise in advanced and more frequent cyberattacks, traditional IDS systems are unable to identify new threats and can create a lot of false alarms. In this paper, a new IDS called DeepCyber-IDS is discussed, which addresses these issues by using deep learning to spot complex network traffic patterns and require less fine-tuning or creation of new features. The system combines CNN, GRU, and LSTM layers to draw useful information from the raw traffic data in terms of space and time. Unlike legacy IDS systems, DeepCyber-IDS learns on its own from unrefined input, meaning it can fit into various network situations without needing to be guided by outside features. It was tested on three commonly used datasets to see how well it performed. NSL-KDD, UNSW-NB15, and SmartGrid. The model got a perfect F1-Score and a 100 % precision and recall rate on UNSW-NB15 and SmartGrid, and an accuracy score of 99.84 % on NSL-KDD. The processing times varied from 469 milliseconds to about 53 seconds, meaning it can be used in real-time monitoring situations. They reveal that DeepCyber-IDS can protect against a wide array of threats and is also capable of being easily improved over time.