Genetic Algorithm Based Feature Selection Algorithm for Effective Intrusion Detection in Cloud Networks

Abstract

Cloud computing is expected to provide ondemand, agile, and elastic services. Cloud networking extends cloud computing by providing virtualized networking functionalities and allows various optimizations, for example to reduce latency while increasing flexibility in the placement, movement, and interconnection of these virtual resources. However, this approach introduces new security challenges. In this paper, we propose a new intrusion detection model in which we combine a newly proposed genetic based feature selection algorithm and an existing Fuzzy Support Vector Machines (SVM) for effective classification as a solution. The feature selection reduces the number of features by removing unimportant features, hence reducing runtime. Moreover, when the Fuzzy SVM classifier is used with the reduced feature set, it improves the detection accuracy. Experimental results of the proposed combination of feature selection and classification model detects anomalies with a low false alarm rate and a high detection rate when tested with the KDD Cup 99 data set.

A feasible approach to intrusion detection in virtual network layer of Cloud computing

“…For analysing the overall performance of Hybrid-NIDS, we have derived weighted average for the applied tests. Weighted average can be derived by the relation In table 7, performances of recent NIDS approaches from [44][45][46][47][48][49] are compared to that of Hybrid-NIDS. These approaches use soft computing techniques for intrusion detection, and they are evaluated on KDD dataset.…”

Intrusion Detection Analytics: A Comprehensive Survey

“…With hybrid IDS , first, known attacks are detected and separated by matching with the attack signatures and then among the remaining stream the unknown attacks are detected by observing deviations of such packet features from those of the normal packets. [9], DT [25], Naïve Bayes [12], SVM [21], RF [30], NN [26], etc. are mostly used by many researchers for the construction of IDS.…”

“…Dimensionality reduction is the process of transforming original raw features into the new dimensions (reduced set of features) by preserving the originality of the data. Correlation Feature Selection (CFS) [12], [23] and Principal Co mponent Analysis (PCA) [15], [19], [23] are the examp les of Dimensionality reduction techniques.…”

“…Desale et. al in [12] focused on feature selection technique for selecting best features which will helps in the construction of a good model for IDS. They have used CFS technique by using mathematical intersection principle based Genetic Algorithm (GA ) as a heuristic search algorith m. Features selected after CFS using the GA method are the intersection of co mbinations of population size and Generat ions.…”

Applying bounded fuzzy possibilistic method on critical objects

“…Outliers are important because they might change the behaviour of the model as they are far from the discovered patterns and are mostly known as noise or exceptions. Outliers are useful in some applications such as fraud and anomaly detection approaches as the rare cases are more interesting than the normal cases [66]. Outlier analysis is another concept of learning and mining approaches known as outlier mining [67].…”