Hybrid Deep Learning Model for Automatic Intrusion Detection System

Network attacks constitute the most ubiquitous and pivotal problem confronting modern society. All networks are susceptible to network threats to some degree. Intrusion Detection is imperative to identify these threats. Although deep learning and machine learning are applied in a variety of sectors to prevent attacks, malicious threats continue to be on the ascent, necessitating the implementation of advanced security solutions. The software's purpose is to create an extensive and reliable profound learning Intrusion Detection System in a supply region with a significant allotment of information as well as figuring assets. Because of frequent changes in IP, databases must be modified systematically on a regular basis. The proposed solution advocates the employment of Convolutional Neural Networks to develop an Intrusion Detection framework whose foundation is deep learning., which will predict and categorize hostile cyber-attacks. A raw dataset can be used by neural networks to extract signatures and patterns to anticipate the characteristic and classification of future data at a faster rate. The Automatic Intrusion Detection System (IDS) is developed using the CNN, CNN-BI LSTM and RNN-LSTM architecture, for latent feature abstraction, memory retention, and categorization skills.From the analysis further the paper states that the CNN-BI LSTM model has a better accuracy as compared to both RNN-LSTM,CNN models i.e the former performs better with an accuracy of 98.72%.Utilizing the exchange learning procedure, this proposition exhibits that powerful profound learning-based IDS frameworks might be executed on certifiable gadgets with less assets while safeguarding economy and speed.