With more people using the internet, there are more instances of cyberattacks, when a person may be subjected to threats, extortion, or harassment. The attack could take the form of psychological pressure or the theft of a person's password. The development of intrusion detection systems (IDS) for identifying and categorising both network-level and host-level cyberattacks frequently makes use of machine learning techniques. Due to the lack of a thorough evaluation of the effectiveness of numerous machine learning algorithms in the current approaches. A kind of deep learning model called Deep Neural Network (DNN) is being researched to find and categorise unanticipated and unplanned cyberattacks in order to develop an effective IDS. This kind of research makes it easier to choose the optimal algorithm for reliably identifying upcoming cyberattacks. Extensive experimental testing has shown that DNNs perform better than conventional machine learning classifiers. This paper suggests a framework for hybrid DNNs that is very scalable and can track network traffic in real time.