DRCNN-IDS Approach for Intelligent Intrusion Detection System

In recent times, deep learning technique turns to be an interesting topic, which is extensively utilized to construct an Intrusion Detection System (IDS) for identifying and classifying web attacks at network level in a periodic manner. Moreover, numerous challenges as malicious attack change its characteristics constantly. There are diverse malware datasets that are available publicly. In this investigation, Convolutional Neural Network (CNN), which a kind of Deep learning is explored along with dimensionality reduction for Intrusion Detection (DRCNN-IDS). This approach is used to identify and classify unpredictable attacks in networks. Due to the constantly changing nature of attacks, evaluation of benchmark dataset is also essential. A conceptual evaluation of this investigation with DRCNN-IDS is compared with Machine learning approaches. The irrelevant features of attacks are eliminated for reducing the network dimensionality for the computation of accuracy. The reduced features are extracted automatically by the anticipated CNN model, which is a preliminary advantage of using deep learning concepts. This work utilizes KDD cup 99' dataset for accuracy computation. Simulation was carried out in MATLAB environment and the performance is compared with prevailing Machine Learning approaches. The anticipated model shows enhanced performance in terms of accuracy. Comparison is made with DNN, RNN, RF, SVM, Naive Bayes and so on. Accuracy attained with simulation is 96%.