Hybrid Intrusion Detection System Combining Sparse Autoencoder and Deep Neural Network

Abstract

Networks are secure from new threats when Intrusion Detection Systems (IDS) are used. When you combine sparse autoencoders with Deep Neural Networks (DNN), you get a better mixed intrusion detection system (IDS) that can find attacks more accurately and quickly. The sparse autoencoder is used in unsupervised feature extraction to find hidden patterns in high-dimensional data while lowering noise and repetition. After that, the traits that were retrieved are fed into a DNN. This network then works as a classifier to find actions that are bad. For this method, the best parts of both are used: strong representation learning from sparse autoencoders and deep neural networks' strong decision-making skills. When tested on benchmark datasets, the system finds both known and unknown threats more accurately, more often, and with a higher total detection rate. The system is a good choice for today's network security issues because it can be scaled up and changed to follow new attack paths. This combination method makes IDS technology better by using smart feature extraction and classification.