Network intrusion detection using ACO-DNN model with DVFS based energy optimization in cloud framework

Abstract

Recent technologies and innovations have encouraged users to adopt cloud-based environment. Network intrusion detection (NID) is an important method for network system administrators to detect various security holes. The performance of traditional NID methods can be affected when unknown or new attacks are detected. For instance, existing intrusion detection systems may have overfitting, low classification accuracy, and high false positive rate (FPR) when faced with significantly large volume and variety of network data. For that reason, this system has been agreed by many establishments to allure the users with its suitable features. Because of its design, it is exposed to malicious attacks. An Intrusion Detection System (IDS) is required to handle these issues which can detect such attacks accurately in a cloud environment. To analyze the IDS datasets some of the predominant choices are Deep learning and Machine learning (ML) algorithms. By adopting nature-inspired algorithms, the problems concerning the data quality and the usage of high-dimensional data can be managed. In this study the datasets KDD Cup 99 and NSL-KDD are used. The dataset is cleaned using the min-max normalization technique and it is processed using the 1-N encoding approach for achieving homogeneity. Dimensionality reduction is done using the Ant colony optimization (ACO) algorithm and further processing is done using the deep neural network (DNN). To minimize the energy consumption we have adopted the Dynamic Voltage and Frequency Scaling (DVFS) mechanism to the system. The main reason to set up this concept is to develop a balance between the energy consumption and the time of different modes of VMs or hosts. The proposed model is validated and compared with ACO and Principal component analysis (PCA)-based (Naïve Bayes) NB models, the experimental outcomes proved the superiority of the ACO-DNN model over the existing methods.