Intrusion Detection System using Multi-Layer Perceptron with Grid Search CV

In today’s life all the organization over the globe are facing a major issue with security’s most common challenging issue of intrusion into their network. This intrusion in the network may lead to security concerns hampering the organizations integrity, confidentiality and availability. To solve this issue there are multiple tools in the market which detects the intrusion in a networkby surveillance of network activities and block the unusual activity detected. These tools and technologies monitor the network for sudden change in activity or behavior and processing them further for analyzing if unusual activity is noticed and inform the administrator about the change in behavior of network.Most of these tool uses the traditional machine learning method for intrusion classification into ‘good’ or ‘bad’ network. In this paper we propose a deep learning model whose architecture compromises of Multi-Layer Perceptron used for intrusion classification and uses GridSearchCV to automate the best model selection for the problem. Using deep learning to solve the problem of intrusion detection in an organization by classification of network has numerous advantages as deep learning performs well on large datasets, unstructured data, better self-learning capabilities, cost effective and scalable. In the implementation of the proposed architecture, we have achieved an accuracy of 98.10% for binary classification and 97.62% for multi-class classification.For hyperparameter tuning as we have used GridSearchCV and used five k-fold cross validation for evaluating the best performing model.